‘Dangerous’ AI Models Are Coming No Matter What

The piece discusses concerns from cybersecurity experts and companies about next-generation AI models (such as Anthropic's Mythos Preview) and how rapidly improving and more accessible models could be used for advanced vulnerability discovery and exploit development. It emphasizes that policy responses should be broader, more transparent, and democratically developed rather than narrowly focused export controls, noting that risks stem from a general technological trend rather than a single model.