On Tuesday, Donald Trump signed an executive order to test AI models for safety, but critics argue it’s more performative than practical. The order avoids heavy regulations and instead sets up voluntary government collaboration with companies, promising rapid deployment of secure tech.
The plan seems to be a non-starter: last month, Trump canceled a signing event when the original EO faced backlash from industry leaders. Even invited CEOs couldn’t make it, suggesting lukewarm support at best. Meanwhile, cybersecurity experts clashed with pro-deregulation officials in his administration.
Experts suggest that the leaked draft and final order are largely indistinguishable, despite Trump's claim of major changes. The EO focuses on 'America First' cybersecurity but fails to address real concerns about AI safety and oversight.
The situation is reminiscent of a well-known meme: ‘US security teams were gutted by DOGE’ – although in this case, the joke might be on us if we rely too heavily on voluntary measures for AI safety.
