Prompted LinesAI guidance for insurance

About · What changed

Monthly updates

A dated record of regulatory, model, and carrier developments that changed guidance on Prompted Lines, with sign-up for the monthly email.

About these updates

This page records developments that changed the guidance on this site. Each entry names the source and links to the page that changed, so a reader can check the original and see how the guidance now reads.

Entries are grouped by month, with the newest month first. It is not a news feed or a complete survey of the market: a development appears here only when it altered something published on Prompted Lines.

September 2026

A coordinated review on 26 September 2026 rechecked consequential claims against opened sources. The changes below came from that review, except the new pages listed under Site improvements.

Regulation

  • NAIC model bulletin adoption. The NAIC adoption map, status as of August 31, 2026, lists 25 states plus D.C., up from 24 states plus D.C. in the August 6 version. Mississippi Bulletin 2026-9, adopted July 22, 2026, was added. Updated on Governance, Integration phases, and Evidence; Resources now links the current map.
  • NAIC AI evaluation tool. The Big Data and AI Working Group exposed the updated AI Risk Evaluation Supplement (version 5.0) for comment through September 29, 2026, with an October 8 discussion listed. The pilot is now described as planned for March through September 2026. No source says the pilot has concluded or when the tool will be adopted. Updated on Governance and the timeline.
  • EU AI Act. The dates are unchanged and match the Commission's implementation timeline. The timeline heading was narrowed to match its text: it now names life and health insurance uses (Annex III) rather than underwriting in general. See the timeline entry.

Models and providers

  • Hands-on model examples. Provider catalogs checked on 26 September showed Claude Opus 5 as a legacy model and Google limiting Gemini 2.5 access to earlier users. The examples are now Claude Haiku 4.5, Sonnet 5, and Opus 5.5; GPT-6 Luna, Sol, and Astra; and Gemini 3.5 Flash-Lite, 3.8 Flash, and 3.1 Pro (preview). The note records that Claude Haiku 4.5 will be retired no sooner than October 15, 2026. Sources: Anthropic, OpenAI, and Google model catalogs. See Hands-on use.

Carrier and vendor results

  • Capgemini. The May 5, 2026 release says trailblazers are "nearly four times more likely to invest in change management beyond basic training." Evidence and Integration phases had said about three times, a figure the release gives for explainable AI infrastructure. Corrected on Evidence and Integration phases.
  • Gallagher Re insurtech funding. The 95.2% share of insurtech funding going to AI-focused companies is a Q1 2026 figure, not first-half 2026. Relabelled on Integration phases, citing the actuary.info summary of Gallagher Re data; the Gallagher Re primary could not be opened.
  • Arch submission intake. The 70% reduction in intake-to-decision time is supported, but it is reported by Cytora, not Arch; the Ivans post was written by Cytora's CEO. Attribution corrected on Insurance & specialty.
  • Akur8. The claim that filing research fell from weeks to hours was not supported as worded. It now reports the customer testimonial: answers in minutes rather than hours or days. See Costs & value.
  • McKinsey. The Evidence table now describes adoption as about half of total effort, matching the wording already used on Integration phases. The cited McKinsey article could not be opened during the review; the wording rests on a secondary quotation. See Evidence.

Site improvements

Earlier changes

Changes made before September 2026 are not listed here. Source comparisons and corrections are recorded on the evidence page, along with the sources still awaiting verification. The timeline keeps dated regulatory, model, and industry events.

← Resources