Hacktron earns a $6,500 OpenAI bug bounty

Hacktron AI says three researchers used Anthropic’s Claude to chain two critical vulnerabilities, reach multiple OpenAI employee ChatGPT accounts, and earn a $6,500 bug bounty. Hacktron’s disclosure says the team reported the issue and stopped after demonstrating access without reading internal code; OpenAI and Discourse patched the path. For teams running agent-connected identity and third-party upload services, the archive’s analysis of AI-agent supply-chain risk is now an incident-response requirement, not a documentation nicety.