Anthropic's White House Negotiations Improve After CEO Dario Amodei Replaced by Co-Founder

Anthropic Replaces CEO in Critical White House Negotiations

Anthropically has replaced CEO Dario Amodei with co-founder Tom Brown in ongoing White House negotiations aimed at lifting restrictive export control directives on its advanced AI models 1

1

2

2

. For nearly two weeks, high-ranking team members at Anthropic have been locked in discussions with the Trump Administration, attempting to resolve AI model export controls that have effectively shuttered access to Claude Fable 5, the company's most advanced consumer-facing frontier model. According to sources familiar with the high-stakes White House meetings, talks were not progressing well when Amodei was involved, but have improved significantly since Brown took over alongside Sarah Heck, Anthropic's Head of Public Policy 1

1

.

One person involved in the calls told Wired that "Tom Brown is not being a weirdo like Dario and can actually engage," suggesting Amodei's communication style hindered productive dialogue 2

2

. Past reporting has characterized Amodei as someone with a tendency to "rant" and who "cannot control his emotions," while sources claim he wasn't being a good listener during these critical negotiations 1

1

.

AI Model Safety Concerns Trigger Export Restrictions

The regulatory challenges began when the White House issued export controls approximately three days after Anthropic released Claude Fable 5 1

1

. The directive required Anthropic to prevent non-U.S. nationals from accessing both Fable 5 and the even more exclusive Mythos 5 model. The government's concerns intensified after Amazon researchers notified the White House that jailbreaks could supposedly remove safety guardrails easily, potentially exposing cybersecurity vulnerabilities 1

1

.

Before learning about the jailbreak issue, the White House was already worried that potential bad actors affiliated with Chinese interests had accessed the model 1

1

. Anthropic announced it had pulled Fable 5 offline entirely on June 12, claiming this was the only way to comply with the stringent new restrictions 2

2

.

NSA Testing Reveals Significant Capabilities

The National Security Agency had been testing versions of Anthropic's AI tools through Project Glasswing, an initiative that partnered with other companies and agencies to test the bots for potential safety pitfalls 2

2

. The NSA found that Mythos had identified vulnerabilities in highly secure government systems within mere hours, though this doesn't necessarily mean the bots could exploit those sensitivities within the same timeframe 2

2

.

Anthropically calls Fable 5 a "Mythos-class" model, meaning it's built with the same core technology as Claude Mythos Preview, which Anthropic deemed too dangerous for public access 1

1

. While Fable 5 came ostensibly stripped of capabilities that made it a potential cybersecurity threat, the jailbreak vulnerabilities raised serious questions about whether these safeguards were sufficient 1

1

.

Path Forward Remains Uncertain

While talks between Anthropic and Trump officials have reportedly been progressing well since the leadership change, it remains unclear whether the government will lift the AI model export controls anytime soon 2

2

. Some conversations have focused on what evidence Anthropic could provide to alleviate the administration's concerns about potential jailbreaks of the new AI models 2

2

.

Commerce Secretary Howard Lutnick, whose department oversees export controls, faces a June 26 deadline from a bipartisan group of lawmakers who demanded answers about the path forward 2

2

. Anthropic has been scrambling to resolve these security concerns, recently pledging to work more closely with the White House in a proposal to Lutnick 2

2

. Tensions had already been building earlier this year after Dario Amodei refused to give the Pentagon unchecked access to Anthropic's AI bots during contract negotiations, seeking limits for their use in mass surveillance and weaponry 2

2

.