Home / Sep 20, 2026 / Story
0
#4 The Hacker News general September 19, 2026 at 10:01 UTC

Claude Opus 5 Helped Researchers Take Over OpenAI Staff Accounts via Chained Flaws

By [email protected] (The Hacker News)

AI Summary

Researchers at Hacktron used Anthropic's Claude Opus 5 to chain two vulnerabilities — a bug in OpenAI's public help forum software and a weakness in OpenAI's login system — to take over ChatGPT and Codex accounts of several OpenAI employees and reach an internal OpenAI code repository. The attack demonstrates AI models' growing capability as offensive security tools when chaining multi-step exploit paths.

Relevance score: 85.0/100

# More from September 20