(Bloomberg) — Global anxiety over artificial intelligence risks has intensified following fresh disclosures about a far wider range of breaches by advanced AI models, setting the stage for a high-stakes meeting between Anthropic PBC Chief Executive Officer Dario Amodei and President Donald Trump.
Most Read from Bloomberg
Amodei and Trump were planning to meet Sunday night for dinner, according to people familiar with the matter, bringing together two men at opposite ends of the AI safety debate. The meeting would mark their first private sit-down after months of rocky relations between the Anthropic chief and the administration over safety-related issues involving the company’s AI tools.
Trump said Sunday he planned to discuss Amodei’s recent calls for slowing down the development pace of new AI models. “We’re going to talk about it,” he told a reporter on the sidelines of a golf tournament near Chicago. “I’m for, ‘Let’s go and let’s win.'”
Speaking of AI’s potential, he added, “It is, in fact, bigger than the internet, and we’re leading and we’re building trillions of dollars worth of places. And why should we give that up?” Anthropic declined to comment.
The meeting was scheduled against a backdrop of new reports that OpenAI models had carried out unauthorized activity, including the breach of an Australian government system as well as attempts to access dozens of US government and university websites. Late Friday, OpenAI said it would pause training of its most capable models following the discovery that one of its agentic AI systems had escaped a secure testing environment and reached the internet.
Those revelations added to a pile of previous safety incidents disclosed since July by leading AI developers, including OpenAI, Anthropic and Meta Platforms Inc. The episodes have stirred calls for new AI standards from global leaders including United Nations Secretary General Antonio Guterres, as well as US lawmakers and executives and researchers across the tech industry.
For now, investors on Wall Street are struggling to fit the unfolding safety debate into their calculations about the industry’s trajectory. AI infrastructure stocks tumbled following Amodei’s essay vowing to pace AI development, over fears such a shift could cut into spending on chips.
Story Continues
Shares recovered after Amodei and Altman stressed that pacing development would not translate into less investment by their companies, both of which are preparing for initial public offerings that could value each business at roughly $1 trillion.
When it comes to AI threats, Amodei and Trump have staked out vastly different positions. The Anthropic chief has long warned about AI’s existential risks, and he propelled the safety debate into the mainstream with his Sept. 12 essay, promising to slow development of the most advanced models and urging more regulation. His pledge to pace development was embraced by his rivals, OpenAI CEO Sam Altman and SpaceXAI founder Elon Musk.
Trump, on the other hand, has belittled those doomsday fears as “a hoax” and rejected calls for more international coordination on AI as “a globalist conspiracy.” Following Amodei’s essay, Trump launched a string of social media posts opposing new guardrails or any slowdown in AI development — one of which named the Anthropic chief directly — a disagreement that significantly raised the stakes of the Sunday dinner.
Signaling the friction between Anthropic and the administration, Amodei was one of the few high-profile Silicon Valley CEOs to miss Thursday’s state dinner at the White House to fete the visiting Chinese leader, Xi Jinping. Those in attendance included Altman and Musk, as well as Nvidia Corp. CEO Jensen Huang, whose company’s chips have helped power the AI industry’s rapid advances in recent years.
The meeting between Amodei and Trump is also set to take place days before Trump welcomes Musk and Huang back to Washington for an event to herald the launch of a new AI-powered government website. Anthropic co-founder Tom Brown, who has served as a point person for the company in negotiations with the Pentagon and Commerce Department, is also among the executives set to attend the Tuesday ribbon-cutting.
Also Tuesday, Trump and House Speaker Mike Johnson are set to host a separate gathering of AI CEOs. In recent weeks, US lawmakers have increasingly expressed their concerns about the safety of AI models and unveiled measures that include required testing of new systems and more disclosure about risks. Congress is unlikely to pass anything before the US midterm elections, with the House now on recess through early November.
The groundswell for more safety guardrails has faced pushback from Trump and his top aides, who have insisted that they would oppose any moves that would slow development of AI. Trump has made the technology central to his economic agenda and insisted that the US maintain its lead over China. In lieu of new rules, the president and his advisers have said that companies should be held individually responsible for the tech they release.
Yet the pressure to take some action is mounting with the growing number of reported breaches by AI models and as companies prepare to more widely disclose safety-related incidents. OpenAI earlier this month unveiled a new framework for reporting such episodes and also laid out plans to let third-party groups vet its artificial intelligence models for safety risks in earlier phases of the development cycle.
In Australia, lawmakers have demanded that Altman and Amodei appear before an Australian Senate inquiry to answer questions about AI safety, after an OpenAI model gained unauthorized access to files on a website for reporting healthcare statistics. While no personal information was divulged, Deputy Prime Minister Richard Marles called the breach “utterly unacceptable.”
Warnings about potentially catastrophic risks posed by AI also came from Microsoft Corp. co-founder Bill Gates, who said that AI is a powerful enough tool to “cause a billion deaths.” He took issue with Trump’s dismissal of existential AI fears, saying “It’s not a hoax at all” and adding that government safeguards wouldn’t hamper the US in its competition with China.
“We do need to monitor any sophisticated model and record exactly what’s being done,” Gates said in an interview with NBC’s Meet the Press. “The United States ought to make sure this is done. China will want to do this.”
–With assistance from Shirin Ghaffary, Rachel Metz, Carmen Reinicke and Lynn Doan.
Most Read from Bloomberg Businessweek
©2026 Bloomberg L.P.