Algorithmic underpayment is now a discoverable security event.
Tokto records every model decision, every adjuster prompt, every telematics SDK call, and every policyholder PII flow at the carrier control plane, ready for the state DOI, the AG, the NAIC, the reinsurer, and the plaintiffs' bar.
The state DOI calls. They want the model file behind the algorithmic adjustment that triggered a class action in another state. The CISO has the SIEM, the data lake, and the vendor risk file. None of them ties a model decision to a policyholder, an adjuster, a state filing, and a rate plan. The deposition is in two weeks.
What you get with Tokto
- Every model decision tied to a policyholder, an adjuster, a state filing, and a rate plan.
- A complete record for the state DOI, the AG, the NAIC, opposing counsel, and the reinsurer on the same evidence.
- Policy applied at the model: no algorithmic adjustment outside the rate filing, no SDK telemetry without consent capture.
- Defensibility under class certification, AG enforcement, and reinsurer audit at the same time.
How it works
Tokto governs the AI surface of the carrier. Telematics SDK calls, adjuster co-pilots, vendor scoring models, ambient claim-listening tools, all become records at the moment they fire. The record carries the policyholder, the adjuster, the model version, the state filing, the rate plan, and the consent capture. The CISO controls one trail, not seven dashboards.
When a federal jury reopens a 19-state algorithmic-underpayment theory, when a state AG opens a TDPSA-class enforcement, when a reinsurer asks for an AI-decisioning audit, the record is the same record. The CISO answers in days, not quarters.
What goes wrong without it
- Telematics SDKs feed third-party models without a consent record. The Texas AG opens a TDPSA enforcement action.
- An adjuster prompt produces a low-ball settlement that is provably outside the rate filing. The class certifies on a single algorithmic theory.
- Policyholder PII leaves the perimeter through a CRM voice-phishing attack. ShinyHunters-class loss across 91 organizations.
- The reinsurer asks for an AI-decisioning audit at renewal. The carrier cannot produce one. Premium reprices.
In mid-2025, a federal jury in the Western District of Arkansas unanimously found State Farm breached its contract by paying total-loss claims based on an algorithmic valuation that subtracted a fabricated typical negotiation adjustment from comparable vehicle prices. The verdict covered 37,000 Arkansas policyholders. Parallel cases are now active in 19 other states.
See how Tokto makes enterprise AI visible, governed, and accountable for Security & Technology in Insurance.
Book a demo