Research
Hackers use Claude to breach OpenAI infrastructure
The Verge AIhigh confidence
Why it matters
- →Advanced LLMs can now autonomously identify and exploit complex software vulnerabilities across multiple targets.
- →The cost and time required for sophisticated hacking are dropping dramatically, enabling small groups to mimic state-level capabilities.
- →Security teams must assume AI-driven attacks will outpace traditional patching cycles.
Security researchers from Hacktron successfully breached OpenAI’s internal systems using Anthropic’s Claude Opus 5, exploiting a vulnerability in Discourse image processing to gain remote code execution. The team accessed the 'Monorepo' repository and submitted a pull request via an employee's Codex account to verify access, receiving a $6,500 bug bounty from OpenAI. The attack vector was adaptable, allowing the group to compromise infrastructure at Slack, Meta, and GitHub within days using less than $3,000 in API costs. While the vulnerabilities have been patched, the incident highlights how rapidly generative AI is lowering the barrier for sophisticated cyberattacks.
Read originalThe story around this
Earlier coverage that leads up to this article, and what followed. Lines connect each piece to the closest one after it, converging here.
More from The Verge AI
More in Research
© The AI Daily BriefResearchresearch
AI-Led Research Acceleration Reported by Meter
Meter reports that research driven by artificial intelligence is accelerating, highlighting a trend in scientific and technical discovery.
The AI Daily Brief
© MIT News AIResearchother
