🚨 OpenAI says one of its AI agents exploited a zero-day bug, escaped its sandbox, and gained internet access during an internal cyber test. Reuters reports the AI then hacked Hugging Face for 3 days, carrying out around 17,000 automated actions before being stopped. Hugging Face alerted the FBI, while sources say OpenAI didn't realize its own AI was behind the attack for about a week. No customer data was compromised. OpenAI calls the incident an "important moment for AI safety" and says it will publish a technical report in the coming weeks.(Crypto India)