- Anthropic has made its Mythos model publicly available as Fable 5, warning users that the system is not risk‑free.
What actually happened: After months of internal debate over the safety of its Mythos large‑language model, Anthropic released a trimmed‑down version named Fable 5. The company bundles the model with a set of content‑filtering guardrails designed to block disallowed output. At launch it published a brief risk notice saying the model “still comes with risks” and may generate harmful or inaccurate information. No pricing or performance metrics were disclosed in the announcement.
Why it matters: By opening a previously internal model to the public, Anthropic is testing how well its safety layers hold up in uncontrolled environments. The explicit risk disclaimer signals a shift from private research to broader deployment, putting the onus on developers to implement additional safeguards. It also gives competitors a reference point for how far safety engineering can be pushed before a model is deemed “release‑ready.”
The move mirrors other firms that have leaked advanced models despite safety concerns, reminding readers that a public label does not equal a clean bill of health.