Anthropic says it will allow participating partners to share cybersecurity findings generated with its Mythos model with other parties, following concerns about the sensitivity of such information. According to reports, the change is tied to contractual terms used when Mythos is provided to partner organisations. These agreements include protections intended to address partners’ requests for assurances before they disseminate results that may involve sensitive security observations. The arrangement is designed to give partners a pathway to communicate their Mythos-related cybersecurity findings while maintaining safeguards specified in the agreements. The details of what partners can share, and under what conditions, are described as being included in the participation agreements that govern the model’s use. The move reflects a balance between enabling collaboration in cybersecurity research and managing confidentiality and information-sharing risks. Both outlets report the same overall policy shift: Anthropic is granting partners the ability to share Mythos cybersecurity findings with others, but only within the boundaries set by the agreements.