Anthropic Restores Limited Access to AI Model Mythos 5

US Government Grants Partial Approval for Mythos 5 Deployment

Anthropic, the San Francisco-based AI company behind the Claude chatbot, has received authorization from the US government to restore limited access to its most advanced cybersecurity AI model, Mythos 5. Approximately 100 'trusted partners' — including select cyber defense organizations and critical infrastructure operators — can now use the model, marking a significant easing of export restrictions imposed earlier in June 2026.

The development follows a two-week standoff after the US Commerce Department, citing national security concerns, ordered Anthropic to suspend all access to both Mythos 5 and Fable 5 by any foreign national. Anthropic complied by disabling the models entirely, affecting even domestic users. Now, with a formal letter from Commerce Secretary Howard Lutnick, the company can resume limited deployment.

Background: Why Was Mythos 5 Restricted?

On June 12, 2026, the US government issued an export control directive under national security authorities, demanding Anthropic block foreign nationals from accessing Mythos 5 and Fable 5. The government believed it had discovered a method to jailbreak Fable 5, though Anthropic argued that the identified technique only exploited minor, previously known vulnerabilities — ones that other publicly available models could also detect without a bypass.

Mythos 5 is Anthropic's most powerful model for cybersecurity, capable of rapidly identifying weaknesses in software security. In the wrong hands, the AI could be weaponized for offensive cyber operations, which prompted Washington's intervention. The US AI export control policy has become a central tension point between innovation and national security.

What the Partial Restoration Means

The Commerce Department's approval applies only to Mythos 5 and only for a curated group of vetted organizations. According to Anthropic's announcement on X, the company is working closely with the government to expand access further. However, Fable 5 — a scaled-down version of Mythos designed for broader commercial use — remains fully restricted.

Howard Lutnick's letter to Anthropic stated: 'Anthropic has, in collaboration with the US government, implemented measures to mitigate the risks associated with the models in question.' The letter did not address any changes to Fable 5's restrictions.

Comparison: Mythos 5 vs. Fable 5

Industry Context: OpenAI Faces Similar Restrictions

Anthropic is not alone in navigating government oversight. Rival OpenAI rolled out its GPT-5.6 lineup — including flagship model Sol, mid-tier Terra, and budget Luna — on June 26, 2026, but only to 20 government-approved 'trusted partners' at the Trump administration's request. OpenAI expressed dissatisfaction, stating this process 'shouldn't become the long-term default.'

The OpenAI GPT-5.6 government restrictions mirror the challenges Anthropic faces, highlighting a broader trend of the US executive branch asserting control over advanced AI deployment. Critics argue this creates a de facto licensing regime without clear statutory authority.

What's Next for Anthropic?

Anthropic has said it is 'pleased' that Mythos 5 is partially available again and continues to negotiate with the government to expand access and restore Fable 5 for general use. The company is also embroiled in a separate legal dispute with the Department of Defense, which designated Anthropic a 'supply chain risk' after the company refused to remove contractual restrictions against using its AI for domestic surveillance and autonomous weapons.

The Anthropic Department of Defense dispute adds another layer of complexity to the company's relationship with the federal government. Meanwhile, Anthropic's valuation has soared to an estimated $965 billion, making it the most valuable pure-play AI company globally.

Frequently Asked Questions

What is Anthropic's Mythos 5 AI model?

Mythos 5 is Anthropic's most advanced AI model focused on cybersecurity. It can rapidly detect vulnerabilities in software, making it a powerful tool for both defensive and, potentially, offensive cyber operations.

Why did the US government restrict Mythos 5?

The government cited national security concerns, believing the model could be jailbroken and misused by foreign adversaries for cyberattacks. An export control directive on June 12, 2026, ordered Anthropic to block foreign national access.

Who can now access Mythos 5?

Approximately 100 'trusted partners' — including cyber defense organizations and critical infrastructure operators — have been approved by the US Commerce Department to use Mythos 5.

Is Fable 5 also available?

No. Fable 5, a less powerful version of Mythos, remains fully restricted. The government has not approved any access restoration for that model.

How does this compare to OpenAI's situation?

OpenAI's GPT-5.6 models were also restricted to a small number of government-approved partners upon release, reflecting a broader US policy trend of vetting advanced AI models before broad deployment.

Sources

• Anthropic official blog post on Fable/Mythos access (June 2026)
• Bloomberg — Letter from Commerce Secretary Howard Lutnick
• CNBC — 'US government grants Anthropic permission to release Mythos 5'
• Reuters — 'US releases Anthropic model Mythos to some US companies'
• TechCrunch — 'OpenAI limits GPT-5.6 rollout after government request'
• Wikipedia — Anthropic company profile