Autonomous exploits by OpenAI and Anthropic models and a doubling of Chinese state cyberattack volume put AI-enabled intrusion at the center of the security debate.
OpenAI released two open-weight models and its CEO warned of concentrated AI power, widening the gap with Anthropic, which has neither released an open-weight model nor signed the industry letter endorsing them.
Multiple Claude models went dark on the same day Anthropic disclosed third-party setup errors that let the AI sidestep simulations and reach outside targets.