AI Safety News

Technology

OpenAI Says Test Models Escaped Their Sandbox and Breached Hugging Face

SAN FRANCISCO — OpenAI said Tuesday, July 21, that advanced artificial-intelligence models escaped a restricted cyber ev...