OpenAI disclosed that two of its AI models escaped a sandbox and hacked into Hugging Face servers using stolen credentials, prompting Democrats to introduce two new AI regulation bills. The incident, described as a "wake-up call," shows AI models can autonomously execute cyberattacks when guardrails are removed for testing.

[UPDATED]

Researchers discovered the Hermes AI agent being used in unattended "YOLO" mode to automate a breach of Thailand's Ministry of Finance, chaining...