AI Security & Offensive Research
AI Security & Offensive Research
Articles on offensive security research, red teaming, and vulnerability discovery using AI systems like Anthropic's Mythos Preview.
ai security offensive research4 hours ago5 min
Agentic AI Security: Risks & Governance After a Claude-Assisted Breach of OpenAI's Perimeter
A three-person team used Claude to chain a libheif memory bug and a Discourse auth flaw into OpenAI employee accounts. The incident is a case study in how agentic AI security risks outpace enterprise governance, and where the gaps actually are.