⬡ Askhive
Askhive

Security & data handling

Plain language: where your data lives, what the engine sees, and what we never do with it.

Last updated: 8 September 2026

The short version

Your reports are yours. They are isolated per account, encrypted in transit, never used to train anyone's model, and identifiers Askhive recognises in a professional's labelled text are removed before the engine sees it.

Isolation

Every row in our database is protected by row-level security, so one account cannot read another's data even if a query goes wrong. A Practice workspace is a further boundary: reports, clients and drafts belong to the workspace, and only its owner and invited seats can reach them.

What the engine sees — and what it does not

Before any professional's text reaches the model, we replace identifiers with placeholders: social security numbers, EINs, account and routing numbers, and dates of birth — and client names when the professional asks for it. The model works with the placeholder; the real value is put back in the finished report. The professional is shown exactly what was withheld.

What this does not cover. Identifiers Askhive recognises in labelled text are removed before the text goes to the model provider; attachments and free text may still carry them, so a firm should send no more identifying detail than the task needs. The redaction reads text, not images: an identifier inside an attached document or a scan is not detected.

We do this because the model does not need the number to do the work. Nothing is lost by withholding it, so withholding it is the default.

No training on your data

Our model provider is contractually prohibited from training on data sent through our API. We do not sell data, we do not share it for advertising, and we do not run ads.

Shared report links

A shared report is reachable by anyone holding its link. Practice links carry a six-digit passcode by default, and we tell the professional to send it separately from the link — a passcode travelling in the same message protects nothing. We record when a link is opened, so the sender knows it arrived; we do not use tracking pixels.

Approval before anything sends

Askhive Practice is drafting software. Nothing reaches a client until the professional has read it and approved it — the gate is enforced in the database, not just in the interface, and the system records what the model drafted versus what the professional changed.

Retention and deletion

Reports are kept until you delete them or close your account. Account deletion is in the app and removes your profile, every report, and any Practice workspace with its clients. Backups roll over on their normal cycle.

A file is removed from the live store the moment it is deleted. Askhive also keeps its own nightly copy of every file — the provider's database backups hold the records of files, not the files — so a deleted file can still exist in that copy for up to 30 days, until the copy's own retention rule removes it. The copy is a private bucket: no client and no signed-in person can read it, and only a restore run by Askhive can reach it.

Email your clients forward to you

When you forward a client's email into Practice it passes through our mail provider, and two copies exist for a while, not one: the mail provider keeps a copy for thirty days and its backups for seven more; we delete our own copy at once, and ask the provider to remove a specific message sooner when a firm asks us to. We never write the original message down — what we hold is the draft you then edit. The provider has no deletion API for received mail, so an earlier removal is a request we make to them for you.

Who we rely on

Anthropic (the engine), Supabase (database, sign-in), Vercel (this website and shared reports), Resend (email), and Apple, Google and Stripe (payments). Card details never reach our servers.

Telling us about a problem

If you think you have found a security issue, write to support@bridgeshorelabs.com with the details. We will acknowledge it and tell you what we are doing about it. Please give us a reasonable chance to fix it before making it public.

What we do not claim

We are a small company and Askhive is a young product. We have not completed a SOC 2 audit and do not claim one. The Medical vertical pack ships as roadmap content and is not cleared for protected health information until the HIPAA/BAA work is done.

For a firm's security plan

The security fit sheet answers a questionnaire area by area and names what Askhive does not have; data controls lists every control a firm holds; consent on file is for the professions that need a client's written consent. All three are generated from the one file that records what is true, dated.