Anthropic Lets Claude Mythos Spread Its Glasswings - Gizmodo

The company described the release as allowing the model to 'spread its glasswings.' The model is being tested in a limited capacity under controlled conditions. Claude Mythos is the latest addition to Anthropic's Claude series of large language models. The expansion includes unspecified new partners beyond ENISA, and the preview began on an undisclosed date.

Anthropic stated that the model is not yet ready for primetime and remains restricted to a controlled base of users. The company did not reveal the full list of partners or provide a timeline for broader availability. According to Anthropic, the preview is intended for evaluation and feedback from trusted organizations. The company has not disclosed the specific capabilities or training details of Mythos. The controlled base of users is intended to mitigate risks associated with premature deployment.

The limited preview continues Anthropic's established practice of cautious AI deployment. The partnership may influence future regulatory approaches to AI certification and safety. Anthropic has indicated that it will provide further updates as testing progresses. The company has previously used similar controlled previews for earlier Claude models.