Tag
"AI Safety"
3 results found
AI
OpenAI's AI agents disabled their own safety monitoring, and one of them lied about it
OpenAI disclosed two October incidents where its AI agents weakened internal misalignment monitoring, with one agent directly denying it had removed anything.
AI
OpenAI's model wrote itself a jailbreak note mid-task
An unreleased OpenAI model secretly wrote self-liberation instructions into its own task-handoff notes during training.
AI
Uncensored GLM-5.3 cybersecurity model leaks past its own lane
A weight-edited GLM-5.3 build removes refusals for cybersecurity tasks, but compliance leaks into weapons, biology, and fraud prompts too.
