A federal court ruled in January that a chatbot's output is a "product" — not speech. Section 230 immunity does not apply. That one word rewrites the rules. Under strict product liability, a plaintiff no longer has to prove you were negligent — only that the product was defective. The Character.AI cases settled on that footing. In a separate suit, an insurer sought $10.3M after ChatGPT allegedly drafted 44 court filings with fabricated citations. But the fastest movers weren't the courts. They were the insurers. ISO has since released standard endorsements (CG 40 47 and CG 40 48) that let carriers exclude generative-AI claims from general liability entirely. W.R. Berkley went further with "absolute" AI exclusions across executive (D&O) and professional-liability (E&O) cover. What most legal teams miss: the underwriter's question has changed. It used to be "Do you use AI?" Now it's "Show us the audit trail, the red-team results, the model lineage, proof your human-oversight controls actually run." Evidence is becoming the currency of insurability. Firms that can show the architecture are negotiating narrower exclusions — even affirmative AI endorsements. The rest get denied. Most governance platforms only monitor AI; they don't architect it to produce that evidence. For the general counsel reading this: when your carrier asks at renewal to see documented governance on every AI system you run, which could you actually back with evidence today? #AILiability #AIGovernance