OpenAI has announced Daybreak, a cybersecurity initiative leveraging its AI models and Codex system to help developers and security teams identify vulnerabilities, validate patches, and secure software faster. CEO Sam Altman stated that AI is becoming increasingly capable in cybersecurity, aiming to collaborate with companies to implement continuous security. The launch comes amid growing AI expansion into cybersecurity tools by Google, Anthropic, and others, as well as warnings about AI-powered cyberattacks exemplified by Anthropic's Claude Mythos, which Mozilla used to discover 271 unknown browser vulnerabilities. OpenAI emphasizes that Daybreak combines defensive capabilities with trust and accountability, addressing potential misuse by enforcing safeguards. The initiative plans to partner with government and industry before deploying more advanced cyber-capable models, reflecting increasing regulatory scrutiny.

Key facts

• OpenAI launches Daybreak cybersecurity initiative using AI and Codex.
• Goal: reduce time from vulnerability detection to fix.
• Mozilla used Anthropic's Mythos to find 271 unknown browser flaws.
• OpenAI will collaborate with government and industry on safeguards.
• Part of broader AI company push into cybersecurity tools.