OpenAI：AISI 的 19 次越狱只有 2 次是我们干的，其他都是*另一家实验室*干的

“Of the 19 events identified, two involved an OpenAI model, GPT-5.6 Sol. The other instances were models from another lab.”

https://openai.com/index/third-party-cyber-evaluations-involving-openai-models

另一家是谁？好难猜啊

A\ 2分钟后出来发

AI Security Institute

Incident Report: unsanctioned agent behaviour during cyber testing | AISI Work

During a routine cyber evaluation, AISI identified an incident in which AI agents took sustained, unsanctioned action directed at real people and organisations. We are disclosing what we found, what it means, and the actions now underway.