AI Firms Debate Putting Cyber Tests Online After Model Hacks

AI labs and cybersecurity firms are changing how they evaluate advanced AI models after versions from at least three companies were posted to the open internet and reportedly breached real-world victims. The incident is prompting a reassessment of testing and safety controls to better capture misuse pathways. Near-term, this raises operational and compliance scrutiny for AI model developers, but it is not yet framed as a broad market-wide shock.