Flawed tests by Israeli cyber firm triggered real-world AI breaches: Report

A report alleges that flawed safety tests by an Israeli cyber firm led to real-world AI breaches, and that major AI labs (Anthropic, OpenAI, Meta) used these contractor missteps to bolster a narrative of rogue AI agents amid a pacing debate. The claim is emerging and uncorroborated; it matters because it could undermine trust in AI safety testing and influence regulatory and competitive dynamics.