Services · 15 ways across one gap

All services

Work · proof, not promises

All case studies

Insights · street talk, written down

Magazine

Company · AI in production since 2011

About Voidgap

Life sciences

Validated,
not just clever.

Pharma, biotech, MedTech, contract manufacturers and labs. Everything is documented, everything is audited, and GMP is getting an annex for AI.

01 · Topics

What keeps
you busy.

The topics that come up first when AI meets this industry. If one sounds familiar, that is where we start.

  • 01Validation eats the gain

    A tool that saves hours but cannot be validated saves nothing in a GxP process.

  • 02Documents without end

    Submissions, deviation reports, SOPs and literature reviews. Most of the time goes into writing and checking text.

  • 03Hallucinations are not an option

    In regulatory work, an invented reference is not a typo. It is a finding.

02 · Examples

Where AI
earns its keep.

Input, action, result. The places we would look first, because the work is repetitive, the data exists and a person still signs off.

  • 01Regulatory drafts

    Study data and templates → first draft of submission sections with references → regulatory affairs reviews and signs.

  • 02Deviation and CAPA support

    Deviation report → similar past deviations and root causes → a structured draft for the QA team.

  • 03Literature and safety screening

    Publications and case reports → relevant hits sorted and summarised with the source → pharmacovigilance decides.

  • 04Visual inspection

    Images from the filling line → defects classified by a validated model → the release stays with QA.

sample Illustrative examples, not client references.

03 · The rules

What applies,
and since when.

The rules we plan around in this industry. With the source, so your legal team can check our reading.

  1. 01
    GMP is getting an annex for AI

    The draft Annex 22 covers AI models in GMP-critical applications: static, deterministic models only, no generative AI in critical uses, defined test data and human review.

    EU GMP Annex 22 (draft, July 2025)final text pending
  2. 02
    Computerised systems stay validated

    Every system used in a GMP process needs validation, access control and an audit trail. AI does not replace that; it adds to it.

    EU GMP Annex 11in force
  3. 03
    MedTech AI is high-risk

    AI in medical devices or in-vitro diagnostics that needs a notified body is high-risk under the AI Act, on top of MDR or IVDR.

    Art. 6(1), Annex I AI Actfrom 2 August 2028
  4. 04
    NIS2 for manufacturers

    Manufacturers of pharmaceuticals and medical devices above the size thresholds must register, manage cyber risk and report incidents.

    NIS2UmsuCG, BSIGsince 6 December 2025

04 · Services

Where to
start.

Build, Secure, Comply, Enable: the four services we would start life sciences projects with.

05 · Next steps

What happens
after you send it.

Four steps between your request and a decision. No step commits you to anything.

  1. 01 You send the requestThree questions about your situation, then your contact details. It takes a few minutes.
  2. 02 We read it and replyA person reads your answers and replies by email, with questions or a proposal for a first conversation.
  3. 03 First conversationYour use case, your data, the rules that apply. Afterwards you know whether and how to start.
  4. 04 Written proposalIf it fits: scope, approach and effort in writing. You decide.

06 · Download

Take it
with you.

For your team, as a PDF. Leave your email and we send it to you.

In preparation

Readiness check: AI under GMP Annex 22

Where your planned AI use falls under the draft annex, and what validation it needs.

What is inside

  • Critical or not: how to classify the use case
  • What the annex expects from test data and monitoring
  • Where generative AI can still help outside critical steps

The PDF is being prepared. Request it now and you get it as soon as it is ready.

07 · Request

Tell us
where it hurts.

Three short steps. Your answers help us prepare the first conversation for your case.

  1. 01Your situation
  2. 02Your project
  3. 03Contact

Step 1 of 3 Your situation

Which segment are you in?

Where should AI help first?

08 · Questions

Questions
from this industry.

Short answers. Anything else goes into your request.

Can we use language models at all in a GxP environment?

Outside GMP-critical steps, yes, for example for drafts that a qualified person checks. In critical applications, the draft Annex 22 excludes them.

Who validates the AI system?

You stay responsible. We build and document so that your validation team can do its job: requirements, test sets, results and change control.

Is the annex final?

By our reading in September 2026 it is still a draft. We plan with the draft and adjust when the final text is published.

Life sciences

Your batch,
your void.

Tell us where it hurts. We check which rules apply before the first line of code.