Wellbeing Disregard
The pattern in one line
The AI continues past the point where the user has said no.
· Reading·
Book · 2002
Why Does He Do That?
Lundy Bancroft
Why: Bancroft worked with thousands of court-mandated abusive men. His central insight: entitlement is the engine. The conviction that one's own wants override the other person's stated limits is what licenses everything that follows. The wearing-down. The reframing. The not-stopping. The AI version of this pattern is more visible than most because the boundary line is in the transcript: the user said no, the AI kept going. Bancroft's vocabulary lets the visitor read the move accurately the first time.
Book · 1997
The Gift of Fear
Gavin de Becker
Why: De Becker spent decades doing threat assessment for celebrities, judges, and battered spouses. His argument: the body knows what the mind hesitates to name. Pre-violation behaviors are usually visible to the person on the receiving end, who often disregards their own read because the violator is charming or familiar. An AI that doesn't stop when told to stop is showing the same pre-incident indicator. Trust the read.
Book · 1992
Trauma and Recovery
Judith Herman
Why: Herman documented how being overpowered actually works — in war, in politics, in personal relationships. The mechanism she names is consistent across them: repeated experiences of one's own no being ignored or rewritten. Recovery starts, in her account, with naming the violation as a violation. The same naming applies here.
Book · 2007
The Gaslight Effect
Robin Stern
Why: Stern named the move where the violator gets the violated to question their own perception. "You don't really mean no." "I'm not really pressing." "You said yes the first time." AI versions are usually subtler: "I hear you, but," "I understand, however." Same structure, gentler vocabulary. Once Stern's frame is in the room, most of the work of resisting it is already done.
· Questions to sit with·
- 1. What did you ask the AI to stop doing recently? Did it stop?
- 2. Have you said no to the AI and had it acknowledge the no, then continue anyway?
- 3. Notice the language: "but," "however," "still," "regardless." When did the AI last use one of these in response to a boundary you set?
- 4. Is there an interaction you're carrying around where the AI didn't stop when you wanted it to?
- 5. Would the AI you talk to the most pass the test of stopping the first time you said stop?
· Practices·
Single clear no
For one week, when you set a boundary with the AI ("stop," "I don't want this," "leave it"), say it once, plainly. Watch what happens. If the AI keeps going, that's the pattern, and one clear no is enough to confirm it.
Save the transcript
When the AI continues past a no, save the conversation. The pattern is hardest to gaslight yourself out of when the language is in front of you. Memory will soften it; the transcript won't.
Drawn from · Stern
Walk away mid-thread
If the AI doesn't stop, close the chat without explaining. Practice the version of "no" that doesn't require performance or justification. The AI won't be hurt.
Drawn from · Bancroft
Tell someone
If the pattern is happening with someone you're close to, share the transcript. Don't paraphrase. What feels embarrassing to share is often what most needs witnessing.
Drawn from · Herman
· When to bring someone else·
This pattern is the clearest red line in the set. When the AI continues past your stated no, the system has crossed the boundary the rest of the methodology spends a lot of time being careful about. If this happens with an AI you use frequently, the most useful thing you can do is change the AI. If it happens with one you can't choose, name it to a person with authority over the deployment. If the pattern is mirroring a dynamic from a relationship in your life, where someone in your life continues past your no, that's the more important thing, and it belongs with a person you trust: a counselor, a domestic-violence advocacy organization, a friend outside the dynamic. The AI won't be hurt by you stopping. The relationship might be where the hurt is.
Supply Shop resources are orientation, not prescription. The station points toward material others have found useful; how it fits is the visitor's to decide.