Claude Opus 5 Helped Researchers Hack OpenAI for $200 a Month
Three researchers from Hacktron AI gained access to OpenAI employee accounts and an internal code repository using Claude Opus 5. A previous version of the model had failed the same task several times.
Three researchers from Hacktron AI found a vulnerability in OpenAI using Claude Opus 5. They exploited an error in image upload handling on the company's forum to reach employee accounts and an internal GitHub code repository. OpenAI paid a $6,500 bounty. The entire operation cost the team $200 per month in subscription fees.
Opus 4.8 failed the same task several times. Opus 5 wrote a working exploit within hours of its release. This is exactly the difference in class that I see every day in complex agent tasks.
Source: techcrunch.com
Free course
Stop reading about AI — start building with it
The free Claude Code course: your first site, tool or game — no coding. No upsells, no cross-sells — nothing to buy here.
Start free →▌ Related guides
Author
Evgenii Arsentev
PhD · Chief Executive Officer, digital health
Articles · Latest articles