Trump’s AI testing plan faces hurdle as DOGE gutted security teams
▼ Summary
– Trump signed an executive order expanding voluntary safety testing of frontier AI models, but critics say it offers only performative reassurances and changes little about model deployment.
– Trump canceled an earlier signing event last month, citing concerns the order might impede AI innovation, though reports point to infighting between cybersecurity experts and deregulation advocates.
– The final order promises not to stifle innovation with burdensome regulation and establishes no requirements for AI firms, setting up a voluntary government-industry safety review process.
– The order claims it will “ensure that the best and most secure technology is deployed rapidly” and enhance an “America First cybersecurity effort” for global AI dominance.
– Experts say little changed between the leaked draft that caused industry backlash and the order Trump signed without a major event involving CEOs.
On Tuesday, Donald Trump signed an executive order expanding the government’s voluntary safety testing of frontier AI models, but critics argue the measure is largely symbolic. They warn that it offers little more than performative reassurance about federal oversight of AI risks, while doing almost nothing to alter the timing or conditions under which these models are deployed.
The signing came after a rocky buildup. Last month, Trump abruptly canceled a planned event where he had hoped to unveil an earlier version of the executive order alongside CEOs from major AI companies. Several executives were invited on short notice and could not attend, though they still voiced support for the initiative. Trump publicly explained the delay by saying he feared the original order went too far and risked becoming a “blocker” that would stifle innovation. Behind the scenes, reports described internal conflict within his administration, with cybersecurity experts pushing for stronger safeguards and officials advocating for a deregulated AI environment.
The final version of the order is notably softer. It explicitly states that the goal is not “to stifle this innovation with overly burdensome regulation” and imposes no binding requirements on AI firms. Instead, it creates a voluntary framework for companies to cooperate with the government on safety reviews. According to the order, this process is designed to “ensure that the best and most secure technology is deployed rapidly to confront any and all threats to our country.”
Trump framed the initiative as part of an “America First cybersecurity effort” that would simultaneously strengthen national security and cement U. S. dominance in global AI.
Yet experts reviewing the final text found that little had changed from the leaked draft that initially sparked industry pushback. The executive order Trump signed without the fanfare of a CEO summit is essentially the same document, now stripped of any pretense of mandatory oversight.
(Source: Ars Technica)
