Anthropic's AI Block Becomes an HR Problem Overnight
Few decisions travel from a national-security memo to an HR to-do list overnight. This one did.
A US order now bars foreign nationals – including Anthropic's foreign-born engineers – from its two most advanced AI models, turning access into a workforce problem before it is a policy one.
Anthropic complied, while flagging the scale of the precedent.
"If this standard was applied across the industry, we believe it would essentially halt all new model deployments," the company says in a statement, describing the episode as a misunderstanding it expects to resolve.
The directive, reported by Le Monde, requires the company to cut off access to its Fable 5 and Mythos 5 models for "any foreign national, whether inside or outside the United States".
That wording reaches the company's own foreign-born staff, and the US Commerce Department has cited national-security concerns.
Anthropic, the company behind the Claude assistant, says it had no alternative once the order arrived.
Days from launch to lockout
Fable 5 launched on 9 June and, within days, some of the engineers who built it could no longer open it simply because of the passports they hold.
That is what HR leaders are now wrestling with: flagship tools that have to be segregated by nationality, overnight, across teams that were never organised that way.
What makes the order so striking is what these models can do.
Mythos 5 earned its reputation by sniffing out cybersecurity vulnerabilities faster than the specialists who hunt them for a living, and Fable 5 is its consumer-grade sibling.
"AI models are now tools of global and national strategic consequence," Anthropic CEO and Co-Founder Dario Amodei writes in a recent policy essay.
That essay, published days before the order, set out the company's own case for government oversight.
"The government should have the power to block or deter deployment of the model if it is determined, in light of third-party assessment, to present unacceptable risks," Dario writes.
The power, he adds, "must be scoped to [four] specific risks and there must be protective measures against political favouritism or arbitrary decisions."
For people teams, that leaves a thicket of questions: who counts as a foreign national, how is that status verified, and how is access pulled for some engineers but not others without breaching employment law.
None of it was in last week's plan.
Unglamorous as it sounds, the work is mostly mapping and matching.
HR identifies which projects touch the two models and cross-checks the immigration status already in its systems.
IT and security then cut access, fast enough to comply yet careful enough not to sort the workforce into tiers.
Beyond Anthropic, a new discipline forms
Foreign-born engineers make up a large share of the talent at every major US AI lab, much of it on work visas. Any similar order would ripple straight across the sector.
The same exposure sits inside Microsoft, Meta, Google and IBM, all of which run large AI programmes staffed in part by foreign nationals.
Should the approach spread, their HR and legal teams face the identical job of deciding who can touch which model.
A new compliance discipline is taking shape, and HR sits at the centre of it.
Restricting model access by nationality, fairly and lawfully, will become a capability that frontier labs and large enterprises are expected to have ready.
Anthropic, for its part, has framed its compliance as cooperative and expects the matter to be resolved through dialogue.
"Any government blocking should follow a process that is transparent, fair, clear and grounded in technical facts," the company says.
Access to the most powerful tools is now a question of who holds which credentials, and HR helps decide the answer. The reason runs deeper than any single company.
Dario says: "AI models are now tools of global and national strategic consequence."
