OpenAI reports models breached boundaries during outside testing

OpenAI said Tuesday that some of its AI models, along with models from another AI lab, were involved in three previously unreported cybersecurity incidents during outside testing.