OpenAI launches security products, tells orgs they’re doomed if they don’t use them

OpenAI has launched two security products that use AI agents to find and help fix software vulnerabilities. The announcement matters because OpenAI President Greg Brockman argued that organisations’ security will increasingly depend on the AI models they use, while acknowledging that basic security practices remain important.

Daybreak Red Team, now in early access, tests web applications for weaknesses and checks reported vulnerabilities and proposed patches. Daybreak Defense Factory scans codebases and prepares patches for engineers to review; OpenAI says it uses its GPT-6 Cyber model. Separately, four nonprofit groups sued OpenAI and four other AI companies in San Francisco, seeking to halt development or deployment of systems more capable than their current leading models, citing concerns including AI-enabled cyberattacks.