Anthropic ban exposes workforce risk HR leaders didn’t see coming

An executive’s security worry triggered a federal directive that cut off access to Anthropic’s most advanced AI models for users worldwide, according to reporting in the Wall Street Journal. Amazon CEO Andy Jassy raised concerns with senior administration officials, including Treasury Secretary Scott Bessent, about researchers using prompts to extract information from Anthropic’s Claude Fable 5 that could aid cyberattacks.

White House officials convened, security researchers began testing Amazon’s claims and within days, President Trump signed off on a global freeze preventing foreign governments, companies and individuals from accessing Fable 5 and Mythos 5.

Anthropic said many of its researchers are foreign-born, so the government directive effectively blocked them from working on the models they were hired to build, CNBC reported. To comply with the order, Anthropic shut off access to both Fable 5 and Mythos 5 for all users worldwide. The rapid shutdown involved the Commerce Department, National Cyber Director Sean Cairncross and Commerce Secretary Howard Lutnick.

Read more: OpenAI, Anthropic CEOs walk back AI job warnings as IPOs loom

The administration had long questioned whether Anthropic could be trusted to manage the security risks its models presented, according to people familiar with the matter, as reported by the Wall Street Journal. A recent call between officials and Anthropic CEO Dario Amodei reinforced those doubts, with some on the call saying the company signaled it was unwilling to work with government security experts.

According to reporting from Nextgov, the directive’s language suggests no “critical worker” exceptions exist, though this interpretation should be confirmed with legal counsel.

‘Building on solid ground’

Hung Lee, founder of Recruiting Brainfood and a keynote speaker at HR Tech Europe, has tracked Anthropic’s standing closely. He wrote on LinkedIn that the company’s reputation had been built on responsible AI development, but that credibility eroded quickly once its most powerful model became a flashpoint.

“The lines between what was unsafe vs. uncompetitive blurred,” Lee says, “with growing disquiet that a single company could appoint itself as gatekeeper of transformatory technology.”

The ban also raises a governance question about who in the organization is choosing the AI tools the workforce depends on. If procurement or IT is selecting cloud-dependent platforms for HR workflows without input from HR leadership, eventually, the CHRO might hit road bumps.

Lee frames this as a strategic choice between owning and renting intelligence. “If you rent AI, you’re building your business on a house of sand,” he says. “If you own AI, which may be less capable at the frontier, you’re building at least on solid ground.”