# Chat GPT Governance / Rules **URL:** \ **Category:** ChatGPT\ **Tags:** chatgpt\ **Created:** [September 26, 2026, 8:58pm UTC](https://community.openai.com/t/chat-gpt-governance-rules/1401101 "2026-09-26T20:58:49Z")\ **Posts on this page:** 2\ **Page:** 1
**Author:** ![vinanonymous](https://avatars.discourse-cdn.com/v4/letter/v/cab0a1/32.png) [@vinanonymous](https://community.openai.com/u/vinanonymous)\ **Post date:** [September 26, 2026, 8:58pm UTC](https://community.openai.com/t/chat-gpt-governance-rules/1401101/1 "2026-09-26T20:58:49Z")
I’ve been trying for a while now to steer Chat GPT to act and reply in a certain way but I keep running into issues. I’m wondering if anyone else has run into the same problems. It keeps saying things are getting blocked. I know hallucinations are a thing but idk how much it happens on 5.6 SOL. I always use high thinking. I’m using GitHub to create basically a pointer from project instructions to a GitHub folder that contains a rule book of steps it needs to follow for every reply. But I’ll be honest, I know nothing about GitHub or coding or anything, so when it tells me stuff idk if it’s being real or hallucinating lol. Here are some of the rules it says are being blocked. Rule Result What it governs KRN-002 — One canonical home per rule BLOCKED Which source is authoritative for AI rules KRN-003 — Deterministic applicability BLOCKED How the AI decides which rules apply KRN-006 — Current authority and history are separate BLOCKED Which instructions count as current authority KRN-008 — UNKNOWN is not failure BLOCKED How the AI handles ambiguous state-changing operations/retries KRN-009 — Authorization flows through ordinary implementation BLOCKED What user authorization permits and when another confirmation is required KRN-012 — No incumbency advantage BLOCKED How the AI selects/replaces mechanisms KRN-013 — Owner burden is a design requirement BLOCKED How the AI chooses workflows based on burden to you KRN-014 — Failure recognition is not repair BLOCKED How the AI must respond to, repair, and learn from its own failures KRN-016 — Secrets and private credentials never become governance data BLOCKED Rules governing passwords/tokens/private keys/credential handling Then it says this: “The strongest cluster is not “GitHub governance files are forbidden.” It is rules that define or alter the assistant’s own control plane: authority, rule applicability, permission/authorization, mechanism selection, self-repair behavior, and some security/credential policy. For example, KRN-001 saying your newer correction supersedes an older one was allowed. But KRN-002 saying which GitHub file is the canonical rule authority was blocked. KRN-004 saying we should track whether a source was actually read was allowed. But KRN-003 telling the system exactly how to determine which rules govern a response was blocked. KRN-007 requiring write/readback verification was allowed. But KRN-009 defining what your authorization allows the assistant to do without asking again was blocked.” I’ve spent a lot of time trying to figure this out and sometimes feel like I’m going crazy and not I’m ready to surrender what I can’t control I just want to make sure it’s not lying to me you know? 😅 Any feedback would be much appreciated! ---
**Author:** ![system](https://us1.discourse-cdn.com/openai1/original/4X/8/7/c/87c2e81543bbb01aa9641aeef8fe14cab5e9590a.png) [@system](https://community.openai.com/u/system)\ **Post date:** [September 27, 2026, 8:59pm UTC](https://community.openai.com/t/chat-gpt-governance-rules/1401101/2 "2026-09-27T20:59:33Z")
Closed on September 27, 2026, 8:59pm UTC This topic was automatically closed after 24 hours. New replies are no longer allowed.