OpenAI and Anthropic backed binding AI safety regulations during a UK parliamentary hearing, while Google DeepMind and Meta declined to endorse new legislation, exposing divisions over how advanced AI should be governed. * Regulatory divide: Anthropic said voluntary safety pledges are insufficient. OpenAI supported mandatory incident reporting and international rules, while Google favored standards and Meta pointed to existing regulators. * Oversight gaps: UK officials acknowledged that the AI Security Institute relies on voluntary access to models and cannot prevent companies from releasing models it considers unsafe. * U.S. restrictions: Anthropic said U.S. restrictions initially prevented full UK access to its Mythos 5.1 model, although access has since been granted. * Security concerns: MI5 Director Ken McCallum warned that AI could increase bioweapons risks, while former AI researchers questioned the industry's ability to regulate itself. 13-OCT-2026 05:16 PM ET