Anthropic Expands Project Glasswing to Share AI Cybersecurity Findings

Anthropic has revised its Project Glasswing initiative, enabling partners to communicate AI cybersecurity vulnerabilities to regulators, industry organizations, and the public in accordance with responsible-disclosure guidelines. Previously, such findings were confined to a controlled-access group. This update seeks to enhance transparency and foster collaboration in AI safety. As an AI research firm, Anthropic manages Project Glasswing as a vulnerability disclosure program centered on large language models. The updated policy allows for wider sharing of threat intelligence while upholding responsible disclosure standards, aligning with industry efforts to create uniformity in AI security reporting.