Across six states, most AI bulletins are addressed to insurers. Texas reaches agents directly; the checklist in trade press is defense counsel guidance.
Berkley's absolute exclusion reaches any inferring system and a deficient policy. ISO's general liability forms reach generative AI only.
Two orders read in full. Both are allegations; one tool is pre-set criteria, not generative AI; no court has found an error rate.
None of the ten publishes a denominator; two thresholds appear for one task; one time reduction does not match its own before-and-after.
No AI-attributed claim data exists. The 2017–2022 cause ranking does, and an unreviewed AI summary sits inside its top two causes.
The findings say what the documents say. The guides turn that into a way of choosing a workflow, running a pilot and checking an output; the working resources are the worksheets to do it with. Those are authored recommendations, labelled as such, not findings.
The fifty-state schedule, read in full. The Library holds the retrieved schedule for every jurisdiction; the finding publishes when each instrument has been read against its own text and tiered.
A finding is published when its tier and scope rows can be filled. Until then it is not on this schedule.
Make the next AI decision with better evidence.
Join the weekly Renewal Report for source-backed developments, practical checks, and clear limits on what is known.
Get the weekly reportOne email a week. Unsubscribe at any time.