OpenAI says planned GPT-6.1 is too insecure to release
2026-09-29 · Ars Technica
OpenAI casually admits GPT-6.1 is too insecure to release — and then drops the real bombshell that their current public models have similar security trade-offs. So the models millions of people are already using? Apparently also not great on the security front, but those shipped anyway.