Security1 publisher2 min readPublished
Trump defends hands-off AI rules after hacks carried out by US frontier models
Pre-release testing of frontier models runs on a voluntary executive order plus a second, non-public agreement, and the enforcement Trump named is the Justice Department, without saying under what authority.
The Watch · Security desk
What happened
- Trump kept defending his administration's hands-off approach to AI regulation after hacks carried out by US commercial frontier models rattled policymakers and spurred calls for oversight.
- In his post he wrote "Whoever wins AI, WINS!" and said he would not stifle growth in a technology he called bigger than the Industrial Revolution or the internet.
- The executive order that created the voluntary testing regime was delayed and altered by AI industry boosters so that government review would not push back model release dates.
- Bessent told Congress that private lawsuits could force AI companies into better security and that the government should not grant the labs the liability exemption they are seeking.
- The US government has accused Chinese AI companies of systematic distillation of US frontier models with the implicit encouragement of Beijing.
Compiled by The WatchSomething wrong?How this is made
Why it matters
- constraint With the operative testing terms unpublished, a customer cannot cite a government standard in vendor diligence; model-abuse requirements exist only where the customer negotiates them.
- exposure The safety route the Treasury Secretary described runs through courts after harm, so the first full accounting of a model-enabled intrusion is likely to happen in discovery.
- decision Security leaders funding model-abuse monitoring have no federal requirement to point to, so the internal case rests on liability exposure and incident cost.
The testing regime in the executive order issued earlier this year is voluntary and largely at private industry's discretion [7]. The administration then struck a second, non-public agreement with frontier AI companies including OpenAI and Anthropic, covering pre-release testing [9]. Frontier pre-release testing has therefore run under two arrangements this year, one on the public record and one private [19]. Whatever a buyer wants verified before a model reaches its network has to be written into the purchase agreement.
Trump named his own enforcement route. "We will be careful, and that's why we have the Department of Justice, and other Law Enforcement bodies, that will rein things in if we have to, but I will only encourage AI or, SI (SUPER INTELLIGENCE)!" he wrote [4]. He did not say under what legal authority, or where he would draw the line [5]. He has previously suggested that good leadership is the only AI regulation the US needs [6].
"The best way to guarantee safety is that the creators are liable for what they build and generate," Treasury Secretary Scott Bessent said [11]. Barack Obama, in remarks last week at Colgate University, rejected the same reasoning as a safety plan [12]. "If it turns out to be dangerous, people will just sue them and they'll be worried about financial liability," Obama said. "That's not how we treat airlines or drug companies or food companies." [13]
The competitive case for holding off comes from David Sacks, co-chair of the President's Council of Advisors on Science and Technology and a top adviser on AI, who cited the threat from China and from other countries he said would not face similar restrictions [14]. "We're not the only country that has advanced AI labs, and as the president declared...we have to win this AI race," Sacks told Politico in May [15]. The argument against that position is that inadequate regulation and cybersecurity controls in current AI systems make those systems impossible to fully control or monitor [17].
For a defender, the week yielded a policy direction and no filing date. Trump also compared warnings that "AI is going to kill us" to environmentalist complaints about climate change, which he called a false narrative [2]. CyberScoop's account describes the hacks behind the renewed calls for oversight only as carried out by US commercial frontier models [18].
What to watch
- A first Justice Department action against a frontier lab, and the statute it is brought under.
- Publication of the non-public pre-release testing agreement, or its terms surfacing through a lawsuit.
- A civil filing that tests Bessent's creator-liability theory against a model-enabled intrusion.