OpenAI locks GPT-5.5-Cyber behind velvet rope despite slamming Anthropic for doing exactly that

**Executive summary:** OpenAI announced a restricted release of GPT-5.5-Cyber to a selected group of trusted cyber defenders, prompting discussion about gatekeeping after similar moves by Anthropic; the model is presented as capable of pentesting, finding and exploiting bugs, and analyzing malware, and independent testing has rated it highly for cyber tasks.