Anthropic's Mythos and Fable Models Cleared for Global Access After Three-Week U.S. Export Control Saga

Three weeks after the Trump administration placed Anthropic's Mythos 5 and Fable 5 models under U.S. export controls, the restrictions have been fully lifted — setting off a cascade of product launches and competitive responses that reshaped the frontier AI landscape in late June and early July 2026.

The Export Control Episode

The Commerce Department added both models to its list of export-controlled technologies on June 12, 2026, triggering an immediate and sweeping response from Anthropic: the company disabled all access to Mythos and Fable entirely, including for its own foreign national employees. The directive was rooted in two separate security concerns. According to Wired, the White House acted after learning Anthropic had granted access to a South Korean telecommunications firm believed to have ties to China. Separately, Amazon and the National Security Agency raised red flags to the administration over Fable 5's susceptibility to jailbreaking.

A partial reversal came on June 26, when Commerce Secretary Howard Lutnick sent a letter to Anthropic co-founder Tom Brown approving restored access for more than 100 select U.S. government agencies and companies designated as trusted partners. Fable 5, however, remained blocked at that point. The Verge reported on June 26 that the model appeared to still be in limbo with no clear rollout timeline — a status that persisted until the full ban was lifted.

Full Clearance and Simultaneous Launches

On July 1, 2026, the U.S. government removed its export licensing requirement for both models, restoring global access to Mythos and Fable. Anthropic moved quickly. The same day, the company released Claude Sonnet 5, priced at $2 per million input tokens through an introductory window ending August 31, 2026. Anthropic also launched Claude Science, an AI workbench purpose-built for researchers that integrates with more than 60 scientific databases — a product aimed squarely at professional and academic scientific workflows rather than general-purpose use.

Competitive Responses During the Blackout

The three-week export control window accelerated activity among non-U.S. competitors who framed the moment as an opportunity. China's Zhipu AI (Z.ai) released GLM-5.2, an open-weight model that researchers claimed matches Mythos performance in select cybersecurity and bug-finding benchmarks. Chinese cybersecurity firm 360 separately unveiled Tulongfeng, an AI tool its developers positioned as Mythos-equivalent in capability, reportedly on June 26. Meanwhile, Tokyo-based Sakana AI — co-founded in 2023 by former Google researchers David Ha and Llion Jones, alongside ex-Mercari and Stability AI executive Ren Ito — launched Fugu, which the company described as standing "shoulder-to-shoulder with leading models like Anthropic's Fable 5 and Mythos Preview."

OpenAI Enters with GPT-5.6

OpenAI also moved during the same period. On June 28, the company announced GPT-5.6, a family comprising three models — Sol, Terra, and Luna — with limited preview access initially restricted to select U.S. partners. OpenAI's own benchmarking claims GPT-5.6 Sol is competitive with Mythos Preview on ExploitBench while using roughly one-third of the output tokens, a claimed efficiency advantage the company is positioning as meaningful for production deployments.

The sequence of events — a blunt export control directive, a partial administrative walkback, competitive launches filling the vacuum, and then a full reversal paired with new Anthropic products — illustrates how quickly geopolitical actions can restructure near-term competitive dynamics in frontier AI. Fable 5's delayed clearance relative to Mythos remains the one unresolved thread: while the July 1 lift covered both models formally, the gap between June 26 and July 1 meant Fable 5 spent an additional five days in regulatory uncertainty that its competitors did not hesitate to exploit.