Real Research on Artificial Intelligence is an NRG newsletter covering how people think about, feel about, and use AI, featuring our original survey data. Subscribe for weekly updates, and reach out to david.byler@nrgmr.com with questions.

When X was Twitter, political journalists often quipped, “Twitter is not real life.” Their meaning: the opinions expressed by news-obsessed, always-online Twitter users were nothing like the views of the broader public.

There’s no single, Twitter-like hub for AI discourse. But the most AI-savvy slice of the white- collar workforce has a similar issue. They’re spinning up armies of AI agents, obsessing over benchmarks, and founding one-person startups worth a small fortune, while the public is still learning the basics of the technology. More specifically, our data shows that many Americans don’t understand AI and use it like Google search. And that may be part of the reason societal feelings about AI are significantly worse than personal feelings about the technology.

TOPLINE: AMERICANS DON’T UNDERSTAND HOW AI WORKS

Americans are adopting AI at a breakneck pace. But they don’t know how it works. NRG recently asked a sample of 1,009 Americans aged 18-64 a multiple-choice question about what AI does. Here are the results:

This question was inspired by a similar one asked by the Democratic firm Tavern Research in August 2025. Their sample, question design and wording were slightly different. But the takeaway was the same: a 45% plurality said that AI was looking up the answer in a database, and only 28% said that AI was predicting subsequent words based on past patterns (the correct answer).

• More context: Quiz-like questions like this, while useful, don’t capture the full range of knowledge that Americans have. For example, many Americans can’t correctly define the term “hallucination,” but most believe that AI-generated text is not always correct. People can understand a concept and still miss a multiple choice question.

• But the point stands: AI discussions in Silicon Valley, corporate boardrooms, and newsrooms are likely dramatically more technical than those of everyday Americans.

GOING DEEPER: FOR MANY AMERICANS, AI IS A GREAT SEARCH ENGINE

• The Topline: The same survey also asked how respondents used AI. And the most popular applications sounded a lot like how people use and relate to search.

• Data Points: Three of the five top use cases are, essentially, search.

• 63% are using it to ask simple questions

• The next most popular use, at 48%, was to learn about a new topic

• 38% reported using AI to research products

• And the most sophisticated, powerful uses–coding and automating tasks–were two of the three least popular.

• What it means: These results aren’t a precise, direct measure of AI use. Some people might overreport impressive-sounding technical uses or underreport seemingly basic uses like search. AI companies can (and do) pull user prompts, categorize them, and get a more accurate outlook on what people do. Instead, this question measures how people think they use AI and what they remember doing. And the top use, by far, is search-like information retrieval and summary.

GOING BROADER: HOW SOME AMERICANS (MIGHT) PROCESS THE COSTS AND BENEFITS OF AI

• The era of bad feelings: Americans are becoming more pessimistic about the societal and political impacts of AI, according to a raft of new polls from other firms. In particular, NBC reports only 26% of Americans have a favorable view of AI, while 46% hold an unfavorable view, for a net -20% favorability score. That makes AI significantly more popular than Iran (-53%) but less popular than Immigration and Customs Enforcement (ICE) (-18%).

• Many explanations are in the mix: There isn’t one single explanation for these numbers. Some people might blame AI for a tough job market. Others might dislike some of the AI-generated media they’ve consumed. And some truly do worry that a superpowered AI could threaten the survival of the human race.

• One theory from this data: the cost-benefit analysis of AI isn’t working for everyone. There isn’t enough data in this survey to fully prove this idea. But it’s possible that for many Americans, the upsides aren’t worth the risks. For them, AI feels like a great search engine. But then, on the news, they hear that it’s about to vaporize their job and potentially destroy mankind. That might feel like a bad tradeoff.

• There’s a huge opportunity for tech companies in this data. The good news for tech companies: AI can already do much more than search. If tech companies can find ways to reach ordinary consumers and help them level up their AI usage, they could convert users into superusers and, for many, add benefit to their cost-benefit analysis.

• More research is needed (and on its way!) The survey that produced these results, NRG’s AI Observatory, just finished its pilot study. Every month, AI Observatory will track how consumers feel about AI and use it, and how AI is changing society and culture. We’ll release a few more findings from the pilot study in this newsletter and use subsequent waves to regularly track sentiment around AI over time. Stay tuned.

BELOW THE RADAR

We also want to share reading recommendations from NRG’s top analysts and in-house AI experts—especially pieces we find to be ahead of the curve.

Stephanie Wong, Senior Vice President in NRG’s Devices and Services practice, recommends Is AI Going to Turn Us All into Middle Managers?

Stephanie’s take: “We talk so much about how much potential AI has for individual productivity. But not as much about how that affects company structure, what skills become valuable, and what it means when we try to remove too much human interaction on teams.”

Thanks for subscribing. We’ll have more original data on AI next week!

Methodology: AI Observatory, the source of all NRG data cited in this newsletter, is a March 2026 general population study of 1,009 Americans aged 18-64, weighted to reflect the nation’s age, racial/ethnic, gender, political, and educational makeup.

Have a question you’d like to ask the public? Email david.byler@nrgmr.com.

Want to do business with us? Reply to this email. For enterprise tech inquiries, reach out to rob.barrish@nrgmr.com. For policy or other areas, email natalie.lacey@nrgmr.com.

Real Research on Artificial Intelligence is an NRG newsletter covering how people think about, feel about, and use AI, featuring our original survey data. Subscribe for weekly updates, and reach out to david.byler@nrgmr.com with questions.

Social media and AI companies are facing increased scrutiny from courts, parents, and politicians–particularly on their impact on teen and child mental health. The public has strong pro-safety instincts on this issue as allegations mount that tech companies engineer overly addictive products, leading to mental health issues. So today, we’re using KOL, NRG’s panel of tech policy experts, to gauge whether this issue will retain momentum and suggest measures that might help kids and teens.

THE CONTEXT: POLICYMAKERS ARE PUSHING FOR TEEN/CHILD SAFETY ONLINE, AND THE PUBLIC IS ON BOARD.

Three recent events encapsulate the current public mood on social media, AI, and youth mental health:

• A California family recently accused Google and Meta of targeting teens with harmful, addictive content and products. Meta CEO Mark Zuckerberg, Instagram head Adam Mosseri, and YouTube Vice President of Engineering Cristos Goodrow all testified, and the jury held Meta and Google liable for $3 million in damages.

• Last year, Australia banned social media for anyone under age 16. The U.K. is pilot testing similar measures, as well as a digital curfew and time limits on apps.

• U.S. lawmakers are working on AI child safety laws. There is bipartisan desire in Congress to restrict minors’ access to AI companions. And in October 2025, California Gov. Gavin Newsom signed SB243, a law requiring AI companies to monitor chatbot activity for signs of suicidal ideation among young users.

Americans are receptive to pro-safety arguments.

• U.S. adults agree with the California jury that social media harms young people. We haven’t seen polling about this specific case. But, per YouGov, seven in 10 adults (69%) believe that social media has a negative impact on the mental health of Americans under age 16.

• Americans also agree with Australia’s social media ban. According to a Fox News poll, two-thirds (64%) of American voters want to ban social media for children under 16.

• And the public wants children protected when they’re using AI. A Fathom/Forbes Tate survey found that 90 percent of Americans agreed that AI products intended for children should be verified safe before release.

  • That’s a step further than Newsom was willing to go. While he signed the chatbot-monitoring bill, he vetoed a measure, backed by pro-safety groups, that would require AI companies to ensure chatbots were not “foreseeably capable” of harming children.

• What it means: The public wants aggressive action to protect children online. But the public doesn’t always get what it wants. Example: Americans want their taxes to be filed automatically, but you still have to file your own taxes next week. So we turned to our tech expert panel to help understand the outlook for continued work on this issue.

GOING DEEPER: EXPERTS EXPECT MORE LITIGATION, REGULATION

• The data: NRG recently gave 51 of our KOL tech experts a series of questions on social media, teen mental health, and regulation. When asked whether the number of lawsuits against social media companies regarding teen mental health would increase or decrease, almost all (98%) of tech policy experts said the number will increase.

• They also believe that more Australia-like laws are on the way. In the same NRG survey, three in four (75%) said they expected more regulations like Australia’s to be enacted in various regions of the world in the next 2-3 years. Another 20% said that “many countries” would adopt a similar benchmark, and only six percent said these bans would be uncommon.

• They’re less optimistic about federal action in the U.S. In our Q4 2025 survey of KOL, the panel generally believed that President Trump’s anti-regulation stance might be a barrier to federal legislation on child safety.

• The overall expectation: This issue isn’t going away. More court cases are coming. And, with major elections coming up in the U.S. and around the world, the issue of teen and child safety online will continue to be a part of day-to-day politics.

GOING FURTHER: WHAT POLICIES WOULD ACTUALLY HELP?

We also asked our experts what one policy (if implemented by all social media companies) would most improve the mental health of children and teens. Here are three quotes:

• Non-algorithmic feeds, time limits, and other in-app fixes: “There isn’t a single, silver bullet. However, the age-verified default-off algorithmic feeds for users under 18 years old should be implemented. Algorithms are the amplifier and key cause for mental health harm. Time limits, age gates and content filters are helpful but they’re downstream patches vs. algorithms, (which) are upstream cause. Raising the minimum age… is impossible to enforce without invasive identity verification.”

• A different, hands-on algorithm for kids and teens: “The algorithm showing content to children and teens would be balanced between content created by individuals and content created by established and trustworthy news and educational organizations like PBS, NBC, etc.”

• Sunlight is the best disinfectant: “I think forced external audits and algorithmic transparency rules would be the most effective and least problematic solution.” They continued that transparency would work because “because it varies the threat of lawsuits and legislation when companies don’t correct dangerous problems.”

This isn’t a comprehensive list of possible policies. And AI-related issues, such as deepfakes, will require their own solutions. But the KOLs seemed to agree that part of the solution is transparency, and part of the solution comes from changing algorithms themselves. Meaning: to solve this issue, both tech companies and the government will need to stay involved.

BELOW THE RADAR

We also want to share reading recommendations from NRG’s top analysts and in-house AI experts—especially pieces we find to be ahead of the curve.

Susan Hoxie, Senior Vice President at NRG for platforms and technology, shared this piece from Microsoft on accessibility and AI.

• Core takeaways: Over a billion people worldwide have a disability, and AI opens up new possibilities for this group. Designing for accessibility, from the start, will be critical for the future of AI.

Thanks for subscribing. We’ll have more original data on AI and tech next week!

Methodology: The Q1 KOL survey was conducted in March 2026, using NRG’s proprietary panel of tech policy experts. The total sample size was 51.

Have a question you’d like to ask the public? Email david.byler@nrgmr.com.

Want to do business with us? Reply to this email. For enterprise tech inquiries, reach out to rob.barrish@nrgmr.com. For policy or other areas, email natalie.lacey@nrgmr.com.

Real Research on Artificial Intelligence is an NRG newsletter covering how people think about, feel about, and use AI, featuring our original survey data. Subscribe for weekly updates, and reach out to david.byler@nrgmr.com with questions.

Editorial note: At the time of writing, NRG is fielding a survey of our tech policy key opinion leader panel (known as KOL) and processing a general population survey about AI usage and attitudes. Both will be featured in this newsletter in the coming weeks! But today we’re sharing original data that’s never been published on this channel before: our December 2025 survey on AI and stress. A shorter recap of that data can be found below.

Americans are stressed when they think about AI. But not when they use it. People are worried that AI might destroy jobs or end civilizations. But when they open an AI app on their computer or phone, they generally feel good.

• The topline: When AI users open an AI app, they are much more likely to experience positive feelings than negative feelings. Curiosity, happiness, inspiration, and excitement beat emotions like confusion, intimidation, and stress.

• The context: This question was asked of the 86 percent of respondents who use AI more frequently than “never.” Some respondents who never use AI may refuse to do so precisely because they feel stressed or intimidated by these tools. But it’s difficult to speculate on the emotions of “never” users, and some might enjoy these tools if they tried them.

• What it means: As adoption grows, attitudes could become more positive. When people come into contact with AI, they tend to understand why it might be useful. If AI companies want to improve their image, increased adoption is part of the solution.

GOING DEEPER: Americans were generally optimistic about AI’s impact on their personal lives.

• The topline: We also asked all respondents (those who use AI regularly and those who don’t) whether they thought AI would have a positive impact on their personal lives. In almost every case, the answer was a clear yes.

• The context: In each of these categories, AI users registered positive expectations, and AI non-users registered negative or neutral expectations. But more people have used AI apps than haven’t, so the overall numbers skew positive.

• What it means: The gap between personal and broader societal expectations showed up again. In AI polling, respondents often register real worries about how AI will impact society or the economy. But people often simultaneously see how these tools could positively impact their own lives.

GOING WIDER: Lower-income workers are most worried.

• The topline: At work, lower-income respondents were more likely to feel stress (or expect to feel stress, if AI isn’t in their workplace already) than higher-income workers.

• Why? Two theories.

  • Some of the most worried workers might be low-income, white-collar workers. Young Americans in clerical or entry-level jobs might feel susceptible to AI job loss, while their bosses (or a high-income, blue-collar worker, like the owner of a pest control business) might feel like AI could decrease their stress at work.

  • Lower-income workers also, by definition, have less financial cushion. If a person making $200K loses 20 percent of their income to AI, they’ll be frustrated. If a person making $50K loses 20 percent of their income to AI, they would likely feel real pain. In that way, AI might feel less stressful at the top of the income distribution.

• These numbers might be changing. This survey was conducted before Claude Code, OpenAI’s Codex, and other similar agentic tools saw a spike in public interest. And some recent data shows that people with incomes under $200K are becoming more worried about their job prospects. We’ll be watching these trends and publishing more as time goes on.

BELOW THE RADAR

We also want to share recommendations from NRG’s top analysts and in-house AI experts—especially pieces we find to be ahead of the curve.

From Natalie Lacey, Executive Vice President and Head of Policy: Craig Martell’s recent remarks at Axios’s AI+DC Summit.

Source: Axios

• Core insight: Martell emphasized that AI is “statistics at scale.” The steps are familiar: gather data from the past, build a model, predict the future.

Thanks for subscribing. We’ll have more original data on AI next week!

Methodology: Unless otherwise specified, data featured in this report comes from a survey of 1,000 US adults, aged 18 to 64, conducted online in December 2025. This data has been weighted to reflect national demographics, based on age, gender, and ethnicity.

Have a question you’d like to ask the public? Email david.byler@nrgmr.com.

Want to do business with us? Reply to this email. For enterprise tech inquiries, reach out to rob.barrish@nrgmr.com. For policy or other areas, email natalie.lacey@nrgmr.com.

Real Research on Artificial Intelligence is an NRG newsletter covering how people think about, feel about, and use AI, featuring our original survey data. Subscribe for weekly updates, and reach out to david.byler@nrgmr.com with questions.

Washington D.C. and Beijing agree on one thing: the U.S. and China are locked in a global AI race, and the winner will gain a significant geopolitical advantage. But if public awareness is the first round in the contest, D.C. has a problem: many Americans have no idea the race is even on.

• The topline: Only four in 10 Americans (40%) said that a Chinese victory in the global AI race would have a “very” or “somewhat” negative impact on the U.S. An only slightly smaller share (33%) said the effect would be neutral or that they didn’t know what the impact would be. And a quarter of our sample (27%) said a win for China would be positive.

• The disconnect: According to Pew, over half (56%) of Americans think China is a competitor, and another third (33%) say China is an enemy. But, according to our data, Americans don’t understand or recognize AI as a theater for the U.S.-China competition.

• Why? To gain insight, we also asked an open-ended question: “Do you think China winning the AI race would be a good or bad thing for you, personally?” A few of the camps that emerged:

  • Pro-AI hawks: A solid chunk of respondents believes that what hurts the U.S. would hurt them. As one respondent said, “It would be bad for the U.S.; therefore, it would be bad for me.”

  • Consumers and positive-sum thinkers: One respondent told us, “China produces a lot of stuff we use, and we can benefit from them.” This attitude— rather than a favorable view of the Chinese government—may be what drives some to see AI victory for China as a positive.

  • The unsure: Some of our respondents just don’t have a view. As one said: “I’m not exactly sure how it would affect me personally, as I don’t understand enough about it to make that call.”

• What it means: Anyone who sees American AI dominance as an urgent national priority needs to get to work. The issue is just not salient yet among a large swath of the public.

GOING DEEPER: A massive age difference

We also found a huge age gap: respondents under 35 were much less likely to see a win for China as a negative for the U.S.

• Almost four-in-10 respondents (39%) under age 35 said a win for China would have a positive impact, while only 27% said the impact would be negative.

  • In research-speak, younger Americans think a victory for China would be +12 net positive (39% positive - 27% negative).

• Almost half of respondents (49%) above age 35 said a win for China would have a negative impact, and only 19% said positive.

  • Research-speak again: middle-aged and older Americans see a China win as a -30% net negative (19% positive - 49% negative).

• That’s a massive gap. This survey didn’t fully explain why. We have two theories:

  • Younger voters sometimes register a sunnier view of AI than older voters. They might be more likely to see the upside of AI development, regardless of who wins the race.

  • Older and younger voters’ different life experiences shape their opinions. Across surveys, older voters tend to be more skeptical of socialism and communism than younger voters. Older voters, who remember the Cold War, might instinctively dislike the idea of losing any competition—including the global AI race—to a Communist state.

GOING FURTHER: On policy, voters prefer defense to offense

One more chart from this survey. We asked voters: “If China were to take a clear lead in the AI race, what actions, if any, would you want your government or companies in your country to take in response?”

The results:

Takeaways:

• The home front matters most. The most popular measures involved protecting U.S. citizens digitally and economically. Americans like their privacy. They want worker protections and rules for AI companies. And they want an internet free of deepfakes.

• Economic sanctions against China is the least popular policy listed. Americans aren’t ready to escalate conflict with China (this survey was fielded before the start of the war in Iran).

• The public wants the government, not companies, to act. When respondents said they favored one of the options above, they received a follow-up asking whether the government or a private company would be best positioned to take action. On every policy, the majority of respondents favored the government.

• What it means: There’s some appetite for the U.S. to protect itself, especially in the event that China starts winning the AI race. But, for now, it appears Americans aren’t looking for confrontation.

BELOW THE RADAR

We also want to share reading recommendations from NRG’s top analysts and in-house AI experts—especially pieces we find to be ahead of the curve.

From Rob Barrish, Executive Vice President and Head of Technology Strategy: Alessandro Cannas of Google summarized Anthropic’s 81,000 respondent global survey on AI hopes and fears.

• Highlights

  • The same people who voice hopes about AI, such as increased efficiency, also worry about potential negatives like job loss and the potential for AI hallucinations.

  • Users in different regions of the globe see different uses for AI (the West wants to alleviate stress, the Global South wants opportunity).

Thanks for subscribing. We’ll have more original data on AI next week!

Methodology: Unless otherwise stated, data comes from an NRG survey of 1000 U.S. adults ages 18-64, weighted to be representative of that population by age, gender, race/ethnicity, education, region, and party. Data was collected via online panel, Feb. 10-12, 2026. Margin of sampling error is plus or minus three percentage points.

Have a question you’d like to ask the public? Email david.byler@nrgmr.com.

Want to do business with us? Reply to this email. For enterprise tech inquiries, reach out to rob.barrish@nrgmr.com. For policy or other areas, email natalie.lacey@nrgmr.com.

National Research Group (NRG) regularly interviews members of our Key Opinion Leader (KOL) panel, KOL Compass. The panel represents over 100 tech and policy insiders from Fortune 500 companies, NGOs, top law firms, academia, and more. We survey this group to obtain unique insights about the future of tech policy and publish the results. To receive more tech policy updates from NRG KOL Compass and our Public Affairs practice, subscribe here.

Tech insiders are worried about AI and mental health. Politicians from both parties are seriously discussing the impact of AI and the internet on child and teen mental health. Major AI companies have demonstrated that only a tiny fraction of AI chats contain signs of mental health emergencies—but they’re taking the problem seriously and trying to steer struggling users to helpful resources. And in our latest survey of KOL Compass, NRG’s panel of tech policy experts, 49% said “losing touch with reality or suffering from distorted thoughts or emotions as a result of prolonged exposure to AI” was a somewhat big or very big problem.

We asked KOLs what the government and tech companies could do to solve the issue. Specifically, we asked them what policies the Trump Administration would likely pursue to protect children in the context of AI and technology and what messages companies could use to build trust with parents. Our experts were pessimistic about government action. But they believe tech companies can build trust with parents by taking direct action and telling parents what they’re doing.

Most of our experts believe the Trump Administration will take, at most, minimal action on AI and mental health. Trump is generally against AI regulation, and they believe that will shape the Administration’s approach. One of our panelists, a VP at a lobbying firm, explained: “I don’t expect any [action] at all unless there is some high-profile incident that focuses attention on the matter. Essentially, there must be a political requirement that they act, one that has not happened yet.” Some KOLs disagreed, saying the Administration might push to give parents more control over content or back laws similar to the Kids Online Safety Act. But over half said that the federal government will do very little.

The KOLs believe that tech companies can change minds on this issue—if they take direct action and communicate effectively about what they’re doing. Here are the messages KOLs believe would build trust between parents and tech companies:

If tech companies act directly—never using children’s data for training, instituting direct parental controls, and ensuring outputs are age appropriate—and tell parents what they’re doing, they have a path to building trust. If tech relies on third parties (schools or non-profits) or focuses on providing resources, their efforts will seem less direct and less convincing.

Some specific recommendations from the KOLs:

• One tech consultant in our panel said “Use AI to create and maintain a perceived age of the user, to determine if they’re a child or adult, and concurrently, the level of information access that they should have. Use actual ID verification to verify the age of users. Refuse to provide any information to a verified child discussing self harm, or armed violence to an individual or group.”

• And one of our right-leaning AI experts suggested tech companies “create customized user interfaces for children that minimize exposure to harmful content. ‘Harmful’ includes seemingly benign things like advertising overly expensive things that children are likely to want their parents to buy for them. In general, targeted advertising to children using social media or the internet should be restricted.”

The bottom line: AI and mental health is a political issue. But politicians might not be able to allay worries. Tech companies themselves might need to take action and build trust to solve this issue.

The NRG KOL Compass is an integral part of our Public Affairs practice, helping businesses and organizations build better messages and communications. For more information contact our team at KOLcompass@nrgmr.com. And subscribe to the NRG Expert Panel Substack for regular KOL updates.

National Research Group (NRG) regularly interviews members of our Key Opinion Leader (KOL) panel, KOL Compass. The panel represents over 100 tech and policy insiders from Fortune 500 companies, NGOs, top law firms, academia, and more. We survey this group to obtain unique insights about the future of tech policy and publish the results. To receive more tech policy updates from NRG KOL Compass and our Public Affairs practice, subscribe here.

NRG policy team member David Byler authored a guest essay in the New York Times. The essay draws on data from NRG’s new white paper, Green Intelligence, which includes numerous quotes from KOL panelists. In the piece, he argues that Republicans have a more coherent plan on AI than Democrats now—but that Democrats have multiple paths forward on the issue.

Read the piece here.

Programming note: Next week, the KOL newsletter will once again feature original data from the latest survey of the panel. After that, the newsletter will go on break for the holidays.

The NRG KOL Compass is an integral part of our Public Affairs practice, helping businesses and organizations build better messages and communications. For more information, contact our team at KOLcompass@nrgmr.com. And subscribe to the NRG Expert Panel Substack for regular KOL updates.

National Research Group (NRG) regularly interviews members of our Key Opinion Leader (KOL) panel, KOL Compass. The panel represents over 100 tech and policy insiders from Fortune 500 companies, NGOs, top law firms, academia, and more. We survey this group to obtain unique insights about the future of tech policy and publish the results. To receive more tech policy updates from NRG KOL Compass and our Public Affairs practice, subscribe here.

NRG just released Green Intelligence, a new white paper on AI and sustainability. In it, we dig into the issues surrounding AI and sustainability, using a general population survey and targeted interviews with KOLs. The report covers:

• Whether Americans are worried about AI’s impact on the environment—and whether they’re more or less worried about other issues

• How much water LLMs consume—and how much water people think they consume

• Generational and political differences: the specific worries that Gen Z and political liberals have about AI’s impact

• Messaging: what messages about AI and the environment land with consumers, and what falls flat.

Download the full report here.

The NRG KOL Compass is an integral part of our Public Affairs practice, helping businesses and organizations build better messages and communications. For more information, contact our team at KOLcompass@nrgmr.com. And subscribe to the NRG Expert Panel Substack for regular KOL updates.

The Trump Administration loathes AI regulations. When they listed their priorities in the White House AI Action Plan, they put “remove red tape and onerous regulation” first. Populist conservatives who distrust big tech companies have lost influence in Washington. And, for the moment, Trump seems content to leave the AI industry to its devices.

But not all moments last. We asked our KOL panelists a simple question: “What warning signs or conditions might cause the government to step in more aggressively to regulate or intervene in AI and tech markets?” We expected uniform, political answers (e.g. “when Democrats retake the White House”). Instead, the KOLs talked about control. Their view: when people lose control—or even see others lose control—policy can change fast.

The most obvious way people could lose a sense of agency: losing their job to AI. As a conservative KOL working at an NGO noted, “The main warning signs are jobs and industries outside of the technology sector being significantly disrupted, which would then begin to lead to a massive amount of job loss and unemployment.” If people feel economic pain—and the shame of losing their jobs to robots—they might push harder to regulate the industry.

But some KOLs imagined situations where people felt a loss of control over the broader economy. According to one KOL, head of technology and innovation at a private-sector company, the federal government might act “when AI makes financial decisions that affect the stock market swiftly & negatively.” Multiple KOLs flagged the stock market as a danger for AI. Unwise human decisions have crashed the stock market in the past. If AI-driven decisions do the same, the public could demand regulation.

Others mentioned potential social crises alongside economic dangers. A KOL in academia said the federal government might regulate AI if there are “adverse events—fund trading errors that significantly impact Wall Street’s ability to operate, drone strikes that are unauthorized, cars that are taught to drive themselves causing accidents due to limitations with the training set of data.” The public may also prove to have low tolerance for even small mistakes made by AI – just look at the outcry after a Waymo accidentally struck and killed KitKat, a local San Francisco cat.

An expert in tech regulation listed more AI-related events that could diminish feelings of agency and bring on regulation: “increased incidence of deepfakes that are used to defraud people and spread dangerous propaganda, increased instances of AI-driven fraud, and AI models going ‘rogue’ and behaving in a manner that their makers did not envisage.” The particulars of these scenarios differ. But each one ends the same way: with humans feeling like machines have taken power from them.

For now, the Trump Administration can leave AI unregulated without paying too much of a political price. AI is still a new technology, and people don’t feel like they’re losing control over society or their own lives just yet. But if voters start to think that AI is taking choices awayin any domain, the political atmosphere could change rapidly. Tech companies, activists, politicians—anyone in the tech policy space—should be prepared in case that moment arrives.

National Research Group (NRG) regularly interviews members of our Key Opinion Leader (KOL) panel, KOL Compass. The panel represents over 100 tech and policy insiders from Fortune 500 companies, NGOs, top law firms, academia, and more. We survey this group to obtain unique insights about the future of tech policy and publish the results. To receive more tech policy updates from NRG KOL Compass and our Public Affairs practice, subscribe here.

In the June/July wave of our KOL survey, we asked our expert panelists which company was leading in the “AI race.” Several politely told us that the question was too simplistic and they had no idea what it meant to “win” an undefined competition.

So the NRG Public Affairs Team, with our egos only slightly bruised, fashioned a new concept: the AI Olympics. We broke the “AI race” into multiple events and asked the KOLs to rate whether some key players seem likely to “win” each one in the future.

The KOLs spread victories across five companies: NVIDIA, OpenAI, Amazon, Anthropic, and Google. These results are subjective. With a different list of events or different panel of experts, we might see a different lineup of leaders. But they reveal, in the KOL’s view, what each company’s unique strengths are.

OpenAI and Amazon lead in the headline-grabbing events. In a previous post, the KOLs said OpenAI had a first-mover advantage. They argued that ChatGPT caught fire before other consumer LLMs and built hard-to-overcome leads in areas like brand strength and model quality.

The three AI Olympic events where OpenAI leads—total LLM subscriptions, the race for AGI, and the ability to attract AI talent—add detail to that story. These events are analogous to swimming, track, and gymnastics at the actual Olympics—they’re highly visible to the public and are commonly associated with a broader sense of success in AI competition overall.

Similarly, Amazon is a leader in a key event: number of jobs automated. If AI is going to reach its full potential, it will likely automate (and create) some new jobs along the way. And Amazon—a tech giant that manages a massive physical supply chain and retail operation—is among the companies best suited to learn what jobs can (and can’t) be automated.

NVIDIA and Anthropic are specializing and reaping the rewards. NVIDIA and Anthropic led in categories that give them special potential for longevity. NVIDIA, unsurprisingly, leads on chip production. As long as others need GPUs to train and run AI, NVIDIA will remain critical to the economy.

Anthropic built credibility on ethics by regularly publishing their safety work and successfully elevating it in the broader media. This is a smart move. If politicians start to scrutinize AI companies on safety or transparency, Anthropic will start with more of a halo than most. And, if enterprises worry more about safety, Anthropic can use their work as a selling point.

Google is a leader in science—a less splashy but critically important area. After we asked the KOLs to rank the companies in each event, we had them rank which event was most important overall. The top choice: scientific breakthroughs.

Google should be glad it’s a leader in science. The race for “most popular chatbot” is crowded. And if a firm creates AGI or starts to kill jobs, they may see significant public backlash. But if Google can use AI to cure diseases, they’ll rake in cash and goodwill.

Conclusion: The AI race is still a good gut check. We’ll continue to poll KOLs and the general public on the “AI race” question. It’s a solid barometer for any population’s big picture feelings about the major players in AI.

But the AI Olympics reveals that choosing the right event is just as important as performing in that event. AI companies can be wildly successful without having the most total users or automating jobs away. The right way to “win” an AI race might not be to outpace ChatGPT on users or produce more chips than NVIDIA. It might be to find the event only you can win.

The NRG KOL Compass is an integral part of our Public Affairs practice, helping businesses and organizations build better messages and communications. For more information contact our team at KOLcompass@nrgmr.com. And subscribe to the NRG Expert Panel Substack for regular KOL updates.

National Research Group (NRG) regularly interviews members of our Key Opinion Leader (KOL) panel, KOL Compass. The panel represents over 100 tech and policy insiders from Fortune 500 companies, NGOs, top law firms, academia, and more. We survey this group to obtain unique insights about the future of tech policy and publish the results. To receive more tech policy updates from NRG KOL Compass and our Public Affairs practice, subscribe here.

The global AI race feels like a Rorschach Test. Some experts study the data deeply and say that the US is winning. Other experts, with the exact same data available to them, argue that China is winning. Both sides are confident. And non-experts are left to guess who is right.

We wanted to understand the true expert consensus, so we polled 53 members of KOL Compass, NRG’s panel of tech policy insiders. Eight in ten (79%) of these panelists said that the US leads China in the AI race. They also revealed why, in their view, the US is ahead and assessed whether China can catch up.

According to the KOLs, the US built its lead by designing better chips. When we asked what it would take for China to take the lead, the most common answer was that China needs a better hardware stack. As one panelist explained: “China needs to have advanced infrastructure in semiconductors, not just processors, but all the other infrastructure, such as connectivity, memory and software that extracts and runs in parallel.” Another said, “a Chinese based chip manufacturing company would need to be able to create GPUs that exceed the top benchmarks of US companies.”

A few panelists said that social advantages kept the US ahead. One KOL said that for China to pull ahead “it would require China to return to the pre-Xi regulatory model where there is a lot more tolerance of aggressive entrepreneurs.” Another noted that China’s illiberal social policies were a barrier and that “they need to earn trust with the global community”. But, generally speaking, the KOLs focused more on hardware than softer, societal factors.

The KOLs were less clear on who would win the global AI race in the long term. As one KOL said, “I think China will overtake the U.S. in the AI race. Realistically, odds are that AI growth will taper—a ‘breakthrough’ to achieve actual human intelligence is quite unlikely. The US market cannot capitalize enough to become economically profitable as it stands. Simply put, the Chinese market is bigger and better equipped to offset the labor perils of AI. The US is poorly equipped, and no one in power (and almost no one out-of-power) actually wants to course correct. It may take a few decades, but the US will lose on its current trajectory.”

Others were more bearish on China’s long-term prospects. Another said “China would need a domestic chip manufacturing capability they don’t currently have. If they can get the chip supply correct, they would be in the game, but the race is already getting close to being over (by industry estimates, at least).”

And one expert expressed the minority view—that China is already ahead and will maintain their lead. “China has an advantage because it is an authoritarian state. I’m not recommending the US follow suit, but the unfortunate reality is when a state has more control there is more investment/an advantage.”

According to the KOLs, the future of the global AI race is still a Rorschach Test. No one knows whether the US or China will win this race, or even what it means to “win.” But the present isn’t so subjective. The KOLs think the US is ahead. And the expert consensus is much clearer than the headlines imply.

The NRG KOL Compass is an integral part of our Public Affairs practice, helping businesses and organizations build better messages and communications. For more information contact our team at KOLcompass@nrgmr.com. And subscribe to the NRG Expert Panel Substack for regular KOL updates.

The KOL newsletter is part of NRG’s Public Affairs practice. Subscribe to the NRG Expert Panel Substack for general population surveys, updates from our 100+ member tech expert panel and more.

Programming note: The KOL newsletter will not be posted on Oct. 23 due to staff being out of office. Regular publication will resume on October 30.

In December 2024, NRG conducted a study exploring the “rise of the “newsfluencer.” The core findings were:

• Newsfluencers—individual creators who use social media, podcasts, and other new mediums to build audiences—are taking over the news media.

• Newsfluencers post frequently and earn credibility by building a personal relationship with their audience. Many win fans by leaning into controversies and taking a political or ideological side.

• During the 2024 election, many of the most popular —and fastest-growing—newsfluencers were conservative.

Since we published this report, the newsfluencers ecosystem has continued to grow and evolve. Here’s some of the most interesting data points we’ve seen, along with some takeaways from the KOL team.

1. The 2024 election was all about podcasts. TikTok could take over in 2028.

The 2024 election was the “podcast election.” Donald Trump toured a string of popular conservative and conservative-adjacent culture podcasts, ultimately earning the endorsements of top creators like Joe Rogan and Barstool’s Dave Portnoy. Many analysts think Trump’s podcast push was an important part of his winning strategy. But in 2028, TikTok might become king.

In 2020, one in 33 US adults received news through TikTok. Now, one in five do. This trend is moving quickly because it’s driven by changing behavior and generational replacement.

Every age group, except for the 65+ cohort, is steadily increasing its TikTok news consumption. Gen Z uses TikTok the most. But Americans aged 50-64 are now ten times more likely to see news on TikTok than they were in 2020. And the group that’s most resistant to TikTok—those aged 65 or older—are the most likely to exit the electorate.

Takeaway: Newsfluencers will be central in the 2028 election—but the mediums will change quickly. Some analysts have attributed Zohran Mamdani’s meteoric rise in the New York Mayoral race to his skill with viral vertical video. Other candidates will need to master this medium—and whatever emerges next—to reach new voters.

2. Substack is now a major player, on par with legacy outlets. And their readership is growing.

Newfluencers aren’t just using new forms. They’re dominating old ones. Earlier this year, Substack—a newsletter and video platform where independent news creators can build an audience and steady income from paid subscribers—matched The Wall Street Journal’s web traffic.

Updated data shows that Substack is still growing faster than many of its establishment rivals. And major media players have noticed. Paramount recently bought The Free Press, a conservative Substack publication, for $150 million and appointed its editor Bari Weiss, as Editor-in-Chief of CBS News.

Takeaway: The key to winning in the newsfluencer economy is building a relationship with the audience. New mediums, like short-form vertical video, can help. But Substack shows that newsfluencers can use one of the oldest tools in media—text sent via email—and still grow their reach.

3. Can liberal newsfluencers catch up? Conversion is the X factor.

After Trump won his second term, Democrats searched for a “Joe Rogan of the Left.” They wanted an influential new media creator who could build a massive audience, reach beyond their base and persuade swing voters. Some liberal streamers, such as Hasan Piker, have built a solid online audience. But it’s not clear whether these creators can reach apolitical viewers as effectively as Rogan or match his overall reach.

But Democrats aren’t completely lost. Joe Rogan and Theo Von, two of Trump’s most popular backers in 2024, have been critical of ICE and Trump’s immigration policies. Unlike some of Trump’s allies, these podcasters aren’t purely political. They’re cultural phenoms whose politics is downstream from their broader brand.

Meaning: The newsfluencer ecosystem felt conservative in 2024, in part, because Trump won over creators who were not partisans. If these same creators sour on Trump, the newsfluencer ecosystem may start to feel, at least temporarily, a little bit less red.

Putting it all together: Most of the trends spotted in our December 2024 report have continued through 2025. Substack, the hub for individual newsfluencers, is growing. More Americans are finding news on TikTok. And news audiences are still hungry for creators they can relate to. The main question is whether the newsfluencer ecosystem remains red or tilts purple as the 2026 and 2028 elections approach.

The NRG KOL Compass is an integral part of our Public Affairs practice, helping businesses and organizations build better messages and communications. For more information contact our team at KOLcompass@nrgmr.com. And subscribe to the NRG Expert Panel Substack for regular KOL updates.

The KOL newsletter is part of NRG’s Public Affairs practice. Subscribe to the NRG Expert Panel Substack for more general population surveys, updates from our 100+ member tech expert panel and more.

On October 1, 2025, New York State’s “Stop Hiding Hate” law took effect. Under the law, large social media companies must report if and how they define, flag, and manage posts that include hate speech, racism, extremism, disinformation/misinformation, harassment, or foreign political interference. Elon Musk’s X Corporation has already sued the state and intends to continue fighting the law.

New York’s law may or may not stand. But if the law remains in place—or if more states successfully enact laws like it—social media companies will need to figure out how they can moderate content and keep users happy.

NRG recently polled 1,039 US social media users aged 18 to 64 on exactly this topic. In January 2025, we found:

A plurality of social media users prefer less government involvement. Four in ten (39%) want the government to be less involved in influencing how social media platforms moderate content. Only 25% want the government to be more involved. Most users said the government was either too strict (31%) or about right (36%) in its approach to content moderation—and only 32% said “too lenient.”

Users want platforms to moderate objectionable or potentially offensive content. When presented with various types of posts—those that encourage violence, spread misinformation, include racism, lure someone into a scam and more—most users wanted to remove the post or allow the post to stay up with restrictions.

Community notes are an answer, but the debate remains open. In theory, 44% of users want social media companies to use independent fact-checkers to moderate content (36% say community notes, 12% say no moderation). But, in practice, 57% of X users say the Community Notes program improved the quality of content on the platform (only 13% said notes had a negative impact). Multiple solutions are possible, but there’s no silver bullet.

Voters may have warmed towards content moderation since these numbers were published. Trump, who was at one time banned from Twitter, has repeatedly criticized social media content moderation. And Musk’s company, X Corp, is pushing back against the Stop Hiding Hate Act. Both men have become less popular since Trump took office. And, generally speaking, American voters tend to sour on the president’s agenda the longer he spends in the White House.

New York’s position on content moderation may be more popular now than it was when the law was initially signed in December 2024. More research is needed. But for now, you can find our latest deep dive on content moderation here.

The NRG KOL Compass is an integral part of our Public Affairs practice, helping businesses and organizations build better messages and communications. For more information contact our team at KOLcompass@nrgmr.com. And subscribe to the NRG Expert Panel Substack for regular KOL updates.

National Research Group (NRG) regularly interviews members of our Key Opinion Leader (KOL) panel, KOL Compass. The panel represents over 100 tech and policy insiders from Fortune 500 companies, NGOs, top law firms, academia, and more. We survey this group to obtain unique insights about the future of tech policy and publish the results. To receive more tech policy updates from NRG KOL Compass and our Public Affairs practice, subscribe here.

“Jimmy Kimmel Live!” returned to the air September 23, just six days after ABC suspended the show. That controversy ended quickly, but the government and major media companies may very well find themselves at odds again in the coming years. Donald Trump sued the New York Times, The Wall Street Journal, CBS News, and ABC News within the last year alone—claiming defamation or other harms—with CBS and ABC settling for $16 million and $15 million, respectively.

The smart move for media companies—or any organization in the free speech business—is to be prepared, just in case they find themselves in a similar fracas. So we asked members of our Key Opinion Leaders (KOL) panel how they would advise media companies to handle situations like the Jimmy Kimmel episode. They had three recommendations.

Stay the course: outcomes vary and patience can pay off

The KOLs pointed out that media companies aren’t destined to lose in court. As one said, “Many of these orgs are folding under the mere *threat* of pressure. It’s clear this government wants to blow a lot of hot air, but almost none of it is holding up in court. The best way these organizations can defend themselves is to stand strong all the way through.”

The takeaway: if a media company wants to win a free speech fight, it shouldn’t give up too quickly. Trump’s net approval rating has fallen by 16 percentage points, from +2% net approval in February to -14% today. The administration has a mixed record in court, including one dismissal, a pending case, and settlements elsewhere. According to some of our panelists, media companies can win if they stick it out.

Messaging strategy: stay calm and proactively explain your values

Two of our experts offered related pieces of advice: stay calm and remember that politicians can and do misread the zeitgeist.

As one KOL said, “Explain to the American people what actually happened—without emotion and political bias (don’t assume what the administration is doing is unjustified). Get that word out so audiences on the conservative side also hear.”

Another said, “All media organizations can do is make sure that Americans feel part of their team... When Jimmy Kimmel was kicked off ABC, it felt like we lost our star player… In such unprecedented and unpredictable times, all media organizations can do is make Americans feel like we are on the same side.”

Calmly explaining the values that led to a decision is a good, unifying political strategy, our experts suggested. Americans—by a margin of 68% to 13%—believe that it’s unacceptable for the government to pressure broadcasters to remove shows it disagrees with. That includes a 42% plurality of conservatives. (Another 29% of conservative respondents said government pressure was acceptable and 28% were unsure).

Politicians live and die by public opinion—and they don’t want to be on the wrong side of it. Public support for Kimmel seemed to grow organically. Should other media companies find themselves in a similar position, their best move might be to proactively appeal to universally held cultural and social values, such as honesty and fair play.

Pick your battles and play the long game

Other KOLs advised media organizations to keep a low profile and work behind the scenes. As one said: “First, try to keep your head down. Not every political crisis requires a comment or post. Second, mobilize your base. Speak with customers directly. Engage them to show support publicly (tweets, posts, likes) for all that you do. Third, get to know your federal representatives early and make sure they know the economic and social value your brand provides. They may help in blunting the President’s pressure. Fourth, don’t make this a war with the President. It’s a struggle to continue serving your loyal customers and the community.”

This playbook won’t work for every brand. Some media companies can’t maintain a low profile while producing the content their audiences want. And other companies have too much at stake to risk a fight with the government. But quiet stakeholder work, direct customer outreach, and early engagement on the Hill can make a game-changing difference.

Overall, the KOLs broadly agree that this is a challenging moment. Their advice: prepare for fights in advance, try to pick only winnable battles, stick to your core values, and don’t panic—especially in the first 72 hours.

The NRG KOL Compass is an integral part of our Public Affairs practice, helping businesses and organizations build better messages and communications. For more information contact our team at KOLcompass@nrgmr.com. And subscribe to the NRG Expert Panel Substack for regular KOL updates.

National Research Group (NRG) regularly interviews members of our Key Opinion Leader (KOL) panel, KOL Compass. The panel represents over 100 tech and policy insiders from Fortune 500 companies, NGOs, top law firms, academia, and more. We survey this group to obtain unique insights about the future of tech policy and publish the results. To receive more tech policy updates from NRG KOL Compass and our Public Affairs practice, subscribe here.

AI is still a new policy issue. As we’ve previously written, only Republicans have articulated a formal AI policy platform and voters don’t yet see the issue through a partisan lens. So, when we asked our KOL tech policy panel about AI and sustainability—an issue that’s getting serious news coverage and stirring significant debate—we expected a similarly messy, partially-formed discourse.

But our experts surprised us.

We interviewed a dozen KOL panelists with deep knowledge about the intersection of tech and the environment. They fell neatly into two groups: pro-growth AI optimists and AI-skeptical environmentalists. And, at this early stage, the optimists have a political advantage.

School of thought #1: Abundance-minded AI optimists

Roughly half of our sample believed that AI had a positive impact on the environment. They made two arguments: that the environmental toll of AI is small and dropping, and that AI will make up for any short-term harms by speeding up technological progress and improving energy efficiency across society.

As one KOL told us, “The energy efficiency of AI inference has improved by orders of magnitude in the past two years and we should expect it to continue as data centers develop even better protocols to take advantage of server racks operating as interconnected units.”

Optimistic KOLs believe that AI will accelerate the development of green technology directly and indirectly. One said, “there are a lot of second-order and third-order effects to consider here. We have reasoning models today that I think a reasonable person ten years ago would consider to be AGI [artificial general intelligence]. These are being integrated into education and job training. The faster this is done, the quicker the world can adapt to environmental needs.” AI optimists also often cite ways that AI will increase the energy efficiency of existing technologies (such as smart grids, supply chain optimization, etc.)

Experts in this school of thought sound like “Abundance” Democrats. Some say that AI will have environmental costs in the short run. But fundamentally they all believe in a future where AI creates sustainability gains across the economy and society. And they don’t want to regulate that future out of existence.

School of thought #2: AI-skeptical environmentalists

KOLs from the opposing school of thought sound much more like traditional environmentalists. They cite data centers’ water and power usage as problems and call for greater regulation.

One KOL cautioned that, “data centers that generative AI relies on for model training are causing environmental devastation due to excessive energy and water use.”

AI-skeptical panelists often favor regulations that directly target AI companies. One KOL from this camp said, “impose environmental regulations [on] tech companies [and] data centers and require environmental ‘nutrition labels’ for consumer-facing products. We need to ask what we’re racing towards and not just blindly think AI is a race”

Still, these KOLs differ in their exact policy prescriptions. Some want to incentivize AI companies to become greener, while others want to restrict the size of data centers, their water usage, and energy consumption. But they tend to emphasize a compromise: feeding green energy into data centers—suggesting that, like the AI optimists, they too are looking for ways to help AI progress in the long run.

AI optimists have a better political position

The AI optimists start with a key advantage: a bigger tent. Republicans will almost certainly oppose any environmental regulation on AI companies. Abundance liberals may be wary of overregulating an industry that could create, well, abundance. Old guard environmentalists, for now, can only capture one part of one of the political parties. At this early moment in the debate over AI and sustainability, that advantage is enough to give the optimists an edge.

But the story isn’t just which side has an early advantage. The range of viable positions matter too. Many on the pro-regulation side agree that AI is an important, permanent part of the economy. The debate is over how much downside there is and how to manage it.

The NRG KOL Compass is an integral part of our Public Affairs practice, helping businesses and organizations build better messages and communications. For more information contact our team at KOLcompass@nrgmr.com. And subscribe to the NRG Expert Panel Substack for regular KOL updates.

National Research Group (NRG) regularly interviews members of our Key Opinion Leader (KOL) panel, KOL Compass. The panel represents over 100 tech and policy insiders from Fortune 500 companies, NGOs, top law firms, academia, and more. We survey this group to obtain unique insights about the future of tech policy and publish the results. To receive more tech policy updates from NRG KOL Compass and our Public Affairs practice, subscribe here.

This chart from Apollo Academy recently went viral among the wonkiest set of AI analysts, researchers, and policy watchers. The data is from the US Census Bureau, and Apollo created the chart.

It’s not hard to see why this data stirred emotions. Other highly publicized datasets show skyrocketing AI usage and a clear, positive return on investment for AI. AI is attracting hundreds of billions in investment—arguably propping up an otherwise sluggish economy. Apollo’s trendlines don’t fit neatly with other data points.

To get better insights into what’s happening here, we presented a small group of KOLs with the chart and asked for their thoughts. Their reactions can be slotted into roughly three categories.

Perspective #1: The AI skeptics have a point.

One member of our KOL panel summarized the AI skeptic reaction succinctly. They said the chart proved “that alleged productivity gains are mostly hype and the ROI is unclear.”

Gary Marcus, a leader in the AI skeptic movement, added a bit more detail in a post on X.

Skeptics have long argued that the cost of training models and generating outputs—in terms of energy and pure cash—are too high. To them, this Apollo chart is more proof that the costs outweigh the benefits and an AI bubble is close to bursting.

Perspective #2: Bureaucracy—not AI tools—are the bottleneck.

Another KOL pushed back against AI skeptics. They said: “Enterprises are still trying to find the use cases that work best. There has been a top-down ‘use AI’ edict, and many of these pilots have not panned out. The future is finding, use case by use case, what makes the business or process more efficient, then powering up with those.”

According to this perspective, the most important trend isn’t a small drop in adoption among large companies. It’s the uninterrupted rise in adoption among firms with 1–4 people.

At small companies, individual employees can experiment with AI and change how the business runs in a matter of days. Larger organizations, by contrast, might put an AI mandate into place, face some initial resistance and realize ROI as time goes on. Moreover, workers at large firms sometimes partake in Shadow AI (using unsanctioned AI tools for work), which may obscure the true adoption trends at bigger companies.

Put simply, some KOLs believe these trendlines represent AI growing pains—which is far from a fatal condition—at some large firms.

Perspective #3: This chart misses the forest for the trees.

One KOL passed on commenting on the short-term dip in the AI adoption trendline among large businesses. Their view on the data: “Larger companies are scaling AI before smaller companies—which could increase their revenues and efficiencies.”

There’s wisdom in seeing the big picture. AI adoption is highest at firms with 250 or more employees. Those firms have the potential to dramatically increase their efficiency using AI, harnessing the technology to create new, innovative products. The real story might be the level of AI usage across companies, not a short-term, potentially reversible trend.

Synthesis: Watch the trend, but don’t buy into the panic cycle just yet.

AI has improved dramatically and spread like wildfire over the last three years. Many companies, large and small, are reporting real ROI from these tools. Adoption trendlines can wobble without erasing those facts.

The best approach, for now, is to wait and see. No one knows how those trendlines will evolve in the next month, quarter, or year.

The NRG KOL Compass is an integral part of our Public Affairs practice, helping businesses and organizations build better messages and communications. For more information contact our team at KOLcompass@nrgmr.com. And subscribe to the NRG Expert Panel Substack for regular KOL updates.

This article is based on recent surveys conducted by the public affairs practice at National Research Group (NRG). We regularly poll the general population and our proprietary panel of tech experts (the Key Opinion Leader—or KOL Compass—panel). More posts can be found here.

Corporate executives aren’t exactly America’s sweethearts. According to Gallup, only 12% of Americans think business executives have high ethical standards and 43% say they have low standards (putting them just above car salespeople and just below state politicians). And executives don’t have a good read on how the public feels about them or their businesses. PwC found that 90% of business executives believe that consumers trust their companies, while only 30% actually do.

This is a bad situation for everyone. Consumers would be happier if they felt they could trust businesses and the people who run them, and companies would likely gain more business as a result. To get greater insights into this trust gap, we asked our two groups of respondents—913 members of the general public and 104 high-powered decision-makers—to play a simulation game. We instructed them to pretend to be a CEO focused on protecting their business’s reputation. Each respondent was given 10 “business protection points” and instructed to allocate those points across 18 different strategies for protecting their reputation.

Our results showed a fairly sharp contrast between our two sets of respondents. The average American thinks the best way to improve a company’s reputation is to provide direct, material help to employees or consumers. Our elite, decision-maker demographic was more focused on profits and losses.

Overall, our general-population respondents gave these five strategies the highest total number of “business protection points”:

1. Giving workers better benefits and wages

2. Making products and services more affordable

3. Improving workplace conditions

4. Enhancing customer privacy and data protection

5. Standing up for corporate values like equality, inclusion and fairness

The top four strategies directly help workers or employees by lowering prices, increasing wages, improving the workplace, or protecting their data. Only one of the top five strategies—endorsing values like equality—provided an indirect benefit.

Our decision-maker demographic—those who reported having a mid-level or higher management job in finance, government, marketing, media, or legal services—agreed that giving workers better wages was the best option. But, after that, they focused on promoting innovation and small business while protecting themselves from risk. They ordered their priorities this way:

1. Giving workers better benefits and wages

2. Increasing funding for customer and product experience innovation

3. Increasing efforts to support small businesses and entrepreneurs

4. Monitoring issues that could lead to boycotts or customer backlash

5. Working closer with regulators to ensure compliance with laws and standards

There’s nothing inherently wrong with corporations trying to innovate or attempting to protect themselves from a boycott. Those strategies can boost revenue and minimize losses. But these results suggest that parts of the managerial elite may be out of touch with customers and employees. The average American wants better products at lower prices and better jobs with higher wages. Unlike management, they are less concerned with end-of-year profits and losses.

There are lessons here for managers and executives in high-prestige industries. Strategies that earn applause in the boardroom won’t always land with the public. If trust is the goal, the surest path is to deliver tangible benefits to the people who rely on you most—your customers—and the people you rely on most—your employees.

Our Public Affairs practice helps businesses and organizations build better messages and communications. For more information, contact our team at KOLcompass@nrgmr.com. And subscribe to the NRG Expert Panel Substack for regular updates.

National Research Group (NRG) recently interviewed over 100 members of our Key Opinion Leader (KOL) panel, KOL Compass. The panel represents tech and policy insiders from Fortune 500 companies, NGOs, top law firms, academia, and more. We regularly survey this group to obtain unique insights about the future of tech policy and publish the results. To receive more tech policy updates from NRG KOL Compass and our Public Affairs practice, subscribe here.

Americans are divided on most major political issues. But not on AI. According to YouGov, 50% of Republicans and 46% of Democrats think AI will have a negative effect on society. And, according to NRG’s own survey data, majorities of liberals and conservatives believe the government should heavily regulate AI to protect consumers and society. Voters know what they think (and what both parties think) about other important issues, such as abortion and immigration. But, for voters, AI isn’t a partisan issue yet.

The situation in Washington DC is dramatically different. We recently provided our KOL tech policy expert panel 15 different messages and asked them which messages would increase trust the most if deployed by a tech company. The goal was to see what messages landed with the operators and thinkers who are shaping tech policy on both sides. We found that Republican and conservative-leaning tech experts have found their message, but Democrats haven’t. And that gap presents a huge opportunity for anyone who wants to influence tech policy.

Republican and conservative-leaning tech experts are unified: they want MAGA for AI. The five messages that resonated most with that group were:

1. INNOVATION GROWS SMALL BUSINESS: We believe in a competitive marketplace; our innovations empower small businesses, not crush them.

2. CYBER PROTECTIONS: Our platforms defend US citizens against foreign propaganda and cyberattacks.

3. GEOPOLITICAL THREATS: We protect American technology from geopolitical threats, especially China's economic and digital aggression.

4. AMERICA FIRST: Our top priority is keeping technology innovation, and economic power, rooted in the US.

5. RESHORE: We're committed to reshoring key operations to support American workers and reduce dependence on foreign governments.

These messages center on helping small businesses and protecting the homeland. They’re compatible with The White House’s AI Action Plan (which emphasized winning the global AI race, pushing homegrown AI technology forward and preventing AI chatbots from being politically biased). Put simply, conservative policy elites seem mostly aligned with the tone and substance of Trump’s AI agenda. And Republican voters will likely support this agenda once they start processing AI as a political issue.

Democratic and liberal-leaning experts don’t yet have a unified approach. Here are the top five messages that built trust with our panelists on the left:

1. ACADEMIC & CIVIC PARTNERSHIPS: We work with local institutions, veterans’ groups, and community organizations to ensure tech benefits every American at the national and local level.

2. CYBER PROTECTIONS: Our platforms defend US citizens against foreign propaganda and cyberattacks.

3. INNOVATION GROWS SMALL BUSINESS: We believe in a competitive marketplace; our innovations empower small businesses, not crush them.

4. BROAD AMERICAN SUPPORT: We're proud to create high-paying American jobs, not just in Silicon Valley, but across the country and the heartland.

5. LEGISLATION ISN'T TRUST BUILDING: We believe tech companies must earn the public's trust and we're committed to doing so without waiting for legislation to force our hand.

None of these messages are politically toxic. Voters, if polled, likely would favor partnerships between tech companies and local civic organizations. And our battery of messages did not include some progressive messages (such as environmental sustainability) that have more recently become flashpoints. But, generally speaking, the Democratic Party hasn’t settled on a single, coherent AI plan.

Individual Democrats have ideas on AI. Joe Biden tried to lay out a regulatory framework for AI as he left office. House and Senate Democrats wanted to investigate the environmental impacts of AI last year. And Bernie Sanders has floated using AI productivity gains to shorten the work week. But, unlike Republicans, Democrats don’t have a singular leader who can weave these ideas into a platform and unite the party behind it. And many Democrats may opt to wait and see whether AI produces crushing job losses, massive economic growth, both, or neither before they commit to a single economic or regulatory plan.

The upshot: now is the best time to push on AI policy. Eventually, the battle lines on AI policy will harden. Democrats will find their message. Republicans will refine theirs. And voters from both sides will start to see AI through a partisan lens. But for now, the debate is still malleable. If a tech company, citizen group, politician—or really anyone—wants to mold tech policy, they should do so before the cement dries.

The NRG KOL Compass is an integral part of our Public Affairs practice, helping businesses and organizations build better messages and communications. For more information contact our team at KOLcompass@nrgmr.com. And subscribe to the NRG Expert Panel Substack for regular KOL updates.

National Research Group (NRG) recently interviewed over 100 members of our Key Opinion Leader (KOL) panel, KOL Compass. The panel represents tech and policy insiders from Fortune 500 companies, NGOs, top law firms, academia, and more. We regularly survey this group, obtain unique insights about the future of tech policy and publish the results. To receive more tech policy updates from NRG KOL Compass and our Public Affairs practice, subscribe here.

We recently polled two groups—our KOL panel of tech experts and Americans in general—and both agree that OpenAI is leading in the AI race. But on the slightly different question of “Who is most likely to win the AI race?” our experts are much more bullish on NVIDIA but bearish on Amazon and Meta.

The reason: experts believe first-mover advantages in the AI race are hard to overcome. That’s why competitors are taking big swings—such as hiring sprees, finding markets outside of consumer apps and more—to try to catch up or even redefine what the AI race means.

Here are the top eight companies, according to a recent NRG survey of 913 US adults.

In the public discourse, “winning” the AI race typically means dominating the market for consumer AI apps or creating artificial general intelligence—AI that matches or surpasses human performance on all cognitive tasks. The public, very reasonably, said that either OpenAI or one of the best-known tech giants (Meta, Google, Amazon and Microsoft) is leading.

Our expert panel thought about the AI race differently.

This data is from earlier in the summer and doesn’t fully price in Meta’s hiring spree, the release of GPT-5 or other potentially relevant recent developments. But, when the KOLs explained their choice, they cited long-term advantages in infrastructure or market share.

Here’s one KOL from June/July 2025 on their vote for OpenAI: “They were a major first mover with the release of ChatGPT, tend to monopolize media coverage, and keep innovating. Everyone else is playing catch up.”

An expert who chose NVIDIA gave a similar explanation: “NVIDIA leads the AI race due to its dominant position in GPU hardware, which powers the majority of AI training and inference workloads. Its CUDA ecosystem and software stack are deeply integrated into nearly every major AI research and production environment. The company has strong partnerships...NVIDIA is not just a chipmaker—it’s building the entire AI infrastructure stack.”

Put simply: experts believe that, once a company is ahead in a general category of the AI race—such as dominating the GPU market or creating the best-known consumer AI brand—other companies can’t easily close the gap. Bolder strategies are needed.

Some companies are hedging their bets. They’re trying to win an entirely different race while still competing for everyday LLM consumers. According to one KOL: “[Google] DeepMind is way ahead of all the others on the list when it comes to applying AI methods to scientific problems and questions. While LLMs will have [an] impact on the nature and shape of white collar work, AI deployed in the services of chemistry, biology, medicine, physics, astronomy, materials, fluid dynamics will have profoundly transformative effects on humanity.”

Others are throwing Hail Marys to try to take the lead. Meta recently offered AI researchers nine-digit compensation packages in an attempt to poach talent from competitors. Chinese companies are chaining lower-quality chips together and pivoting hard to open source to try to create strong AI sans NVIDIA.

These companies recognize that early leads matter in AI—and they can’t catch up by simply hustling a little bit harder.

The NRG KOL Compass is an integral part of our Public Affairs practice, helping businesses and organizations build better messages and communications. For more information contact our team at KOLcompass@nrgmr.com. And subscribe to the NRG Expert Panel Substack for regular KOL updates.

National Research Group (NRG) recently interviewed over 100 members of our Key Opinion Leader (KOL) panel, KOL Compass. The panel represents tech and policy insiders from Fortune 500 companies, NGOs, top law firms, academia, and more. We regularly survey this group, obtain unique insights about the future of tech policy, and publish the results. To receive more tech policy updates from NRG KOL Compass and our Public Affairs practice, subscribe here.

In the early days of Donald Trump’s second term, many CEOs rushed to end DEI, ESG, and all other public displays of corporate social liberalism. Businesses believed that, in order to be on the right side of public opinion (and the president) they needed to drop progressive stances on social issues and adopt MAGA positioning—or, at the very least, go silent.

But data from our KOL panel suggests that companies have some freedom to choose a socially liberal position—if the right conditions are met.

Specifically, our panelists understand that the progressive protest summer of 2020 is over and the political atmosphere in corporate America has shifted right. But, our experts believe that if a company’s core business is affected or a true crisis of democracy emerges, there’s room to stake out a center-left or even liberal policy position.

The KOL panelists clearly articulated this dynamic in a series of questions on immigration. They agree with the general GOP consensus that enforcement is good for business.

But seven in ten KOLs (including 55% of Republican or conservative-leaning Independent KOLs) believe that companies have room to speak on immigration if their business interests are threatened.

In crisis situations, where democratic norms are an issue, panelists also think that companies have a duty to act. Only 34% of the panelists overall think companies should stay silent at these critical moments.

Businesses don’t have political carte blanche, though. When immigration is merely an issue of the day—and not posing a threat to their business or impacting democracy itself—KOLs were split on whether businesses should do anything.

These even numbers mask division between conservative and liberal KOLs. Roughly two-thirds of the conservative panelists said that companies should avoid public comment and stay focused on the business, while a similar share of our liberal experts said companies have a responsibility to address these issues.

Moreover, our experts don’t believe companies should align with protest movements. Businesses can’t control what happens at a protest or whether public opinion turns against a movement. A majority of both conservative and liberal KOLs said companies should avoid allying with protests. The overall numbers:

Taken together, these results communicate a clear message. Businesses can take a centrist or even left-of-center position if it’s carefully tied to a business interest (such as keeping non-citizen employees in the country) or a truly unignorable crisis. Companies would be unwise to race to the left as quickly as they did during the protest summer of 2020. But they may have just a bit more running room on the left than they did in November 2024.

The NRG KOL Compass is an integral part of our Public Affairs practice, helping businesses and organizations build better messages and communications. For more information contact our team at KOLcompass@nrgmr.com. And subscribe to the NRG expert panel substack for regular KOL updates.

National Research Group (NRG) recently interviewed over 100 members of our Key Opinion Leader (KOL) panel, KOL Compass. The panel represents tech and policy insiders from Fortune 500 companies, NGOs, top law firms, academia, and more. We regularly survey this group, obtain unique insights about the future of tech policy, and publish the results. To receive more tech policy updates from NRG KOL Compass and our Public Affairs practice, subscribe here.

The Trump Administration wants US companies to build AI fast—and face as few regulations as possible. But the states aren’t so sure. According to the National Council of State Legislatures, legislators in all 50 states have introduced AI-related proposals, and 38 states have adopted or enacted roughly 100 new AI-related laws just this year. The result: a chaotic policy landscape, where national and various governments are moving in wildly different directions simultaneously.

We asked our KOL panel how to navigate this constantly changing patchwork of laws—that is, which areas of AI policy are amenable to state action, which ones require the federal government to step in and where policy will be made by states and the nation. Their verdict: the federal government is well-positioned to lead on data governance, but groups who want to influence any other policy area need to work outside DC as well.

The easiest way to think about this data is to break the issues into three groups.

First, there are data governance issues: copyright protection, data privacy, deepfakes, and misinformation. The federal government is naturally positioned to lead on these issues. As one KOL put it, “AI (and most other tech) does not stop at state boundaries. Ever since the Internet became widespread, there's been a difficult push/pull with federal vs. state regulation around technology. How do you establish a state policy when digital information flows everywhere?”

But state governments will have more to say on issues related to discrimination and legal battles. Red and blue states will have very different views on how law enforcement uses AI, the right role for AI in a state-by-state patchwork of election systems, and whether AI introduces bias in employment, loan approval, or consumer pricing. If a corporation or group wants to influence policies in these areas, they’ll need to go through—not around—the states.

And states will have the most influence when AI collides with physical infrastructure—such as autonomous vehicles, energy, healthcare, or education. Data centers, Waymo cars, hospitals, and schools aren’t digital abstractions. They’re located in specific states, and the states will have a strong say in how they’re governed.

Maybe most importantly, state-level efforts are a hedge against federal inaction. When the Senate voted to kill the moratorium on state-level AI regulations in Donald Trump’s budget, they ensured that states could move legislation forward, even if DC remained gridlocked. As one KOL observed: “The federal government is unlikely to do any comprehensive legislation—except [on] kids—and the states continue to come up with new lines of legislation and regulation.”

Put differently: if a group wants to lead on AI policy, they have to move through the states.

The NRG KOL Compass is an integral part of our Public Affairs practice, helping businesses and organizations build better messages and communications. For more information contact our team at KOLcompass@nrgmr.com. And subscribe to the NRG expert panel substack for regular KOL updates.