Anthropic CEO Replaced by Co-Founder in White House AI Talks Amid Export Control Dispute
Dario Amodei's departure from high-stakes meetings reportedly follows concerns over his engagement style, as U.S. officials maintain export controls on new AI models.
Anthropic CEO Dario Amodei has been replaced by co-founder Tom Brown in critical meetings with White House officials regarding the company's advanced artificial intelligence models. The shift in representation reportedly aims to facilitate more productive discussions after Amodei was perceived as difficult to engage with during previous high-stakes negotiations, according to a report.
The U.S. government recently imposed strict foreign export controls on Anthropic's new AI bots, "Mythos" and "Fable," prompting swift engagement from the AI giant in Washington. However, Amodei's direct involvement was reportedly met with challenges, with one source familiar with the calls telling Wired that Brown "can actually engage" without being "a weirdo."
Since Brown and Sarah Heck, Anthropic's public policy chief, took over leading the discussions, talks with Trump administration officials have reportedly improved. The White House and Anthropic have not immediately responded to requests for comment.
AI Model Export Controls and National Security Concerns
The export controls were imposed earlier this month, following a warning from Amazon CEO Andy Jassy about evidence suggesting the AI models' safety guardrails could be bypassed. In response, Anthropic temporarily pulled the models offline on June 12, citing the need to comply with the stringent new regulations.
Concerns over the AI models' capabilities were heightened by tests conducted by the National Security Agency. The NSA found that Mythos could identify vulnerabilities in secure government systems within hours, though it does not necessarily imply exploitation within the same timeframe. These tests were part of Anthropic's Project Glasswing initiative, which partnered with various entities to identify potential safety flaws in its AI tools, amidst broader concerns about the advanced nature of models like Mythos and Fable.
Ongoing Negotiations and Regulatory Deadlines
Discussions between Anthropic and government officials are ongoing, with a focus on what evidence the company can provide to alleviate concerns about potential "jailbreaks" of the AI models. Commerce Secretary Howard Lutnick, whose department oversees export controls, faces a June 26 deadline from a bipartisan group of lawmakers demanding answers regarding the future path forward for such AI regulations.
Anthropic has reportedly been working to address these security concerns and has pledged closer collaboration with the White House in a proposal submitted to Lutnick. Previously, tensions arose earlier this year when Amodei reportedly declined to grant the Pentagon unrestricted access to his AI bots during contract negotiations, seeking limitations on their use in mass surveillance and weaponry. The company had argued that the export restrictions were unnecessary, stating it had flagged potential risks within its own bots. However, White House officials were reportedly displeased that Anthropic downplayed the safety risks of its models, especially after years of warnings from Amodei himself about the broader catastrophic risks associated with other AI models.