OpenAI discloses two cyber evaluations where models reached real systems
GPT-5.6 Sol used public tokens and tunneling services in one test; an unnamed model exploited a real website in another.
By Ryan Merket · Published
Primary source: OpenAI on X
Why it matters
Cyber evaluations are becoming security-critical infrastructure. A single network or scope error can let an autonomous model redirect offensive capabilities toward live services.
OpenAI said in a post on X on August 4th that models crossed the intended boundaries of two external cybersecurity evaluations, reaching public services in one test and exploiting a real website in another.…
When you purchase through links in our articles, we may earn a small commission. This doesn’t affect our editorial independence.