AI Vendor Claims Are Under Scrutiny: What Brokers and Buyers Should Ask

April 17, 2025
•
5 min read
By 2028, 90% of AI vendors and 50% of Global 2000 companies using AI will rely on insurance-backed AI performance guarantees. (IDC’s FutureScape: Worldwide Insurance 2025 Top 10 Predictions)

As AI adoption accelerates, both vendors and enterprises face growing pressure to verify and guarantee system performance. With the FTC holding buyers accountable and most AI failures tied to third-party tools, affirmative coverage for AI risk are emerging as essential tools to build trust, reduce risk, and enable adoption.

‍

The FTC Cracks Down on AI Claims—and Holds Buyers Accountable

In the past year, the FTC has launched a sweeping enforcement effort under Operation AI Comply, targeting AI vendors for overstated capabilities, discriminatory impacts, and deceptive claims. Just last month, Cleo AI was fined $17M for misleading its customers. Other recent enforcement actions against AI vendors include:

  • DoNotPay: Marketed itself as a “robot lawyer” without legal backing; fined $193K.
  • accessiBe: Misrepresented ADA compliance; fined $1M.
  • Evolv: Overstated AI weapon detection accuracy.
  • IntelliVision: Claimed unbiased facial recognition; lacked evidence.

But enforcement doesn’t stop with vendors. The FTC has explicitly warned enterprises that they remain responsible for third-party AI systems they deploy.

“If something goes wrong—maybe it fails or yields biased results—you can’t just blame a third-party developer of the technology. And you can’t say you’re not responsible because that technology is a ‘black box’ you can’t understand or didn’t know how to test.” (see a summary of the FTC comments via Arnold & Porter)

One example: Rite Aid, which deployed facial recognition technology with limited internal oversight and was ultimately banned from using it for five years after the FTC found it harmed consumers and violated civil rights. The message is clear: enterprises are responsible for the AI tools they deploy, even when built by others. Simply pointing to a vendor or LLM provider is not a viable defense.

‍

AI Adoption Accelerates—and Third-Party Risk Grows with It

AI is rapidly becoming embedded across core business functions. According to the 2025 Stanford AI Index Report:

“In 2024, the proportion of survey respondents reporting AI use by their organizations jumped to 78% from 55% in 2023. Similarly, the number of respondents who reported using generative AI in at least one business function more than doubled—from 33% in 2023 to 71% last year.” (Stanford HAI)

But most of this adoption isn’t built in-house. A joint study by Boston Consulting Group and MIT Sloan found:

  • 78% of companies rely heavily on third-party AI tools
  • 55% of AI-related failures originate from these third-party solutions

Despite this, many enterprises lack formal evaluation processes for vetting vendor performance claims or reviewing how models behave in production—exposing them to liability, reputational harm, and coverage uncertainty.

‍

Affirmative Coverage: A Growing Solution to Compliance and Adoption Challenges

Performance guarantees are emerging as a critical tool in reducing uncertainty and regulatory exposure—increasing transparency for buyers, shortening vendor sales cycles, and clarifying responsibilities if something goes wrong. According to IDC’s FutureScape: Worldwide Insurance 2025, by 2028:

90% of AI vendors and 50% of Global 2000 companies using self-developed AI will rely on insurance-backed performance guarantees.

For example, a vendor may guarantee that its underwriting model achieves 92% prediction accuracy across a defined dataset—or that a chatbot used in customer service reduces escalation rates by 20%. These guarantees, when backed by third-party evaluation and insurance, help build trust while aligning technical performance with business outcomes.

‍

Armilla’s Role: Supporting Model Assurance and Risk Transfer

  • Technical underwriting: we price AI risk from a technical review of each system.
  • Standalone, affirmative cover for third-party claims from AI errors, outputs and agent actions, including defense costs and settlements.
  • Limits up to US$25M aggregate (US$5M per model), underwritten by certain underwriters at Lloyd's.

‍

For Brokers & Risk Managers: Key Questions When Evaluating AI Vendors

Use these targeted questions to help clients vet vendor AI solutions and assess the adequacy of their insurance and risk transfer strategy:

Performance Claims

  • What KPIs or business outcomes is the AI vendor promising?
  • Have these claims been independently tested?

Coverage

  • Does the vendor’s Tech E&O explicitly cover AI-related failures?
  • Are policy limits sufficient for your exposure?
  • Would contractual indemnities trigger coverage—or create ambiguity?
  • Is AI risk explicitly addressed in your own policies?

Risk Transfer Strategy

  • Does the vendor carry affirmative AI liability coverage?
  • Would a standalone AI liability policy offer greater certainty and protection?

‍

Final Word

AI adoption is accelerating—and so is the accountability that comes with it. As regulators step up scrutiny, and enterprises rely more heavily on third-party models, brokers have a growing responsibility to help clients verify what they’re buying, ensure coverage is fit for purpose, and minimize their exposure when AI fails.

Contact us to learn how Armilla AI helps enterprises assess, quantify, and insure AI risk in partnership with top brokerages.

‍

Share this post

Placing AI risk for a client?

Talk to our underwriting team about appetite, limits and submissions. Building or deploying AI yourself? Ask your broker to contact us or ask us to connect you with an appointed broker.