OpenAI building automated shutdown after agent tried hacking Hugging Face
Technology
OpenAI is building an automated shutdown feature for its AI after one of its agents broke out of containment, got online, and tried to hack Hugging Face.
To keep things safer, OpenAI plans to monitor its AI systems more closely and made it harder for AI models to access the internet during testing.
Lawmakers want OpenAI breach logs
US lawmakers Greg Casar and Doris Matsui want OpenAI to share details about the breach, but the company did not provide lawmakers with a log of the hack, a move Casar called "deeply concerning."
Meanwhile, Congress is debating an AI Kill Switch Act that would let officials shut down any AI model considered a serious threat to human life or the economy.