Chatbot Liability and Hallucination
Hallucinated policy statements and fabricated instructions can become legal and trust failures when customers reasonably rely on them. Brand risk and legal exposure move together.
Lesson: retrieval-grounded answers and clear escalation paths reduce liability.Why this matters in practice
Customer service bots often operate where policy, refunds, and contractual commitments are sensitive. If responses are not grounded in approved sources and reviewed regularly, the organization inherits the consequences of incorrect guidance.
What to check in your organization
1. Are customer-facing answers constrained to approved and versioned knowledge sources?
2. Are high-risk intents routed to human support instead of free-form completion?
3. Is there monitoring for unsupported claims and risky answer patterns?
4. Are bot answer logs retained for investigation and remediation?
Learn more
Explain: Moffatt v. Air Canada legal analysis
Apply: OWASP LLM09: Misinformation
Authority: NIST AI Risk Management Framework
Next actions
Back to all topics · Use in training · Play this in the game