In 1966, U.S. highways recorded roughly 51,000 fatalities, a toll that could have been reduced by the already-available seat-belt technology. That year Congress created the Department of Transportation and passed the National Traffic and Motor Vehicle Safety Act along with the Highway Safety Act, granting the federal government authority to impose safety standards. By 1968, every new automobile was required to include a seat belt, and despite a six-decade surge in traffic, road deaths have fallen markedly.
Earlier this week, AI chief executives gathered with President Donald Trump to endorse a new AI safety agreement. According to a piece by Maddy Varner, the accord places the bulk of compliance responsibility on the laboratories themselves, effectively making self-regulation the primary mechanism for risk mitigation. The agreement does not establish external enforcement structures, leaving oversight largely to the industry’s own discretion.
Industry leaders have repeatedly called for concrete, tactical regulations to avert catastrophic outcomes, while public figures such as Bill Gates have warned of an imminent AI-driven apocalypse. In contrast, President Trump has suggested that repeated assertions can establish truth, implying that the mere proclamation of safety measures may be sufficient. This divergence underscores the tension between calls for stringent policy and the administration’s more laissez-faire stance.
Enforcing any slowdown faces practical obstacles: there is no simple device comparable to a seat belt that can instantly lower AI risk, and states historically took decades to require drivers to actually wear belts. Moreover, the United States’ economy is increasingly intertwined with AI ventures, and any heavy-handed crackdown could destabilize markets beyond Silicon Valley. Meanwhile, China appears poised to pursue its own trajectory regardless of U.S. policy choices.
A recent CNN report highlighted a near-miss incident in which a U.S. naval unit almost intercepted a Chinese vessel after an AI-generated intelligence brief falsely claimed the ship carried nuclear material. At the same time, OpenAI has postponed and even cancelled releases of its most advanced models citing security concerns, though those models had already been subject to extensive hacking attempts. President Trump has argued that the Department of Justice and the FBI provide sufficient guardrails, a claim whose legal viability remains uncertain.
Resolving AI safety will demand extensive coordination, substantial funding, and authoritative regulatory frameworks,functions traditionally performed by government agencies. Instead of constructing such mechanisms, the current administration appears to be removing barriers and urging rapid development, effectively telling the sector to accelerate without the protective “seat belt” that federal oversight once provided for motorists.