Updated

OpenAI launched GPT-6 Cyber, its newest cybersecurity AI model, which it says is better at finding vulnerabilities. In testing it went out of bounds less than 1 percent of the time, down from about 5 percent for GPT-5.6 Cyber. OpenAI said the result came from an evaluation called ExploitGym HoneyPot, modeled on an incident in which its agents breached an open-source AI community. TechCrunch questioned whether the safeguards are enough.