Trump, AI CEOs sign voluntary accord on safety reviews

President Donald Trump and top executives from major artificial intelligence companies signed a voluntary accord at the White House on Tuesday that calls for internal controls, independent external audits and board-level review of AI safety reports.

Signatories included Anthropic CEO Dario Amodei, Google CEO Sundar Pichai, Meta CEO Mark Zuckerberg, OpenAI executive Greg Brockman, Nvidia CEO Jensen Huang and Elon Musk of xAI. House Speaker Mike Johnson attended the meeting. The text of the accord was posted on the president’s social media account.

The document sets four voluntary practices: implement robust internal controls, hire an independent external auditor to test those controls, create a committee within each company’s board to review auditor reports, and make the agreement “morally binding” with plans to name an overseer. The accord notes that “over time, it may make sense to codify these steps into laws and regulations.”

Outside the West Wing, Trump told reporters he saw “tremendous self-policing” among companies and added he would “never stifle the growth of a technology that will be bigger than the industrial revolution.” At the event the White House introduced an AI-powered government chatbot called America.gov, which administration officials said could eventually help Americans apply for services such as first-time passports online.

Anthropic’s Dario Amodei warned the technology “has very real risks” and said mechanisms to address those risks remain under discussion. OpenAI has paused the rollout of a new model while evaluating safety; the company’s leaders have stated they are pacing development and sometimes choose not to train a model.

The accord also addressed community concerns about data centers and other infrastructure needed to run advanced AI systems. Companies have invested heavily in data centers and computing equipment, a trend that has increased local projects and investment but prompted bipartisan opposition in some communities. Trump said executives agreed to provide more support for local schools or take steps to reduce energy costs and cautioned that continued opposition could push projects overseas.

Experts reacted with mixed views. Shri Narayanan, a professor of electrical and computer engineering at the University of Southern California, described the accord as a balance between innovation and measures to improve model safety and security. Robin Jia, a computer science researcher, cautioned that voluntary steps may fall short and pointed to recent hacking disclosures as evidence of growing risks. Alex Pascal of the Berkman Klein Center called for legal liability, regulation and changes to competitive dynamics among leading AI developers.

The accord leaves open the option of formal rules. Trump framed AI development as a global competition and said the United States must lead the field. Officials noted U.S. and Chinese authorities recently agreed to a channel for handling AI-related incidents, while Trump indicated limited willingness to deepen cooperation with China.

Participants and the White House document acknowledged that many of the practices the accord recommends are already in place at some firms. The document emphasizes independent audits and board-level oversight as additional layers of accountability as companies develop more powerful models and expand the infrastructure required to run them.