Anthropic Allows Glasswing Partners to Share Mythos-Based Findings

Anthropic has revised Project Glasswing policies to allow participants to share findings, tools, and code developed with the Mythos model to promote broader defensive coordination; the article discusses benefits for collective vulnerability remediation, challenges in triaging AI-generated disclosures, the need for automated validation pipelines, and concerns about potential weaponization and competitive pressure from OpenAI's Daybreak.