U.S. President Donald Trump has reached an agreement with Silicon Valley leaders that supports having external organizations evaluate the safety of artificial intelligence systems, aiming to address growing concerns about AI risks without introducing new regulations.
Trump hosted nearly twenty technology industry executives for lunch at the White House on Tuesday, during which the agreement was reached. The attending executives also pledged to hold regular meetings to jointly set an industry example.
Although the document is not legally binding, Trump told reporters it carries "moral force." "In a sense, it's almost like a constitution," Trump told reporters on the White House driveway after the event. "The most important people in the world signed it, and I signed it as president. It truly is a form of protection."
The agreement was reached after nearly two hours of talks between Trump and his guests, amid mounting global concerns over the potentially catastrophic risks of AI.
Signatories include leaders of the most prominent AI development companies: Anthropic PBC CEO Dario Amodei, Nvidia Corp. CEO Jensen Huang, and Elon Musk of SpaceXAI. According to the agreement text posted by Trump on his Truth Social platform, other signatories include OpenAI President Greg Brockman, Meta Platforms Inc.'s Mark Zuckerberg, and Sundar Pichai, CEO of Google parent company Alphabet Inc.
Several companies, including Anthropic and OpenAI, had previously committed to hiring independent evaluators. Under this agreement, the Trump administration maintains its previous position that AI companies should primarily rely on self-regulation to ensure safety. However, the document also signals a new openness to establishing certain safety guardrails.
The agreement recommends that AI companies "should" establish a four-layer safety control mechanism, and notes that it may become necessary in the future to formalize relevant recommendations into laws and regulations.
Accepting third-party evaluation organizations' involvement in reviews marks a shift in the Trump administration's stance on AI safety issues. Although the industry has been calling for independent evaluators to assess the most advanced AI models, Trump administration officials had previously been reserved about which organizations would be most suitable for such work, fearing they might carry specific ideological biases.