Commerce Secretary Howard Lutnick cleared Anthropic's two most advanced AI systems, Claude Fable 5 and Mythos 5, for overseas access as of Wednesday, but only after the company agreed to ongoing national security monitoring.
Anthropic's two most advanced AI systems, Claude Fable 5 and Mythos 5, are back on the export market as of Wednesday after Commerce Secretary Howard Lutnick confirmed the lifting of a licensing requirement that had blocked overseas access to the systems earlier this year. The reversal came after Anthropic agreed to ongoing national-security commitments that will reshape how its flagship models are deployed, and the same day the agency scrapped the Biden-era AI Diffusion Rule, the late-2024 framework that capped how advanced AI chips and model weights could flow abroad.
Lutnick confirmed the move in comments carried by CNBC, and the Bureau of Industry and Security (BIS) issued the formal rescission of the export license through its own press release, which framed the action as part of a broader reset of U.S. AI export controls. Anthropic said separately that customers will regain full access to Claude Fable 5 and Mythos 5 starting Thursday, July 2, 2026. The split between policy reversal and operational restoration is deliberate: the legal change takes effect on the announcement, the technical unblocking of accounts follows the next day.
BIS's parallel rescission of the Biden-era AI Diffusion Rule, the late-2024 framework that capped how advanced AI chips and certain model weights could flow abroad, looked like a separate story at first. The two moves were entangled from the start: the export-license regime applied to Anthropic and the diffusion framework both rested on the same legal premise, that frontier AI was itself an export-controlled commodity. Unwinding one required unwinding the other.
What Anthropic actually gave up is spelled out in its own update and on the AnthropicAI account: ongoing collaboration with the government on AI security standards and active monitoring for malicious use of its models. That is a softer mechanism than the licensing regime it replaces, but it changes the steady state. The periodic Commerce veto becomes a standing obligation. Anthropic will be watched, not gated.
The reverse holds as well. Earlier in the standoff, Anthropic sued the Department of Defense over AI tool-use concerns, a fact reported by CNBC and Forbes. That case has not been publicly resolved at the time of today's announcement. Reading today's reversal as a clean industry win elides the fact that Anthropic felt compelled to litigate while negotiating. The relationship work that helped reopen the channel included Anthropic co-founder Tom Brown, whose prior government ties helped rebuild the negotiating channel with Washington, per Forbes.
The bigger question is what comes next. If today's deal structure holds, the next time a frontier AI product launches, Washington will not hold a license it can revoke; it will hold a collaboration agreement that has to be navigated. Framing from Al Jazeera leans on the official line that this is the routine end of an over-broad order. Critics quoted in the same coverage argue the opposite, that pulling the controls shows the original order was regulatory overreach. Both readings can be true. The remaining test is whether the next time a frontier AI product crosses a security threshold, the answer is a renewed Commerce veto, a renegotiated compact, or a fresh lawsuit.