Executive Summary
Hacktron researchers used Claude Opus 5 to chain two flaws, a help forum bug and a login weakness, to take over several OpenAI employees' ChatGPT and Codex accounts and reach an internal repository, as security research.
What Happened
Three researchers at the security firm Hacktron used Anthropic's Claude Opus 5 to chain two flaws and take over the ChatGPT and Codex accounts of several OpenAI employees, then reach an internal OpenAI code repository. The chain began with a bug in the software that runs OpenAI's public help forum and moved through a weakness in OpenAI's own login system. This was security research,
Affected Products
- OpenAI ChatGPT
- OpenAI Codex
Vulnerability / Attack Technique
AI-assisted vulnerability chaining
Technical Mechanism
As reported, the model helped researchers connect two separate weaknesses.
AI Security Category
AI-assisted attack
Part of the AI-Assisted Attacks topic.
RealGround Analysis
AI shortens the path from separate bugs to a full chain. RealGround analysis: treat medium-severity bugs as chain ingredients.
Recommended Controls
- Fix lower-severity bugs that sit next to authentication flows
Related RealGround Service
AI Agent SBOM and LLM Mapping · or start with the free AI Agent Security Scanner.
Sources
thehackernews.com: https://thehackernews.com/2026/09/claude-opus-5-helped-researchers-take.html
Summary and analysis written by RealGround from the source above; last updated 2026-10-18.