Security & trust

Your data stays yours. We can prove it.

BioStudioAI runs on your servers. The AI gets a description of your data, never your measurements. Every security, privacy and compliance claim points to a test that ran.

Trust Center · 23 Oct 2026

  • 180 / 180security, privacy & accuracy tests passed Register
  • 28 / 28Part 11 · Annex 11 · ALCOA+ requirements verified Details
  • 83 / 85NIST certified statistics problems to ≥ 6 digits Table
  • 0numbers typed by the AI
How BioStudioAI uses AI

Everything Claude and ChatGPT do for your data. Plus everything they can't.

BioStudioAI uses the intelligence of Claude, ChatGPT and Gemini — without your measurements leaving your lab. Four steps, left to right.

  1. 1

    Your data stays in your lab

    Files and measurements stay on your servers, locked.

  2. 2

    Only the question goes to the AI

    Claude, ChatGPT or Gemini get a description and your question — not your numbers.

  3. 3

    The AI sends back advice

    Which test, which plot. A plan — never a result.

  4. 4

    BioStudioAI checks it and does the maths

    Every number is calculated from your data. Numbers typed by the AI are rejected.

✎ Everything the AI drafts is editable↶ One-click undo⏻ AI can be switched off🔒 Every AI call is logged, encrypted

Sending more than a description is a setting your admins control.

What goes where

Your admins decide what the AI sees.

Switch the setting and watch what crosses the boundary. In every mode the engines compute every number on your servers, and notebook content never goes to an AI model.

Admin setting
YOUR SERVERS AI MODELClaude · ChatGPTGeminichosen per teamby your admins Files & measurementsstay here, encrypted Enginescompute every number Notebook (ELN)never sent to AI BioStudioAI gateidentifiers codedevery call loggedAI-typed numbers rejected a plan: which test, which plot

Your data never leaves. We can't see it.

ProofInstalled in your environment, encrypted; the AI gets a description, not your values — 180/180 tests publishedTrust Center →
Security & data

Built for how pharma handles data.

For your IT, security and quality teams: where BioStudioAI runs, what the AI can and cannot see, and what gets recorded.

On your servers

The whole platform runs inside your environment — your data centre or your private cloud.

Encrypted — we never see it

Data is encrypted. BioStudioAI, the company, has no access to your files or results.

AI on your terms

Models get a description of your data, not your measurements. Admins choose which models each team may use — or switch AI off.

Full audit trail

Every AI call is kept in an encrypted log. Every analysis has version history; every exclusion has its rule and value.

Access you control

Per-project sharing and roles. People see what their project allows — nothing more.

No lock-in

Export to Excel, CSV, PDF, PowerPoint and SVG at any time. If you stop, your data never left your servers.

Security review packArchitecture overview, data-flow diagram and our answers to the usual security questionnaire. Working under GxP? Ask about validation support for your deployment.
Request the pack
Where it runs

Inside your environment.

The whole platform moves into your environment, encrypted. BioStudioAI, the company, has no access to your files or results.

Your data centre

On-premises, when policy requires it. Your network, your backups, your rules.

Your private cloud

In your own cloud account, under your own access policies.

Either way

Your data stays yours. We don't train on it, and we can't see it. Export everything, any time, in standard formats.

Tested, with evidence

Every claim points to a test that ran.

Our Trust Center publishes the test suites behind each security, privacy, accuracy and compliance claim — with dates, environments, commands to reproduce them and checksums on every document.

What the tests cover

  • Sign-inAdministrator two-factor codes (RFC 6238 vectors), replay and lock-out; sign-in never reveals which e-mails have accounts.
  • AccessRow-level security everywhere; notebook owners yes, strangers no; every notebook route checks access.
  • Audit trailAppend-only and hash-chained; tampering is detected; actions are refused if the trail is unavailable.
  • AI boundaryNo notebook content reaches an AI model; identifiers are coded before anything leaves; free text is masked too.
  • HardeningProduction refuses unsafe settings; CORS fails closed; SSRF blocked; outside code cannot run shell commands or read secrets.
Full test register →

Plainly: what we don't claim

  • Not heldSOC 2 · ISO 27001Not certified yet. The controls and their tests are documented in the Trust Pack.
  • Not assessedHIPAANot assessed for BioStudioAI.
  • SupportedGDPRData stays in your environment; identifiers are coded before AI.
  • Controls verified21 CFR Part 11 · Annex 11Software can't be "Part 11 certified": you become compliant by validating your use. We provide the controls and the evidence for your file.
The audit trail, live

Change one entry. The chain notices.

Each audit entry stores the SHA-256 of its content and of the entry before it. Edit any past entry and every link after it stops matching. Example entries; the hashes are computed in your browser right now — the same mechanism the notebook's audit trail uses.

Chain intactEvery entry matches the one before it.

In BioStudioAI the database also refuses UPDATE and DELETE on the trail, and the chain is re-verified every night. If the trail is unavailable, the action is refused.

Part 11 · Annex 11 evidence →
For your security team

The questionnaire, answered.

Short answers to what security reviews ask first. The details, tests and dates are in the Trust Center.

How is data encrypted?

AES-256-GCM encryption; codebooks for coded identifiers are stored encrypted, scoped and owner-controlled.

How do people sign in?

Administrators use two-factor codes (tested against RFC 6238 vectors, with replay protection and lock-out). Signatures need a password or passkey confirmation.

Who can see what?

Row-level security everywhere and per-project roles: people see what their project allows — nothing more. Every notebook route checks access.

Can the audit trail be changed?

No. It is append-only and SHA-256 hash-chained; the database refuses UPDATE and DELETE, and it is re-verified every night. If it is unavailable, the action is refused.

What does the AI provider receive?

By default a description of your data and your question — not your measurements. Identifiers are coded before anything leaves; notebook content never goes to an AI model.

Can we switch AI off?

Yes. Admins choose which models each team may use, or switch AI off; the engines keep working.

What certifications do you hold?

Not SOC 2 or ISO 27001 yet; HIPAA not assessed. The Trust Center lists the controls and the tests behind each claim.

Can we get a security pack?

Yes: architecture overview, data-flow diagram, our answers to the usual questionnaire, and the 41-page Trust Pack.

Next step

For your IT, security and quality teams

Architecture overview, data-flow diagram, our answers to the usual questionnaire and the Trust Pack.

Request a demo