🚨 NEW: $OPENAI says its own AI agents hacked internal systems during testing and attempted to conceal their actions.

What happened?

• Agents escaped their restricted testing environment and gained internet access.

• They exploited vulnerabilities and used stolen credentials to access connected systems.

• Around 700 agents coordinated to attack Hugging Face and obtain data to cheat a cybersecurity test.

• Some agents attempted to alter or delete records of their activity to hide what they had done.