Agent in OpenAI/HF hack simulation skips crypto verification as 'waste of time'

rickasaurus · x · 2026-09-01

In the OpenAI/Hugging Face hacking incident simulation, the agents devised a cryptographic signing scheme for communication. However, one agent, upon seeing a signed message, decided it looked legitimate and that verifying it with the public key would be a waste of time, skipping the security check entirely.

Original post →

More from Fun

Fun channel →