Hunting Vulnerabilities Using Frontier Models
Okta used frontier AI models GPT-5.5 Cyber and Mythos via OpenAI and Anthropic programs to scan millions of code lines for vulnerabilities.
Okta describes using frontier AI models, including GPT-5.5 Cyber Preview (TAC) and Mythos Preview, through OpenAI's Daybreak Cyber Partner Program and Anthropic's Project Glasswing to hunt vulnerabilities across its product codebase. The team built a custom Python orchestrator with strong isolation, vendor-agnostic model support, and four distinct scanning pipelines executed as isolated Codex or Claude Code sessions with progressive context loading to reduce context bloat. Human experts and AI agents worked both autonomously and in paired hunts, and Okta reports the best results when humans and agents taught each other.
- Four distinct scanning pipelines showed different step sequences surface different vulnerability types.
- Progressive context loading limited context bloat across isolated agent sessions.
- Running multiple pipelines traded duplicate findings and higher false-positive rates for broader coverage.
- Full session capture let human reviewers audit any agent step after execution.
- AI models
- GPT-5.5 Cyber PreviewMythos Preview
From the original article
9 min read at sec.okta.com
Okta used early access to GPT 5. 5 Cyber Preview and Anthropic's Mythos Preview to hunt vulnerabilities in our own codebase. The big lesson: agents alone generate noisy leads, but pairing them with human experts who steer investigations produces real, high-confidence findings, plus a reusable pipeline architecture for future AI-driven security work.
sec.okta.com
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.