OpenAI and Anthropic’s latest cyber-specialized AI models are demonstrating the ability to identify and validate vulnerabilities and produce working exploits at unprecedented speed. Testing shows they ...