Fable 5’s launch becomes a test of access safeguards

Anthropic launched Claude Fable 5 and Claude Mythos 5 on 9 June with different safeguards around the same underlying model. The launch announcement offered Fable for general use while restricting Mythos to a small group of defensive cybersecurity and infrastructure partners through Project Glasswing.

Fable could route some requests to Opus 4.8 when its safeguards detected risk. Anthropic said the restrictions were deliberately conservative and could catch harmless requests. The advertised API price for both models was $10 per million input tokens and $50 per million output tokens.

The launch page records access being suspended on 12 June. In its subsequent account of the suspension and restoration, Anthropic says a US export-control directive restricted foreign-national access and that it lacked a reliable way to verify nationality immediately, so it suspended both models for everyone.

That account says the controls were lifted on 30 June and global Fable access resumed on 1 July. Anthropic introduced an updated safety classifier and acknowledged that it would also block more benign coding and debugging requests. Mythos access remained subject to partner restrictions. The chronology shows how an announced capability can become unavailable, then return with changed controls; customers’ operational access cannot be inferred from a model’s launch date alone. For Indian organisations, a globally offered model could still be interrupted by a restriction originating in another country. This record follows the availability changes.