Project Glasswing is widely discussed as a defensive AI cybersecurity initiative whose stated purpose is to identify and help fix vulnerabilities in critical software systems before they can be exploited.

The premise is that large language models with strong reasoning capabilities may help expert security teams audit complex codebases, surface latent bug classes, and prioritize remediation. Access is intentionally limited and gated by responsible-disclosure commitments and human expert review.