ISO Enacts Standardized Generative AI Exclusions in Commercial Liability: Eliminating Silent AI and Forcing Affirmative Underwriting

The Insurance Services Office (ISO, a subsidiary of analytics leader Verisk), which authors standard contract policy language across the United States and international admitted markets, has formally implemented a suite of standardized endorsements explicitly excluding Generative Artificial Intelligence exposures. The foundational forms—CG 40 47 01 26 and CG 40 48 01 26—fundamentally transform commercial coverage architecture by amending baseline Commercial General Liability (CGL) policies.

This analysis is provided exclusively for educational and informational purposes, examining the endorsements through the perspective of actuarial risk aggregation, specialized liability line pricing, and unmodeled exposure eradication ("Silent AI").

Mechanics of Endorsement Forms CG 40 47 and CG 40 48

The implementation of ISO's specialized filings aims directly at extinguishing "Silent AI"—the latent legal ambiguity whereby commercial casualty policies omit explicit references to artificial intelligence, allowing policyholders to argue during litigation that unforeseen algorithmic liabilities must be absorbed by casualty underwriters.

The newly deployed exclusions operate across precise coverage layers:

1. Form CG 40 47 01 26 (Total Generative AI Exclusion): Eliminates indemnity and defense obligations across both Coverage A (Bodily Injury and Property Damage) and Coverage B (Personal and Advertising Injury). The exclusionary trigger applies to any claim or suit arising directly or indirectly out of the utilization, output generation, design flaw, or recommendation of generative AI models;
2. Form CG 40 48 01 26 (Generative AI Exclusion - Personal & Advertising Injury): Specifically targets Coverage B, barring coverage for intellectual property infringements, copyright violations, commercial misappropriation, privacy torts, and defamation triggered by machine learning generation;
3. Statutory Definition of Generative AI: The filings introduce an exhaustive standardized definition, describing AI as "a machine-based learning system or model trained on data with the capability to create content or outputs, including but not limited to text, images, audio, video, or software code."

Eradication of Silent AI and Regulatory Scrutiny

The aggressive adoption of ISO's forms by global reinsurers and commercial carriers (including Chubb, Travelers, Munich Re, and Berkshire Hathaway) aligns with intensifying insurance supervision. The National Association of Insurance Commissioners (NAIC) Model Bulletin on AI Systems mandates that carriers demonstrate rigorous capital modeling and eliminate unhedged accumulations of unpriced exposures.

Faced with a complete absence of multi-decade loss experience, commercial casualty underwriters refuse to absorb catastrophic algorithmic risks within standard general casualty loadings. Attaching form CG 40 47 forces commercial insureds to either secure specialized technology liability policies (Technology E&O / AI Liability) or execute comprehensive algorithmic risk audits certifying continuous "human-in-the-loop" operational governance.

Actuarial Implications for Kazakhstan and Regional Emerging Markets

For the insurance sector of Kazakhstan and Central Asia, ISO's precedent provides a critical structural blueprint. As regional financial institutions, fintech ecosystems, and corporate conglomerates rapidly embed large language models (LLMs) and automated underwriting scoring engines into production workflows, permitting Silent AI in Directors & Officers (D&O) and corporate liability books introduces unquantified systemic fragility.

Key actuarial takeaways include:

• Casualty Portfolio Segmentation: Formal bifurcation between classical operational perils and algorithmic indemnity exposures;
• Stochastic AI Liability Reserving: Transitioning from static claims frequency distributions to stochastic modeling of multi-agent cascade failures and systemic digital rights infringements;
• Supply Chain Vendor Auditing: Integrating upstream third-party model developer risk assessments into primary underwriting guidelines.

Actuarial takeaway: The statutory rollout of ISO endorsements CG 40 47 and CG 40 48 marks the formal closure of unpriced 'Silent AI' within commercial general liability portfolios, preventing unmodeled systemic tail-risk aggregation. For actuaries and casualty underwriters, this establishes a structural mandate to construct standalone affirmative AI Liability products, calibrating rate schedules to verifiable 'human-in-the-loop' governance standards and upstream model vendor validation.

Source: Insurance Services Office / Verisk (verisk.com). Also reported by: Business Insurance, Morningstar, Clyde & Co, American Property Casualty Insurance Association.