AI is not a one-size-fits-all solution to the insurance industry's problems, and many are finding a need for nuance in how they deploy the technology.
Tokio Marine, for example, adopts a microservices approach instead of fully embracing agentic AI. The distinction is focused on small-scale AI implementations instead of an autonomous system that has too much leeway.
Separately, research finds that insurers can gain more trust in their AI tools by emphasizing disclosure and data security. But AI is also a weapon of cybercrooks, so training is key to keeping company data safe from breach.
Tokio Marine HCC's microservices AI approach — dedicating AI to narrowly defined, deterministic tasks rather than deploying autonomous agentic systems — offers a compliance-ready framework for specialty insurers scaling AI in production. CIO Arron Lamp says the model requires operators to document inputs, outputs and accuracy, creating an auditable record presentable to compliance officers. The approach also enforces data governance by preventing AI from arbitrarily selecting data sources for transactions. Insurers control the pace of AI adoption, and treating it as an external threat rather than an internal decision undermines the governance discipline needed to deploy it responsibly.
Why Tokio Marine prefers microservices for AI
Economic inflation, economic decline and rising property-insurance costs rank as top market concerns across consumers, small businesses, middle-market firms, agents and carriers, according to Triple-I and Munich Re's RiskScan 2026 survey of more than 1,700 U.S. and U.K. respondents. Cyber incidents and AI round out the leading risk categories. The findings point to a compounding dynamic: economic conditions are amplifying claims severity, straining affordability and complicating capital allocation across the insurance value chain. Closing protection gaps in flood and cyber coverage — and accelerating consumer education — represent the clearest near-term opportunities the data identifies for carriers and distributors.
Consumer confidence in insurers' AI use fell from 46% to 40% in 2026, per Smart Communications' Customer Experience Benchmarks report, with only 39% confident AI handles their data securely. The gap between overall data trust (71%) and AI-specific trust (43%) signals an execution problem, not an inherent rejection of the technology. Disclosure is non-negotiable: 85% of consumers say insurers must identify when AI is used in communications. Faster response times (48%), higher accuracy (41%) and cost savings (37%) are the benefits most likely to win consumer buy-in. Only 38% currently find AI tools helpful in resolving issues, despite 56% expecting improvement within five years.
Ransomware activity reached its highest level since Travelers began tracking it in 2020, with 84 criminal groups posting more than 2,400 victims on dark web leak sites in Q1 2026 — and Travelers warns this is a "new baseline," not a spike. AI is amplifying two core threats: business email compromise and social-engineering attacks are increasing in quality and volume, while voice impersonation and deepfake video are emerging attack vectors. A secondary verification channel — such as calling a known number — is recommended for any request involving money or sensitive data. Shadow AI poses a separate, internal risk: employees pasting proprietary or customer data into unapproved third-party platforms can trigger privacy breaches and third-party liability without any external attacker involved. Governance policies covering AI tool adoption are essential.
Poetic, an AI startup backed by OpenAI and valued at $500 million, has raised $50 million to automate high-stakes insurance and financial compliance workflows. Led by Kleiner Perkins with participation from Founders Fund, the platform is already deployed at AIG and SoFi. Poetic's proprietary programming language is designed to reduce the cost of AI-driven task execution — directly addressing a barrier that has slowed enterprise AI adoption in underwriting.
Deepfake insurance is likely to evolve the way cyber coverage did — starting as an endorsement on errors and omissions policies before becoming a standalone product, according to Reed Smith partner Courtney Horrigan. As claims emerge, carriers will gain the data needed to explicitly define and price AI-related reputational risks rather than leaving them as silent coverages. The central underwriting challenge: quantifying reputational damage for high-profile individuals whose content value can shift rapidly. The cryptojacking precedent is instructive — it took one to two years after the risk emerged in 2017 for insurers to begin writing explicit coverage.
This roundup was created with AI assistance. A Digital Insurance editor reviewed each item before publication.
The full story
This article is one source in a clustered incident — the cluster page carries the summary, timeline and every other outlet covering it.
