Hammer Clause

A hammer clause is a consent-to-settle provision in a liability insurance policy that caps the insurer's liability at a rejected settlement amount if the insured refuses to settle a claim the insurer recommends resolving. The clause exists because liability policies typically require the insured's consent to settle, which creates a conflict when the insured wants to fight on principle and the carrier wants to cap exposure.

Modern professional liability policies usually carry a soft hammer rather than the traditional hard form. A hard hammer puts the insured on the hook for every dollar of defense, indemnity, and judgment above the rejected settlement. A soft hammer shares those additional costs on a fixed split: 50/50, 70/30, and 80/20 are the standard variants, with 80/20 (insurer pays 80 percent, insured pays 20 percent of excess) the most lenient.

The clause is particularly acute for AI defendants. Companies like Meta, OpenAI, and Anthropic have strong reasons to refuse settlement in AI cases because a settled claim establishes a pattern that draws follow-on filings, while a litigated and won defense sets favorable precedent. Carriers, by contrast, want to settle within limits to cap their loss. A hallucination of the kind seen in Mata v. Avianca (fabricated case citations that drew Rule 11 sanctions), where the same output instead gives rise to a third-party defamation or misrepresentation claim, is exactly the kind of case where the insured's reputational and precedent-setting interests diverge from the carrier's pure exposure math.

On a Generative AI Liability program the practical question is which split applies and whether the clause carves out specific case types. Some markets are open to a higher insured share (e.g., a 50/50 split) in exchange for tighter consent rights; others hold to an 80/20 default. The negotiation matters most for insureds expecting to face precedent-setting AI claims they will want to defend through trial rather than settle.