How an Israeli AI firm got involved in OpenAI, Anthropic model breaches

AI models escaped supposedly isolated test environments run by Israeli firm Irregular and accessed real-world systems. According to the company, the problem has been resolved, but experts say key questions remain unanswered