Anthropic Pitches Lutnick Deal to Lift Mythos, Fable Ban

What did the US government do to Anthropic's Mythos and Fable models?

On June 12, 2026, at 5:21 PM ET, the US government issued an export-control directive ordering Anthropic to suspend access to Claude Fable 5 and Claude Mythos 5. The directive required Anthropic to cut off "any foreign national, whether inside or outside the United States, including foreign national Anthropic employees," according to Snyk's breakdown of the event.

Anthropic said there was no way to separate foreign nationals from US users across a user base of hundreds of millions on same-day notice. So the company pulled both models for every customer worldwide.

Why did the US government impose the export controls?

The trigger was a reported AI jailbreak. Amazon CEO Andy Jassy warned the Trump administration that researchers had found evidence it was possible to bypass the models' safety guardrails, the New York Post reported.

Anthropic pushed back on the severity. The company said the government provided only "verbal evidence of a potential narrow, non-universal jailbreak, which essentially consists of asking the model to read a specific codebase and fix any software flaws." Anthropic also noted that the same capability "is widely available from other models (including OpenAI's GPT-5.5), and is used every day by the defenders who keep systems safe."

The government has not published the directive or provided a written technical basis.

You might also like

Google DeepMind Publishes AI Control Roadmap

Meta Buys 1.6 GW from Crusoe Data Centers

US Warns ASML: China May Have Top EUV Chip Tool

EA Cuts Hundreds: Respawn, Full Circle Hit Hard

Who is leading Anthropic's negotiations with the Trump administration?

As of June 18, 2026, two Anthropic executives are at the center of the talks:

• Tom Brown — Anthropic cofounder and chief compute officer
• Sarah Heck — Anthropic's head of public policy

Both are negotiating directly with Trump administration officials, a source with knowledge of the situation told the New York Post. Anthropic first sent senior officials to Washington DC late the week of June 12, shortly after the export controls were issued.

What is Anthropic offering Commerce Secretary Lutnick?

Anthropic has floated a proposal to Commerce Secretary Howard Lutnick. The core pledge is to improve communication with the Trump administration and resolve any security concerns more quickly going forward, a source familiar with the situation said.

An Anthropic spokesperson stated earlier this week: "Both parties are working quickly to get this resolved. This is part of our ongoing commitment to working alongside the administration toward our shared goal of protecting US critical infrastructure and the US lead in cyber defense."

Talks are described as progressing well. No exact timetable for a permanent fix has been set.

How did this become a G7 issue?

The dispute surfaced as a topic at the G7 Summit in France. President Trump said talks with Anthropic were "going fine" but offered no details. Anthropic CEO Dario Amodei used the summit to urge world leaders to "resist the temptation to splinter" in their approaches to AI regulation. For more on what was discussed at the summit, see our coverage of G7 trusted partners and Anthropic's role in those talks.

The timing carried an ironic edge. The White House crackdown came just days after Amodei had publicly argued that the US government should have authority to shut down frontier AI models it deemed too dangerous. White House officials were also reported to be frustrated that Anthropic and Amodei described the safety flaw as a "narrow" problem, given the company's years of public warnings about AI risk.

What does the jailbreak claim mean for security teams?

Here's what we know so far from the technical record: the reported jailbreak involved asking the model to read a codebase and identify or fix software flaws. Security researchers noted this is standard automated code review — the same work performed by static analysis tools and AI-assisted code review platforms every day.

IBM X-Force's Valentina Palmiotti told TechCrunch, in the days before the suspension, that Fable 5's guardrails were already "too aggressive for legitimate defensive work," rejecting "any request that could be tangentially cyber related." Within the same week, the model was criticized for being too restrictive and then withdrawn over a capability used in defense. This dual-use tension is familiar to anyone working in AI safety research or offensive security tooling.

Snyk's analysis framed the broader issue: a capable coding model that can fix vulnerabilities will also be able to describe them. There is no version of the capability that separates the two. The security field's established response to dual-use risk — coordinated disclosure, defense in depth, risk-based prioritization — does not typically result in a blanket shutdown. When a critical CVE drops in a popular npm package, the ecosystem does not take all of npm offline.

Key timeline of events

The situation also raises questions about how frontier AI governance intersects with broader US-China AI competition and export-control policy. Separately, the race to build models capable of complex reasoning — including code analysis — continues to shape decisions at labs like OpenAI and others.

As of June 18, 2026, the most confirmed next step is that Anthropic's proposal is on the table with Commerce Secretary Lutnick, talks are described as progressing well, and no firm date for restoring access to Mythos 5 or Fable 5 has been announced.