Trust & security

Built for sensitive claim data.

Field photos, addresses, carrier names, and loss details run through Mensuria. We treat each as the sensitive record it is — isolated per account, encrypted in transit and at rest, never used to train a model, and never the final word over a licensed adjuster. Below is an honest account of what's built today and what's still on the roadmap.

Per-account isolation · Not used to train AI · You approve every estimate
Per-account
isolation, enforced in code
TLS
encrypted in transit & at rest
No
model training on your data
You
approve every estimate
What runs through Mensuria
Loss photosAddressesCarrier namesPolicy contextEstimateshandled as sensitive
How your data is protected today

Protections that are actually built in.

These are live in the product right now — ownership checked on every request, encryption you can't turn off, and location metadata stripped from photos. What isn't built yet is in the roadmap section below, labeled honestly.

Per-account isolation

Every claim, photo, and export is tied to your account, and that ownership is re-checked on every request — in the API, not just the UI. A non-owner gets a not-found, never another adjuster's file.

Encrypted in transit & at rest

All traffic runs over TLS, and data is encrypted at rest by our cloud provider. Photos are served only through authenticated, access-controlled requests — no public buckets, no anonymous file URLs.

Location metadata stripped

Field photos are re-encoded to remove EXIF/GPS metadata before they're stored or sent for processing, so we never persist or transmit the insured's location.

AI drafts, you approve

Mensuria proposes line items against a verified code base and never invents a selector it can't ground — it flags it instead. A licensed adjuster reviews and clears every flag before anything is final.

Deletable on request

Ask us to delete a claim or your whole account and we remove the photos, drafts, and exports within 30 days. Self-serve deletion and data export are on the roadmap below.

Least data, least access

We collect only what's needed to draft and bill — card details go straight to Stripe, never to us. It's a small team with minimal production access, and we don't sell your data.

Data lifecycle

Where your claim data lives, end to end.

Follow a single claim from the field to deletion. At every stage: encrypted, scoped to your account, and never mined into a training set.

See the full pipeline
01
Capture

Photos upload over TLS into account-scoped storage, and EXIF/GPS metadata is stripped on the way in.

02
Process

Vision and drafting run via Anthropic's API. Inputs draft your scope, are not used to train models, and are retained by the provider only transiently.

03
Review

Drafts live in your account only. The original AI draft is preserved separately from your edits, so you can always see what the AI proposed versus what you approved.

04
Retain

Active claims persist for as long as you keep the file, so history and exports stay available for your records and any rework.

05
Delete

Ask us and we purge the photos, drafts, and exports within 30 days. Automated self-serve deletion and backup purging are on the roadmap.

AI posture

Your claims are inputs, not training data.

The biggest fear with an AI tool is “where do my photos go?” The straight answer: they're sent to our AI provider once, to draft your claim, are never used for training, and are retained by the provider only transiently.

Never used to train models

Your photos, claims, and estimates are not used to train, fine-tune, or improve any AI model. Per our provider's (Anthropic's) API terms, API inputs are not used for training.

Used once, for your draft

Each photo is read only to draft that claim's line items. No cross-claim mining, no shared embeddings pool, no secondary use of your inputs.

Provider retention is transient

Anthropic may retain inputs briefly for abuse monitoring, then deletes them. A separately-contracted zero-retention agreement is on our roadmap — we won't claim it until it's signed.

No human eyes by default

We don't read your claim content to operate the product. We'd only look at a specific file with your explicit request to help troubleshoot.

Accountability model

AI drafts. A licensed adjuster approves.

Mensuria is a drafting tool, not a decision-maker. It accelerates the write-up and shows its work — but export stays locked until a human clears every flag. Your license, your judgment, your name on the estimate.

01
AI drafts

Mensuria reads the photos and proposes Xactimate-ready line items against a verified code base. It never invents a selector it can't ground — it flags it instead.

02
Evidence is attached

Every line carries the exact photos behind it and a confidence band. Anything uncertain surfaces as an open question rather than a silent assumption.

03
A licensed adjuster approves

Export stays locked until a human clears every low-confidence line and flag. The estimate ships under your license, with your judgment on every line.

mensuria.com / claims / CLM-2041
Approve & export
M. Alvarez · Water
State Farm · 118 Birch Ln, Montauk NY
In review
31 line items● 19 high● 4 low0 demoted
Laundry Room
WTR/EXTS80 SFWater extraction — Category 32 photos ✓
WTR/DRYWLF36 LFFlood cut drywall, up to 2ft
WTR/INS72 SFTear out & bag wet insulation
WTR/GRM72 SFApply anti-microbial agent
WTR/WTRDHM3 DADehumidifier (per 24 hr)
DRY/1/272 SFDrywall — hang, tape, float, sand
FNC/B36 LFBaseboard 2 1/4" — R&R — confirm
Hallway
WTR/EXTS72 SFWater extraction — Category 3
WTR/DRYWLF36 LFFlood cut drywall, up to 2ft
WTR/GRM72 SFApply anti-microbial agent
FCW/72 SFHardwood install — confirm code
+ 20 more lines · open questions (3)
Access

Only you reach your files.

Owner-scoped access

Every claim and photo is checked against its owner on each request. Ask for a file that isn't yours and you get a not-found — never another adjuster's data.

MFA available

Multi-factor sign-in is available through our auth provider (Clerk), and we recommend enabling it. Sessions are revocable if a device is lost.

Minimal production access

It's a small team. Production access is limited, and we'd only view a specific claim with your explicit request to help — not as standing access.

Sub-processors & status

Honest about where we are.

We'd rather show you the real status than flash a badge we haven't earned. Here's every vendor that may touch claim data, and exactly what's done versus planned.

Anthropic

AI model inference — reads photos + notes to draft line items

Clerk

Authentication & account management

Stripe

Subscription & payment processing (card data never touches our servers)

Render

Cloud hosting & data storage (United States)

Available now

Sub-processor list

A current list of every vendor that may process claim data — shown on this page and in our Privacy Policy, updated as it changes.

In progress

Data Processing Addendum

A DPA for firm accounts is being prepared. If your firm needs one to get started, contact us and we'll work through it.

Planned

Off-site encrypted backups

Automated off-instance backups (and deletion that propagates to backups) are being set up to harden durability before scale.

Planned

Formal certifications

We're an early, focused team and have not yet pursued SOC 2 or a third-party penetration test. We'll be transparent about timing as we grow — and won't flash a badge we haven't earned.

What we won't overclaim
We don't claim SOC 2, HIPAA, or PCI — claim photos aren't payment or health data, and we haven't pursued formal certification yet.
We don't claim a zero-retention guarantee with our AI provider until it's contracted — today, inputs are retained only transiently and never used for training.
We don't claim application-level encryption or an immutable audit log we haven't built — data is encrypted by our cloud provider and access is owner-scoped on every request.
We don't position the AI as the decision-maker. A licensed adjuster reviews and signs every estimate.
Responsible disclosure

Found something? Tell us.

If you believe you've found a vulnerability, email us and we'll respond quickly. We don't pursue good-faith researchers who report responsibly, and we'll keep you posted through remediation.

// responsible disclosure
security@mensuria.com
Good-faith reports welcome

Questions before a real claim?

We'll share our sub-processor list and answer anything your firm needs to know before you run sensitive data through Mensuria.

security@mensuria.com · sub-processor list above · DPA in progress