Trump backs independent AI audits in shift toward stronger safety measures

By Maggie Eastland

 

President Donald Trump endorsed using outside auditors to assess the safety of artificial intelligence systems through an accord with Silicon Valley leaders that seeks to sidestep new government rules in addressing a groundswell of concern over AI risks.

 

Under the agreement forged at a White House lunch Tuesday with nearly two dozen top tech industry executives, AI companies also committed to regularly meet to set best practices for the industry. The document carries no legal weight, though the president told reporters it was “morally binding.”

 

“It’s almost like a constitution, in a way,” Trump told reporters in the White House driveway following the event. “The biggest people in the world signed that, and I signed it as president, and it really is a form of protection.”

 

 

The accord emerged after a nearly two-hour meeting between the president and his guests that took place against a backdrop of rising global anxiety surrounding AI’s potentially catastrophic risks. Signers included leaders of the AI industry’s most prominent developers: Anthropic PBC Chief Executive Officer Dario Amodei, Nvidia Corp. CEO Jensen Huang and SpaceXAI’s Elon Musk.

 

Other signers included OpenAI President Greg Brockman, Meta Platforms Inc.’s Mark Zuckerberg, and Sundar Pichai, CEO of Google parent Alphabet Inc., according to a copy posted by Trump on his Truth Social network. Several of the firms including Anthropic and OpenAI have previously committed to having independent evaluators. 

 

Trump’s willingness to engage on safety after previously dismissing doomsday fears as “a hoax” reflects how AI has increasingly become politically charged ahead of the November elections. Trump has emerged as one of the loudest champions of unfettered AI development, putting him at odds with voters who are skeptical of the technology. More than seven in 10 voters said they prefer candidates who back stricter AI guardrails, according to a new Quinnipiac University poll.

 

Public fears about AI’s risks have grown following a series of high-profile hacks by rogue OpenAI agents — including one that targeted an Australian government website — and recent warnings from former AI company employees that the tech poses an existential threat to humanity.

 

With the accord, the administration held firm to its previous insistence that AI companies are responsible for policing themselves, but the document signaled a new openness to some guardrails. It suggested that AI companies “should” implement four layers of controls and noted that “it may make sense to codify” its recommendations into law or regulation in the future.

 

The embrace of third-party evaluators marks a shift in the administration’s posture toward AI safety. While the industry has clamored for using independent assessors to vet their most advanced models, Trump administration officials have regarded the groups best-suited to conduct those reviews with skepticism over their ideological leanings.

 

Ties between many safety reviewers and a movement known as effective altruism had earlier threatened to turn third-party evaluators into a nonstarter with Trump. Popular in Silicon Valley, the movement’s adherents seek to ground every decision in mathematical probabilities they believe would maximize benefit for other people, and supporters generally favor slowing AI development over fears it could lead to human extinction.

 

Trump’s former AI czar, David Sacks, who was present at the White House lunch on Tuesday, has dismissed outside assessments as “pseudoscience.” Even so, some of the firms poised to conduct those reviews have already met with the Center for AI Standards and Innovation, a unit inside the Commerce Department, Bloomberg News previously reported. 

 

Beyond independent evaluators, the accord recommends that companies take safety precautions by implementing internal controls for risks related to cybersecurity, biosecurity, chemical threats and unintended actions by AI models. Companies should also build an internal team to monitor those controls and designate an independent committee of the board of directors to oversee safety teams and independent audits, it says.

 

Despite Tuesday’s nod at safety concerns, the images of Trump ingratiating himself with the group of billionaire executives whom he called the “the most brilliant people in the world” could provide fodder for his political opponents who accuse him of being too close to the tech industry.

 

During his appearance with the executives in the driveway, Trump showed a change of heart toward Amodei, who has long staked out a starkly different position from the president on AI safety. The Anthropic chief’s pledge to slow cutting-edge AI development earlier this month drew a rebuke on social media from Trump, yet the vibe was different Tuesday.

 

Trump said Amodei “has been fantastic” and summoned him to address reporters, leaning closer to him and admonishing him to “be careful.” Amodei,  who met with the president for a private dinner on Sunday night, embraced Trump’s call to preserve the US lead in AI yet indicated that the core safety questions remained unresolved.

 

“The technology has very real risks,” Amodei said. “We all need to work together to make sure that we can win and we can win safely if we do this right,” he added before stepping out of the camera’s view as Trump complained about “fake news.”