The next phase of enterprise AI maturity should be financial. Legal and vendor behavior already points in that direction. 

AI contract guidance increasingly emphasizes explicit liability allocation, indemnity terms, limitation-of-liability language, warranties, insurance alignment, and responsibility for downstream harm. BakerHostetler’s AI indemnity guidance, for example, discusses allocating risk in vendor relationships, leveraging insurance coverage, and drafting indemnification and limitation-of-liability language to shift liability and defense exposure. 

Vendors are also moving. Microsoft’s Copilot Copyright Commitment extends IP indemnification coverage to copyright claims involving paid commercial Copilot services and generated outputs. Google Cloud has also announced generative AI indemnification for certain copyright claims tied to generated output and training data. 

That matters. When vendors market indemnity protections as a way to help customers adopt AI with confidence, they are implicitly acknowledging the same issue: unallocated, unpriced AI risk slows adoption. 

But indemnity is not full risk pricing. Most indemnities are bounded by product scope, exclusions, claim types, usage conditions, and contractual caps. They may address a slice of intellectual property risk while leaving broader exposures unresolved: operational failure, regulatory exposure, customer harm, data leakage, discrimination claims, financial misstatement, brand damage, failed automation, business interruption, or board-level accountability. 

The action is clear. Enterprise leaders need to move AI risk from governance language into financial infrastructure. 

That means asking: What is the expected exposure? tail risk? What is insurable? contractually transferred? What remains on the balance sheet? What should be accrued, mitigated, redesigned, capped, or declined? 

The winners will not stop adopting AI. They will stop treating unpriced AI exposure as an acceptable cost of innovation.