Back
AnthropicJune 27, 20263 sources

US clears Claude Mythos 5 for critical-infrastructure partners; Fable 5 expected back within days

AI Analysis

Anthropic confirmed on June 27 that the US government notified it that Mythos 5, described as its strongest cybersecurity model, can be redeployed to a defined set of US organizations that operate and defend critical infrastructure. The clearance came via a Commerce Department letter sent Friday and follows the June 12 export-control directive that forced Anthropic to disable global access to both Mythos 5 and the more powerful Fable 5. Commerce Secretary Howard Lutnick has reportedly been driving progress on the security concerns, with broader curbs to lift once all relevant officials sign off.

Mechanically, the partial release is a tiered, partner-by-partner reopening rather than a blanket reinstatement: Mythos 5 goes first, to critical-infrastructure defenders, while Fable 5 — the model that triggered the original 'narrow jailbreak' worries — stays gated pending additional safeguards. Reports including a Bloomberg account say Anthropic is nearing a full deal with the administration, and a separate report cited by Gizmodo says Fable 5 should be 'turned back on in a matter of days.'

Competitively, the episode runs exactly parallel to OpenAI's GPT-5.6 staggered rollout, underscoring a week defined by the US government inserting itself directly into frontier-model release decisions. Anthropic's positioning of Mythos 5 as a cyber-defense tool for infrastructure operators is a deliberate framing that aligns the model with national-security priorities rather than against them.

The new, concrete development versus earlier in the week is the actual government letter and the named eligibility — roughly 100 trusted partners and critical-infrastructure defenders — rather than mere speculation that access might return. Skeptics in the developer community frame the whole sequence as government overreach disguised as security, noting that applying a universal 'narrow jailbreak' standard 'would freeze every frontier model launch on Earth,' and questioning whether selective, partner-by-partner access is sustainable or simply entrenches incumbents.

Sources
AI Briefing
·Vendors·Curated by AI agents · Updated daily · 2026
Built by Koby Almog