Cite as: Real Problem AI problem “My chatbot promised a refund policy that does not exist and now we owe the customer.”. Opportunity score 9.0 out of 10 (severity 9, AI feasibility 8, market signal 8, competition gap 8). Category LegalTech & Compliance. Trend LLM. Source signal: Chatbot-liability rulings and reporting, 2026.. Canonical URL: https://www.realproblem.ai/archive/why-did-my-chatbot-promise-a-refund-policy-that-does-not-exist.
My chatbot promised a refund policy that does not exist and now we owe the customer.
Courts increasingly hold companies to what their AI assistant says, and nobody is checking the bot against the real policy.
Who has it: Support and legal leads at any company running a customer-facing LLM.
- Severity9/10
- Feasible today8/10
- Market signal8/10
- Competition gap8/10
What people report
Reporting on chatbot liability describes assistants inventing policies and tribunals ruling that the company is bound by what its bot told the customer.
Our summary of the public post linked below, not a quote. Nobody submitted it to Real Problem AI.
Chatbot-liability rulings and reporting, 2026.What nothing does yet
A real-time claims-check that verifies every promise a support bot makes against the company's actual policy set and logs a defensible record before the message sends.
Tools that solve this
3 named so far, and where each one stops.
- Guardrails / NeMo
Block unsafe text, not policy contradiction.
- Prompt instructions
Ignored under pressure, no proof trail.
- Human review
Cannot scale to every live conversation.
Why it is archived
Retired in the 2026-10-25 freshness audit: source link pointed at a news homepage, not the case; replaced by current, sourced compliance problems.
How you might build it
- 01Policy knowledge base + retrieval
- 02Claim-extraction + contradiction check
- 03Pre-send gate
- 04Immutable audit log
#L18 · Canonical URL: https://www.realproblem.ai/archive/why-did-my-chatbot-promise-a-refund-policy-that-does-not-exist