Anthropic starts to restore Claude Fable 5, Claude Mythos 5 after US export controls lifted

Anthropic has started rolling out Claude Fable 5 and Claude Mythos 5 after the US government lifted export controls on the model. Anthropic said Fable 5 will be available immediately on its platform with hyperscale cloud access available "as quickly as possible."

The company noted that Mythos 5 was enabled for a set of US organizations on June 26 and the company is working with the US government to expand access in its Glasswing program.

Although it's a good development that Fable 5 will be available on Claude's platform across services like Claude Code and Claude Cowork as well as Max, Team and Enterprise plans, the biggest takeaway is that US government approval of frontier models lacks a set process. Models can be launched and they can be pulled at a moment's notice. Enterprises need to plan accordingly and may need to either leverage open models they can own or stay away from the bleeding edge. Refraining from the bleeding edge may make sense anyway due to costs.

Another key takeaway from Anthropic's carefully worded post is that other models could also pull off the same jailbreak that sidelined Fable 5. Anthropic said Amazon researchers found a way to bypass Fable 5's safeguards to identify software vulnerabilities. Anthropic said its testing found that other models had the same ability including Claude Haiku 4.5, Sonnet 4.6, Opus 4.6, Opus 4.7, Opus 4.8, GPT-5.4, GPT-5.5 and Kimi K2.7.

That disclosure about jailbreak capabilities in other models sparked a lot of debate and commentary. The gist: Yes, there may be a framework for releasing frontier models, but the thrash that ensued now means US model providers will be more conservative and more likely to be eclipsed by Chinese models.

• Anthropic launches Claude Sonnet 5
• Anthropic says Mythos 5 gets approval for small set of select customers
• Anthropic’s Fable 5 and Mythos 5 dispute: Still a mess
• Anthropic shuts down Fable 5, Mythos 5 as US govt shuts down foreign national access
• Anthropic’s Dario Amodei wants governments to have the power to block ‘dangerous’ AI systems
• Anthropic’s Claude Fable 5, Mythos 5: What you need to know

Here's what you need to know:

• Fable 5 will be included for up to 50% weekly usage limits through July 7 and then it will be under usage credits.
• Access on AWS, Google Cloud and Microsoft Foundry are being restored.
• Anthropic said "although we have reached a constructive resolution, these events have made clear that the industry needs a consistent way to assess and fix potential “jailbreaks” of AI models (techniques that bypass a model’s safeguards)."
• Recall jailbreak concerns about Fable 5 and Mythos 5 are what led to the export controls. The catch is there's no shared standard for judging the severity of a jailbreak. Anthropic, Amazon, Microsoft, Google and Glasswing partners are working on a framework.
• There will be government "collaboration," which probably looks like regulation to most of us. "We’re also strengthening our level of collaboration with the US government on new pre-release testing, information sharing, and research collaboration," said Anthropic.
• This approach seems to be in play with OpenAI’s latest release.

More:

• Fable 5 and the Shift From Model Capability to Capability Governance
• My internal monologue reading Anthropic's big idea posts
• Claude Mythos and the New Cybersecurity Operating Model

The release framework going forward

Anthropic wrapped together existing processes and new systems to release frontier models that may impact national security. Anthropic proposed:

• Model pre-releases for government access and evaluation. Anthropic said it will expand early access to government partners to independently evaluate capabilities and guardrails. Anthropic will provide technical staff to work with government evaluators.
• An information sharing system on safeguards. Anthropic proposed a system that rhymes with what happens in enterprise software. Significant jailbreaks or misuse patterns will be quickly identified, investigating, triaged and shared.
• Dedicated resources for joint research. Anthropic will provide dedicated teams to work on "shared government priorities." Another wrinkle: Anthropic will "provide a significant compute allocation to support government testing and research."
• The industry will align on a security evaluation standard for frontier model providers.