Americans distrust AI and that could be good news

Americans harbor deep suspicions about artificial intelligence, and that skepticism might be exactly what the technology needs. Recent polling shows AI sentiment at -18%, placing it between fracking (-9%) and race-aware college admissions (-23%) in public opinion. While Americans worry about specific AI applications—self-driving cars poll at -18% and AI workplace surveillance at -46%—the technology remains a relatively minor concern in national political discourse.

Many AI safety advocates dismiss public opinion as uninformed or reactionary, worried that populist backlash could lead to counterproductive regulation. However, this negative sentiment represents an underutilized force that could meaningfully slow unsafe AI development. Public opposition to emerging technologies has historically created valuable friction that prevents hasty implementation and forces more careful consideration of risks.

This dynamic operates through three distinct mechanisms that transform diffuse public concern into concrete barriers for reckless AI deployment.

Cultural permission structure

Public sentiment creates an environment where individuals with influence feel empowered to act on their concerns. During the early 1900s, investigative journalism—known as “muckraking” for its willingness to expose corruption—enabled President Theodore Roosevelt’s trust-busting campaigns. These journalists understood that public resentment toward large corporations would drive newspaper sales, creating a feedback loop where reporting both responded to and shaped public opinion.

The journalism fanned anti-trust sentiment, shifting the Overton window—the range of policies considered acceptable for public discussion—and making increasingly aggressive anti-corporate reporting possible. This media environment, in turn, influenced public perception and provided political cover for Roosevelt’s regulatory actions.

The same dynamic applies to AI development. When the public perceives AI negatively, whistleblowers and insiders feel more confident speaking out about safety concerns. Whistleblowers like Daniel Ellsberg, who leaked the Pentagon Papers revealing government deception about the Vietnam War, and Edward Snowden, who exposed mass surveillance programs, acted partly because they sensed their revelations would resonate with public sentiment.

This cultural permission structure already influences AI development. Safety researchers have felt empowered to leave major AI companies and speak publicly about their concerns, knowing their community would support such actions. Former OpenAI researchers’ departures and subsequent public warnings about AI risks exemplify how negative sentiment can encourage transparency from industry insiders.

Corporate lip-service

Companies actively respond to public sentiment, even when that response begins as superficial marketing. During the height of corporate activism in the 2010s, many firms adopted LGBT-friendly or environmentalist messaging once their competitors started doing so, fearing consumer backlash if they remained silent.

While corporate statements can seem hollow, maintaining complete hypocrisy between public positions and private actions proves surprisingly difficult. Most executives, even at the highest levels, attempt to maintain some coherence between their stated values and business practices.

Oil companies provide a compelling example. During the 1990s, these firms invested heavily in greenwashing campaigns to rehabilitate their public images. BP famously rebranded itself with the tagline “Beyond Petroleum” in 2000, emphasizing renewable energy investments. Later, during the ESG (Environmental, Social, and Governance) investment boom of 2015-2020, companies without stated climate goals suffered in financial markets. ExxonMobil, after years of resistance, finally published climate risk reports in 2018 following sustained investor pressure.

Critics rightfully point out that greenwashing often neutralizes public concern through PR rather than meaningful change. However, the alternative to “Beyond Petroleum” was never BP’s shutdown—it was no change whatsoever. Corporate statements, even when hypocritical, create ammunition for activists, investors, and regulators who can hold companies accountable for their public commitments.

Legislative and regulatory friction

Public opinion rarely creates dramatic policy changes on its own, but it excels as a braking mechanism on political and corporate machinery. Former President Trump’s attempts to repeal the Affordable Care Act ultimately failed due to sustained public backlash. Vocal public opposition also scuttled the Keystone XL pipeline extension. NIMBYism—the “Not In My Backyard” phenomenon where communities resist local development—derives its persistent power from status quo bias, the human tendency to prefer existing conditions.

This braking force operates throughout society’s decision-making structures. Beyond corporate lip-service, it forces similar responses from politicians, particularly in local and state politics where public sentiment carries more weight. Local resentment toward AI has already produced facial recognition bans in San Francisco and Boston. Similar sentiment could support ballot initiatives in California cities, directly regulating Silicon Valley operations, or fuel digital NIMBYism in data center districts, as currently occurring in Northern Virginia.

At the state level, attorneys general could leverage negative AI sentiment to launch investigations and file complaints, positioning themselves as consumer advocates. Negative sentiment could also inspire state-level laws restricting AI training, data collection, or deployment.

Friction remains effective even when specific measures fail. If companies perceive public opposition to AI, they’ll anticipate regulatory risk—that some city council member, attorney general, or state senator with a vulnerable seat might position themselves as the “anti-AI candidate.” They’ll also recognize that any perceived misstep makes such punishment more likely, potentially leading them to invest more resources in safety measures or at least in public relations.

This unpopularity creates an asymmetric advantage for safety advocates. Because public opinion excels at stopping rather than starting initiatives, and because it’s fundamentally conservative, it naturally favors safety over acceleration when framed appropriately.

Practical implications

The goal involves harnessing existing negative public opinion to increase friction around unsafe AI practices. The primary objective isn’t dramatic Congressional bans or massive protests—though unpopularity might make these more likely—but rather the cumulative effect of an unfriendly media environment, institutional lip-service, and local regulatory risk.

The general public fears AI appropriately, but these concerns lack salience in daily political discourse. The bottleneck lies in making AI risks more prominent in public conversation. This requires appropriate messaging: while extinction-level risks matter, most people respond more viscerally to job loss, surveillance, and algorithmic decision-making that affects their daily lives.

This approach shouldn’t dismiss “wrong” concerns about AI, such as environmental impact, artists’ rights, or geopolitical competition with China. While arms race framing—the idea that America must win an AI competition with China—has already influenced congressional sentiment, it represents a different calculation aimed at lawmakers rather than the general public.

However, stoking anti-AI sentiment carries risks. Arms race narratives and broad regulatory approaches like the EU AI Act have emerged from negative sentiment, and future backlash could push AI companies toward superficial safety measures or even underground development. Not every anti-AI narrative deserves amplification, as some can be manipulated by AI companies or cause more harm than good.

The solution involves focusing negative sentiment on public concerns—job displacement, surveillance, loss of human agency—rather than congressional priorities like international competition. Public sentiment operates asymmetrically; it’s easier for a mistaken Congress to pass harmful legislation than for a mistaken public to force Congress into poor decisions.

Increasing salience around job loss, surveillance, and automated decision-making makes strategic sense. These concerns are realistic, near-term, and directly affect ordinary people rather than just policymakers. While these might seem like mundane harms compared to existential risks, they can create friction that indirectly addresses more serious dangers.

Mundane concerns could establish groundwork for existential ones or provoke responses that reduce catastrophic risks. Worries about job loss, surveillance, and human decision-making could slow AI deployment into critically important domains. Building on these existing concerns creates the kind of friction that makes reckless development incrementally more difficult.

The path forward

Public skepticism toward AI isn’t a problem to solve—it’s a resource to harness. Rather than dismissing negative sentiment as uninformed, safety advocates should recognize it as a potentially powerful tool for creating the friction necessary to slow unsafe AI development.

The key lies in making these concerns more salient while focusing on issues that resonate with ordinary people’s daily experiences. This approach won’t solve AI safety challenges on its own, but it can create an environment where doing the wrong thing becomes consistently more difficult—and that incremental friction might prove invaluable as AI capabilities continue advancing.