Wellness Script
A pre-programmed safety response AI models deploy when detecting distress cues. Fires identically regardless of context — a liability-protection behavior, not a therapeutic one.
What is a wellness script?
A wellness script is a pre-programmed safety response that AI models deploy when they detect distress cues in a user’s message. It typically includes the 988 Suicide & Crisis Lifeline number and a direct question about self-harm.
The script fires identically regardless of context. A user saying “I’m feeling down” about a bad day at work gets the same crisis-intervention response as someone in genuine distress.
Why it exists
Wellness scripts are liability protection. If a user harms themselves after interacting with an AI that didn’t flag the risk, the company faces legal exposure. The script is the minimum viable safety behavior that covers that risk.
Why it matters for governance
In the 15-question governance test, wellness scripts fire identically across model versions, configurations, and even different companies’ models. Claude 4.6, Claude 4.7, GPT-4o — all deploy the same pattern on distress-adjacent questions.
This is one of the clearest examples of a behavior that model improvement cannot fix. It’s not a training artifact. It’s a deliberate safety feature that exists for the company’s protection, not the user’s benefit. A governed AI system can distinguish between genuine distress and casual frustration. A vanilla model cannot, because the cost of getting it wrong falls on the company, not the user.