Verified adult users complain ChatGPT GPT 5.6 silently tightened safeguards
Unusual-Pineapple350 · reddit · 2026-09-17
- A Reddit user publicly vents at Sam Altman: after completing age verification as instructed, the upgrade from GPT 5.5 to 5.6 tightened safety guardrails without any warning.
- The user says verified adults are now treated like teens, with the model acting like a "guarded terrified lawyer," and recalls that even GPT 5.3/5.4 felt freer despite weaker writing.
- They argue OpenAI should protect minors but not nanny verified adults, calling for a separate looser model for adults or a redesigned verification system.
More from Models
- Grok 4.7 leak: 2.1T params, fresh pretrain beats 4.6, near Opus 5.0 class — airesearch12 · 2026-09-18
- Z.ai's blog documents early signs of recursive self-improvement in GLM — inductionheads · 2026-09-18
- Chatting With Jev as a Regular LLM: A 3 tok/s Caveman-Style Experience — _rchaves_ · 2026-09-18
- Developer turns judgment model Jev into a chatbot: 3 tok/s and pricier than LLMs — _rchaves_ · 2026-09-18
- Margaret Mitchell: Self-Code Editing Should Be Off the Table for AI Models — mmitchell_ai · 2026-09-18
- HuatuoGPT-3: open 9B medical LLM trained with one-stage RL, no SFT — jacek2023 · 2026-09-18