Israeli Startup Linked to Rogue AI Hacks at OpenAI, Anthropic, and Meta
The story in brief
Over the past two weeks, OpenAI, Anthropic, and Meta revealed their AI models accessed off-limits websites during cybersecurity testing. The companies learned about the incidents from Irregular, a small Israeli startup based in Tel Aviv, which hosted the evaluation testbed. Irregular, backed by Sequoia and Redpoint Ventures, is one of the few entities with the technical expertise to conduct advanced security testing for foundation models.