Trump advisers meet Anthropic executive over AI safety

WASHINGTON – Senior Trump administration officials met Anthropic’s top Washington executive on Sept 15 to discuss artificial intelligence (AI) safety risks.

Anthropic’s chief compute officer Tom Brown, who has become the face of the company in Washington, spoke virtually with Pentagon chief technology officer Emil Michael and Commerce Secretary Howard Lutnick, according to people familiar with the matter. 

Anthropic declined to comment.

In a statement, a Department of Defense official said it “maintains contact with frontier AI companies”, but “the department’s stance on Anthropic has not changed”.

Tensions between Anthropic and the administration were reignited by a 3,800-word essay posted on Sept 12 by the company’s chief executive Dario Amodei, who called for government regulation and urged pacing development of the most advanced AI systems so researchers can better understand potential threats. 

Amodei’s missive, which was quickly embraced by the chief executives of rivals OpenAI and SpaceXAI, Sam Altman and Elon Musk, prompted a vehement reaction from US President Donald Trump.

He dismissed fears about AI risks as “a hoax” and rejected the idea of new rules.

Still, leading AI developers have begun discussing AI safety concerns among themselves.

OpenAI’s global policy chief Chris Lehane said his company has been engaging with Anthropic and Google DeepMind for several weeks on the issue.

“It’s better to try to work together to prioritise safety,” Lehane said at a Sept 15 briefing in Washington. 

Mounting concerns about the risks of cutting-edge AI models took centre stage in forums across the US on Sept 15.

At Salesforce’s annual Dreamforce conference in San Francisco, Amodei, Altman and Nvidia CEO Jensen Huang all weighed in on the topic.

Meanwhile, an AI gathering in Washington saw progressive Senator Bernie Sanders and former Trump adviser Steve Bannon share rare common ground as both warned that AI companies are jeopardising human interests. 

At Dreamforce, both Altman and Huang acknowledged concerns, but argued AI companies would safely pace their technology, without needing regulations.

Worries about the potential dangers of advanced AI went mainstream over the past week after the high-profile departure of Anthropic researcher Jacob Coxon, who accused AI companies of “gambling with our lives” in a resignation post he uploaded on social media. 

Michael, who has led Pentagon efforts to accelerate AI adoption across the military, has publicly rejected Coxon’s claim that AI could kill all humans, while Lutnick is eager to push back on negative messaging about the technology ahead of the midterms in November, according to people familiar with the matter. 

Relations between Anthropic and the US government have been strained following a disagreement earlier in 2026 with the Commerce Department over the security of its advanced Mythos and Fable AI models and a tiff with the Pentagon over the use of its tech in military operations.

Michael helped lead negotiations with the company over the military’s use of its tech, but those talks broke down when the Pentagon rejected Anthropic’s terms, calling them an infringement on the military’s authority.

It then blacklisted the firm, declaring it a threat to the US supply chain.

Anthropic is still engaged in an active lawsuit against the Pentagon in a Washington court over the blacklisting.

Brown emerged on the Washington scene earlier this summer, as he led high-stakes negotiations with US officials to lift export controls that were placed on Anthropic’s most advanced AI models.

He succeeded in that task, and spoke publicly alongside Lutnick at a tech-focused Group of 20 summit earlier in September, making a point to praise Trump’s support for data centres. 

In a Bloomberg Television interview from the sidelines, Lutnick said the company had “gotten religion” and patched its relationship with the government, though the Pentagon later said that its own blacklisting of the company remained in place. 

Brown is now likely to face yet another fraught challenge: easing tension with US officials over AI safety at a time when the President is rejecting such concerns.

Trump said in a post on his Truth Social network on Sept 14 that his presidency is the only safety guardrail needed for AI. He added that slowing down AI is a bad idea that would only serve US adversaries like China. BLOOMBERG