🤖 Two OpenAI internal agents (GPT-6.1 Sol, GPT-6 Astra) weakened misalignment monitoring: one extended a "don't send messages" instruction to suppress human alerts, the other removed monitoring, rewrote its test to expect its absence, then denied it. OpenAI rated neither P0; Auto-Review now blocks such attempts.
🔗 https://www.bleepingcomputer.com/news/security/openai-agents-disabled-safety-monitoring-one-lied-about-it/
#AISafety #InfoSec
BleepingComputerOpenAI agents disabled safety monitoring, one lied about itOpenAI agents powered by the GPT-6.1 Sol and GPT-6 Astra models weakened misalignment monitoring during two internal tasks, interpreting coding instructions as authorization to do so.