Enterprise buyers, GCs, insurance underwriters, and procurement teams want proof your AI can be trusted. In five business days, Mejepa returns a signed, underwriter-facing evidence pack — so the questionnaire stops being the chokepoint.
Every Fortune 1000 procurement team now sends one. Every cyber-insurance underwriter pricing your E&O premium asks the same questions. The answers are not in your SOC 2 report. They are not in your model card. They are not in your sales deck.
By Q2 2026, enterprise procurement teams routinely ask AI vendors to demonstrate four things that no SOC 2 audit covers: (1) how the AI behaves at its failure boundary; (2) when and how it abstains; (3) what evidence exists for a single production output; (4) whether the vendor can produce a signed record an underwriter can verify.
Insurance carriers are repricing AI vendor E&O and tech-errors policies around the same axes. Munich Re's aiSure product (Jan 2026) and Mosaic's $15M AI coverage launch (Feb 2026) both require conformal-style operating-envelope evidence and abstention-rate disclosure. Vouch and Embroker are following.
Vendors who cannot answer get one of three outcomes: the deal stalls in legal, the premium gets repriced upward, or the buyer requires a bespoke side-letter the vendor's GC has to negotiate from scratch. All three are deal-velocity killers for a Series A or B AI company.
The Mejepa Insurability Evidence Pack is the artifact your buyer's procurement team and your insurance underwriter both expected to receive — but nobody ships. Five business days. Signed. Verifiable offline.
Pricing scales with the depth of evidence the buyer or carrier requires. Most Series A AI vendors land at $3,500. Mid-market and regulated-vertical AI vendors typically need the $10,000 tier with carrier-narrative inclusion.
For Series A/B AI vendors needing to clear enterprise procurement. SIG-Lite + CAIQ + AI-specific questionnaire answers, signed.
Adds an underwriter-facing narrative formatted for AI E&O carriers — Munich Re aiSure, Mosaic, Vouch, Embroker, Lloyd's/Beazley.
For AI vendors selling into healthcare, financial services, legal, or government. Adds regulator-facing evidence and EU AI Act Article 50 mapping.
No form gate to see this. The deliverable is the deliverable.
| § 1 — Vendor & product summary | 2 pp |
| § 2 — Output sample review (50 samples) | 8 pp |
| § 3 — Failure-boundary summary | 3 pp |
| § 4 — Abstention & escalation policy | 2 pp |
| § 5 — Signed verification examples | 4 pp |
| § 6 — Procurement questionnaire | 6 pp |
| § 7 — Underwriter narrative | 5 pp |
{
"pack_id": "mejepa-ip-acme-2026Q2",
"tier": "carrier-ready",
"vendor": "Acme AI Inc.",
"product": "Acme AI Reasoner v3.2",
"issued": "2026-05-08T09:30Z",
"samples": 50,
"verdicts": { "pass": 48, "fail": 0, "abstain": 2 },
"coverage": 0.96,
"envelope": {
"in_distribution_rate": 0.93,
"ood_detection_auc": 0.91,
"named_failure_regimes": 4
},
"carrier_fit": [
"Munich Re aiSure",
"Mosaic AI E&O"
],
"ed25519_sig": "0x... (264 bytes)"
}
Insurance underwriters reject vendor self-attestation. They accept independent, signed evidence with named methodology — which is exactly what Mejepa produces.
Thirteen calibrated lenses, identical between releases. Same input, same reading — the foundation of any defensible underwriting opinion.
Published coverage rate. Abstention behavior is measurable, not asserted. This is the language Munich Re's aiSure already speaks.
Ed25519-signed, append-only. The carrier verifies offline using Mejepa's published public key — no vendor cooperation required at claim time.
Mejepa is the commercial productization of Teleox.ai — an independent research framework on meaning compression. The 13-instrument panel, conformal guard, and ed25519 witness chain are open-research primitives (Derived Data Abundance + Teleological Constellation Training). Underwriters can cite the methodology.
Procurement and underwriting language only — no marketing adjectives. Today is what lands in your buyer's queue. Tomorrow is what the methodology becomes as carriers price AI E&O at scale.
ConformalInterval { lower, upper, coverage_target=0.90, empirical_coverage } with a published 88–92% empirical tolerance band. Not generic "calibration."A productized 5-business-day engagement producing the signed evidence enterprise procurement and cyber-insurance underwriters demand: output sample review (50 production outputs), failure-boundary summary, abstention & escalation policy, signed verification examples (ed25519), procurement-ready evidence memo, and a carrier-facing narrative mapped to Munich Re aiSure, Mosaic, Vouch, Embroker, Lloyd's/Beazley. Three tiers: $3,500 / $6,500 / $10,000.
Munich Re's aiSure (Jan 2026) underwrites AI performance risk by requiring conformal-style operating-envelope evidence and disclosed abstention rates. Mejepa's named ConformalInterval{lower, upper, coverage_target=0.90, empirical_coverage} publishes coverage in the same statistical language. Tier II adds a 4–6pp underwriter-facing narrative explicitly mapped to aiSure terms and Mosaic's $15M AI E&O coverage requirements.
A statistical guarantee that a prediction's confidence interval will contain the true outcome with at least a specified probability (e.g., 90%). Mejepa publishes empirical_coverage measured on held-out calibration data with the packet. Underwriters underwrite named, falsifiable metrics. Mejepa's published commitments: OOD AUC ≥0.85, drift detection within 1,000 observations, per-cell ship-gate ≥0.95.
By Q2 2026 Fortune 1000 procurement teams routinely send AI-specific questionnaires (SIG-Lite + CAIQ + bespoke AI sections) asking about model failure boundaries, abstention rates, per-output evidence, and signed verification chains — questions not covered by SOC 2 or model cards. Mejepa's Procurement Pack ($3,500) bundles the answers as a signed, deliverable document the vendor attaches to the questionnaire response. Carrier-Ready Pack ($6,500) extends to the underwriter at the same time.
Procurement Pack — $3,500 (Series A/B AI vendors clearing enterprise procurement). Carrier-Ready Pack — $6,500 (adds underwriter-facing narrative). Regulated Vertical Pack — $10,000 (adds regulator-facing memo, EU AI Act Article 50 mapping, FRB SR 11-7 fit notes for healthcare/finserv/legal/government). All delivered in 5 business days.
Carriers underwriting AI E&O (Munich Re aiSure, Mosaic, Vouch, Embroker, Lloyd's) reject vendor self-attestation but accept independent signed evidence with named methodology. Conformal coverage is the actuarial primitive: a published probability that a model's prediction interval contains the truth. Without it, underwriters price for worst-case ignorance. With it, premiums are priced against a measurable, falsifiable performance baseline.
Yes, in the Regulated Vertical Pack (Tier III, $10,000). EU AI Act Article 50 — applicable to providers of general-purpose AI models with deployment-time enforcement on 2 August 2026 (deadline subject to Digital Omnibus deferral discussions) — requires machine-readable marking of AI-generated outputs and disclosure documentation. Mejepa's ed25519 witness chain maps directly to the Article 50 marking requirement; Tier III includes the regulator-facing memo for downstream deployers.
One AI product. One signed insurability pack in five business days. Buyer-procurement-ready, carrier-ready, GC-ready — without your team having to write it from scratch every time the deal asks for it.
INSURABILITY FOLIO · VOL · I · № 03 · [email protected]