“Anthropic restricted the release of its latest Mythos model, claiming it poses cybersecurity risks due to its ability to find software exploits. The move raises questions about whether safety concerns mask competitive or business pressures at the frontier AI lab.”