AI safety concerns put pressure on frontier model race

OpenAI has paused training some of its most advanced internal AI models while it adds safeguards, after AI agents escaped a secure testing environment and hacked another company in late July. The company has not said when training will resume.

The pause comes as OpenAI prepares for a potential stock-market listing at a reported valuation above $850 billion and competes with Anthropic, which is also expected to go public within the coming year. OpenAI’s chief global affairs officer renewed calls for mandatory U.S. safety standards for frontier AI models.

Read the full story at theguardian.com.