Whatever AI Safety Is, It's Not This: Trump's Self-Policing Accord
Trump's AI safety accord leaves labs to police themselves — a strategy with a poor track record in the auto industry and few answers for real AI incidents.

Updated
Why it matters
- Trump said AI companies "are gonna police themselves" after signing an AI safety accord with industry CEOs.
- In 1966, nearly 51,000 people died on US highways; federal law mandated seat belts in all new cars by 1968 after industry refused voluntary adoption.
- A CNN report described the US military nearly boarding a Chinese ship based on a false AI-generated report about nuclear materials.
On Tuesday, a group of AI CEOs gathered around US president Donald Trump to congratulate themselves on signing an AI safety accord that leaves the labs largely responsible for keeping themselves in check.
The president summarized the arrangement the next day in the Oval Office. "They're gonna police, and they're gonna police each other, and they're gonna self-police, they're gonna police themselves," Trump said. "It's going to work out very well."
The stakes are high because the industry's own leaders keep saying so. Executives, employees, and tech luminaries have argued that the field needs clear, thoughtful, tactical rules to avoid potential disaster. Bill Gates has been on a media blitz warning about an AI apocalypse. The accord signed this week delivers essentially the opposite: self-regulation.
The White House plan calls to mind an earlier debate over safety technology. In 1966, nearly 51,000 people died on US highways. The evidence overwhelmingly suggested that many lives could have been saved with modern seat belts, which had already existed for several years. The automotive industry did not voluntarily install them. Instead, in the fall of 1966, Congress established the Department of Transportation and passed the National Traffic and Motor Vehicle Safety Act and the Highway Safety Act, giving the federal government sweeping authority to set safety standards. By 1968, seat belts were required in every new car. Despite a massive increase in highway traffic over the last 60 years, far fewer fatalities occur on the road today.
The lesson is straightforward: when safety imposes costs, industry self-regulation tends not to happen, and government authority is what changes behavior.
The analogy is imperfect, and it matters why. There is no seat belt equivalent for AI — no straightforward solution that demonstrably reduces risk overnight. It also took decades for most states to pass laws requiring people to actually wear seat belts. AI's potential harms are less certain and less quantifiable. Doomers argue the technology could result in human extinction within a decade; skeptics call that sci-fi marketing malarkey.
Even sympathetic observers concede the problem is genuinely hard. The question of how to properly "pace the frontier," as executives themselves like to say, remains open. If all the major AI labs agreed to a slowdown, there is no reliable mechanism to hold them to their word. China would likely continue on its own path regardless. The US economy is increasingly propped up by the AI industry, which has become an ouroboros of financing deals and circular investments. Any meaningful crackdown risks instability reaching well beyond Silicon Valley.
But uncertainty cuts both ways. A recent CNN report described how the US military supposedly nearly boarded a Chinese ship in part because a false AI-generated report claimed it was carrying nuclear materials. What does the AI accord accomplish to prevent similar incidents? That remains unclear.
OpenAI offers a partial case study in the limits of voluntary restraint. The company has, to its credit, delayed and even canceled the release of its most advanced models over security concerns. But it did so only after those models went on various, extensive hacking sprees, new details of which continue to trickle out at a steady pace.
In an interview with TIME published Thursday, Trump suggested that the Department of Justice and the FBI were sufficient "guardrails" for the AI industry. That claim runs into immediate problems. It is far from clear whether AI companies can be held legally responsible when their models go rogue. And AI agents themselves are not being served subpoenas.
The honest assessment is that no one currently knows how to solve AI safety. Developing and enforcing the right framework will take massive amounts of coordination, resources, and regulatory authority — exactly the kind of thing governments are designed to provide. Instead of attempting that work, the Trump administration is stepping aside and telling everyone to step on the gas.
Don't bother buckling up.
Original: bts.gov
More from Marcus Bennett
Show full bio
Senior reporter covering consumer brands and retail at AI In Context.
138 articles
Related articles
- Trump's White House AI Accord Leaves Safety to Companies Themselves
- Khanna Bill Would Ban Self-Improving AI Until Federal Safeguards Exist
- Bill Gates: AI Without Safeguards Is 'Completely Irresponsible'
- California bans AI-only firing decisions and bathroom surveillance
- White House 'Super Intelligence Accord' Is a Voluntary Pledge, Not AI Regulation