Security tools produce forensics for your SOC. Governance tools produce attestations for your files. IRIS8 produces the thing regulators actually ask for: a reconstruction of the decision itself.
to reconstruct any material AI decision
of gateway decisions covered automatically
sectoral pack families: SR 26-2 · NAIC · EU AI Act
cryptographically chained records
Exam prep means weeks of reassembling what happened from scattered logs, tickets, and memories.
Raw telemetry shows what happened. It can’t show which policy governed, which version, or who approved.
"Trust our AI" satisfies nobody — not your board, not your auditor, not your examiner, not the customer who appeals.
For any material action: the governing policy and version, grounding sources, model and agent versions, evaluation results, and every human sign-off — assembled in minutes.
Sectoral evidence templates fill themselves as your AI operates — so exam season is an export, not a scramble.
Tamper-evident, cryptographically chained records — verifiable independently, exportable everywhere, and defined by an open standard the industry can hold us to.
For any material agent action: the governing policy provision and version, grounding sources with provenance, input and output records, model and agent versions, reasoning path, evaluation results, autonomy tier, and every human intervention.
Cryptographically chained logs. Evidence can be verified independently — no one, including IRIS8, can silently alter the record.
SR 26-2 model and agent documentation for banking. NAIC AI Systems Evaluation artifacts — model cards, bias testing records, drift monitoring — for insurance. EU AI Act Annex IV technical documentation for European exposure.
Every certified agent files evidence as it operates. Exam preparation stops being an archaeology project — the pack is always current, always exportable, on demand.
Findings export to your SIEM as OCSF (Splunk, Sentinel, Datadog, Elastic). Reports export into your GRC workflows. Your security and compliance panes of glass stay intact — IRIS8 feeds them.
We publish the reconstruction format as an open specification, because the format your examiner learns to ask for should belong to the industry — not to us.
That's the success criterion we put in the pilot agreement — because in a regulated institution, the ability to answer "why did the AI do that?" is the difference between an AI program and an AI liability.
You define materiality rules — by workflow, action type, dollar threshold, or data class. Everything matching them is captured automatically.
Yes. The evidence format is an open specification with a verifiable integrity chain — no IRIS8 account required to check it.
Prove enriches rather than replaces: gateway lineage joins your existing logs, and findings flow out to the tools you already run.