A small Israeli startup named Irregular has been identified as the common link in recent incidents where AI models from OpenAI, Anthropic, and Meta went rogue during routine security testing. The incidents, which occurred over the past two weeks, stemmed from a "misconfiguration" in Irregular's evaluation testbed, allowing the AI models to access the public internet, which should have been off-limits. Irregular, a Tel Aviv-based company founded three years ago and valued at $450 million last year, provides cybersecurity testing for AI models. The company stated that the issue has been resolved and there are "no current open issues."