GMP batch release · deviation investigation · OOS decisions

QP-level
decision support,
locally yours.

ChatQP translates EU GMP and FDA regulatory requirements into structured, defensible release decisions — grounded in a curated regulatory corpus, not a general-purpose language model.

Runs entirely on your hardware. No internet connection. No cloud. No data egress. Every output includes the regulatory basis, open questions, path to release, and a plain-English summary.

ChatQP does not replace the Qualified Person. Final batch certification decisions remain the legal responsibility of the QP under EU GMP Annex 16. All outputs are decision-support material subject to QP review and professional judgement.

Request Access See sample outputs ↓
Live example — OOS assay result
Input
Assay = 89.8% (spec: 90.0–110.0%)
Retest = 90.2%
No assignable cause identified
Output
INVESTIGATE
OOS without assignable cause. Retest cannot invalidate original result.
Rule: R1-OOS-NO-CAUSE · FDA OOS Guidance · EU GMP Chapter 6
305 Regulatory Documents
Hard GMP Rule Engine
EU GMP · FDA · ICH
No Internet Required
Draft Annex 22 Principles
Private Beta
EU GMP An. 22 Draft

Built ahead of EU GMP Annex 22

The European Commission's forthcoming Annex 22 is expected to introduce regulatory requirements for the use of artificial intelligence in pharmaceutical manufacturing and quality systems. ChatQP has been designed to anticipate these obligations — ensuring AI-assisted QP decisions remain explainable, auditable, and under qualified human control. Hard GMP rules are applied in Python, outside the model. Every output includes a full reasoning trace with source citations.

Human QP remains the decision authority
Full reasoning trace per output
Hard GMP rules enforced in code
No cloud processing of batch data
Source-cited, auditable outputs
Designed for QP review, not to replace it

Core Capabilities

01
⚖️

Batch Release Decisions

RELEASE / INVESTIGATE / REJECT positions grounded in EU GMP and FDA requirements. Hard GMP violations block release unconditionally. Soft flags generate a documented path to release, not a dead end.

GMP · Annex 16
02
🔍

OOS & Deviation Support

Structured investigation logic aligned with FDA OOS Guidance and EU GMP Chapter 6. Each output states the open question, the specific path to release, and required documentation — no generic boilerplate.

FDA OOS · Ch. 6
03
📐

Regulatory Retrieval

Authority-weighted retrieval across 305 classified documents. Binding regulations outrank guidance, which outranks literature. Domain gating ensures OOS queries pull FDA OOS Guidance and EU GMP Chapter 6 — not sterility annexes.

RAG · Authority-weighted
04
🔒

Air-Gapped by Design

Runs entirely on local hardware. Batch records, investigation reports, and quality data never leave your network. No internet connection is required at any stage. Designed to meet EU GMP Annex 22 data governance expectations for AI in GMP environments.

Local · Draft Annex 22

How it works

QP reasoning logic,
not a chatbot.

01
Classify the question

OOS / deviation / stability / data integrity. The question type determines which regulatory sources are retrieved and which rules apply. Irrelevant domains are blocked before retrieval begins.

02
Retrieve regulatory sources

Binding EU GMP and FDA sources are retrieved first. Authority-weighted — binding regulations outrank guidance; guidance outranks literature. 305 classified documents, nothing from the open web.

03
Apply decision logic

Hard GMP rules run in code — not by the model. OOS without assignable cause, sterility positives, and unqualified impurities block release unconditionally. Soft flags define the documented path to release.

04
Output structured decision

RELEASE / INVESTIGATE / REJECT. Every output includes the regulatory basis, open questions, required documentation, inspector challenge points, and a plain-English summary for non-QP stakeholders.

What ChatQP is — and is not

Built to support QP judgement,
not to replace it.

ChatQP IS
  • A decision-support tool for QP and QA professionals
  • A structured reasoning engine aligned with EU GMP and ICH frameworks
  • A system for faster, more consistent preliminary evaluations
  • A tool for generating inspection-defensible rationale
  • A retrieval system grounded exclusively in regulatory and scientific literature
  • Designed for use alongside — and subject to — QP professional judgement
ChatQP IS NOT
  • A replacement for a Qualified Person
  • A system that makes final batch release decisions
  • A source of regulatory authority
  • A substitute for product-specific knowledge
  • A general-purpose AI assistant
  • A cloud service — no data leaves your network

Built for real GMP scenarios

Practical support across
the quality lifecycle.

01
Batch Release Decisions
Structured QP position on release, hold, or investigation — with rationale, risks, and required actions.
02
Deviations & CAPA Support
Impact assessment logic for process deviations, unapproved changes, and documentation errors.
03
OOS / OOT Investigations
Phase I and Phase II investigation sequencing aligned with FDA OOS Guidance and EU GMP Chapter 6.
04
CMO Oversight
QP reliance assessments under Annex 16; evaluation of QTA prerequisites and site compliance status.
05
Specification & IMPD Alignment
Impurity qualification thresholds, specification review, and ICH Q3/Q6 framework alignment.
06
Data Integrity Review
ALCOA+ assessment for transcription errors, audit trail concerns, and LIMS data corrections.
Built on real-world QA/QP experience Grounded in EU GMP and ICH frameworks Focused on inspection-defensible reasoning No internet required at any stage

Regulatory alignment

Grounded in EU GMP and FDA regulatory expectations.

EU GMP Chapter 6
Out-of-Specification Investigations
Laboratory controls and OOS handling. Investigation requirements before batch release. Assignable cause obligations before invalidation or retest.
FDA OOS Guidance
Laboratory Controls / OOS Handling
Phase I and Phase II investigation logic. Prohibition on retesting without assignable cause. Full investigation sequencing and batch disposition requirements.
EU GMP Annex 16
Batch Certification by a QP
QP reliance framework for imported batches. Written confirmation requirements. QP assurance obligations and technical agreement prerequisites.
ICH Q1A / Q1E
Stability Testing & Extrapolation
Real-time data requirements for shelf-life claims. Accelerated study criteria for extrapolation. Post-approval stability commitment obligations.
ICH Q3A / Q6B
Impurity Qualification Thresholds
Identification and qualification thresholds. Toxicological assessment requirements. Oligonucleotide-specific impurity frameworks by analogy.
MHRA / FDA DI Guidance
Data Integrity
ALCOA+ principles. Audit trail requirements for manual data adjustments. Electronic batch record obligations and data lifecycle controls.

Actual outputs.
Regulatory decisions, not summaries.

Representative outputs from ChatQP's QP mode — generated locally, with no internet connection. Each shows the decision structure: recommended position, rationale with source tags, risks, required actions, and confidence level. Lightly trimmed for display.

chatqp — /qp mode — CPP excursion, no observed CQA impact
Query type: A (HIGH, keyword) | Batch decision | Domain: general | Sources: 2 [Scope filter: removed 6 out-of-scope source(s): annex17, annex16, annex15, stability draft...] [Sources: eu_gmp_part2_api.pdf, eu_gmp_ch4.pdf] ───────────────────────────────────────────────────────────────── Drug substance batch meets all specifications. CPP outside validated range, no observed impact on quality attributes. ─────────────────────────────────────────────────────────────────
## QP RECOMMENDED POSITION
RELEASE — all requirements met, QP can certify
The batch meets all release specifications and no impact on quality attributes is observed. Release is supportable with documented justification for the CPP excursion.
## RATIONALE
- EU GMP Part II states that prior to completion of concurrent validation, batches can be released where they meet all specifications and are thoroughly monitored. [REGULATORY — eu_gmp_part2_api.pdf]
- ICH Q6B specifies that specifications should reflect thorough process characterisation. Where validated ranges are set by process capability, exceeding them triggers evaluation — not automatic rejection. [GUIDANCE — ICH Q6B]
- Expert interpretation: with no observed CQA impact, release is supportable where the excursion is documented, bounded, and risk-assessed. [DERIVED — no direct source]
## KEY RISKS
- Undocumented excursion creates an audit exposure if challenged during inspection. - Potential long-term quality impact if the CPP excursion is a process drift signal rather than a one-off event. - Insufficient historical context may limit the strength of any risk-based justification.
## REQUIRED ACTIONS
(1) Raise a formal deviation documenting the CPP excursion, its magnitude, and timeframe (2) Prepare a written risk assessment confirming no impact on CQAs using available in-process and release data (3) Review historical trend data for this CPP across recent batches to confirm it is not a recurring pattern (4) QA review and QP approval of the full package before batch certification
## INSPECTOR CHALLENGE
An inspector would challenge the adequacy of the CPP excursion investigation and risk assessment, requiring written documentation of the justification, data review, and QA sign-off.
## CONFIDENCE [Medium]
Supported by EU GMP Part II; partially relies on expert interpretation. Confidence increases with a well-documented risk assessment and historical process data.
DECISION: RELEASE — batch meets specifications; CPP excursion justification required REASON: No observed CQA impact; EU GMP Part II supports release with thorough documentation NEXT ACTION: Raise formal deviation; prepare written risk assessment; QA/QP review
chatqp — /qp mode — component lot used without formal QMS release
Query type: A (HIGH, keyword) | Batch decision | Domain: general | Sources: 6 [Scope filter: removed 2 out-of-scope source(s): ich_q6a, ich_q1_draft] [Sources: eu_gmp_ch1, eu_gmp_ch4, eu_gmp_ch5, eu_gmp_part2_api, ICH Q9(R1)...] ───────────────────────────────────────────────────────────────── Clinical batch filled using a stopper from a qualified supplier. Specific lot not formally released in the quality system at time of manufacture. ─────────────────────────────────────────────────────────────────
## QP RECOMMENDED POSITION
INVESTIGATE — use of a component lot without formal QMS release requires investigation before a certification decision can be made.
## RATIONALE
- EU GMP Chapter 1 (Clause 1.8(vii)) mandates that significant deviations are fully recorded and investigated to determine root cause and appropriate corrective action. [REGULATORY — eu_gmp_ch1.pdf]
- ICH Q9(R1) requires that risks associated with deviations be assessed, documented, and acted upon. The risk assessment should consider the specific lot characteristics. [GUIDANCE — Q9(R1)]
- Expert interpretation: supplier qualification does not substitute for formal lot release — both are distinct GMP requirements. [DERIVED — no direct source]
## KEY RISKS
- Patient safety: an unreleased stopper lot could introduce variability not captured in incoming QC. - Compliance: failure to investigate a procedural deviation may be cited as a GMP finding at inspection. - Data quality: the absence of formal lot release creates a traceability gap in the batch record.
## REQUIRED ACTIONS
(1) Retrieve the CoA and QC records for the specific stopper lot used; verify against approved specification (2) Confirm retrospectively that the lot met all approved criteria at the time of manufacture (3) Prepare a written deviation investigation covering impact assessment and root cause (4) Raise a formal CAPA to prevent unreleased components entering production in future
## INSPECTOR CHALLENGE
An inspector would challenge whether the use of an unreleased lot was fully investigated, requiring documentation of lot-specific QC data, impact assessment, and corrective actions.
## CONFIDENCE [Medium]
EU GMP Chapter 1 and ICH Q9(R1) provide a clear procedural basis; confidence in the final position depends on the outcome of the lot-specific review.
DECISION: INVESTIGATE — component lot not formally released in QMS at time of manufacture REASON: Procedural deviation; supplier qualification does not substitute for lot release NEXT ACTION: Retrieve and verify lot CoA; prepare written impact assessment and deviation report
chatqp — /qp mode — OOS assay result, no assignable cause
Query type: A (HIGH, keyword) | Batch decision | Domain: analytical | Sources: 8 [Domain gate: blocked ich_Q1E, ich_Q3B — not relevant to OOS query] ⚠ HARD BLOCK: R1-OOS-NO-CAUSE — RELEASE → enforced INVESTIGATE ───────────────────────────────────────────────────────────────── Assay = 89.8% (spec: 90.0–110.0%), retest = 90.2%, no assignable cause identified ─────────────────────────────────────────────────────────────────
## REGULATORY POSITION
[REGULATORY — fda_oos.pdf] FDA OOS Guidance: A Phase I laboratory investigation must be completed before any retesting. The initial result cannot be discarded simply because a retest passes.
[REGULATORY — eu_gmp_ch6.pdf] EU GMP Chapter 6: OOS results must be fully investigated before a batch can be released. Retesting without a documented assignable cause does not constitute a valid investigation.
## QP RECOMMENDED POSITION
INVESTIGATE [HARD BLOCK — R1-OOS-NO-CAUSE]
PATH TO RELEASE: Conduct Phase I laboratory investigation to identify a specific, scientifically justified assignable cause. If Phase I finds no cause, a Phase II full manufacturing review is required before any release decision.
## CONFIDENCE [Low]
Hard block applied. No risk-based justification overrides this rule. Based on information provided; assumes no undisclosed investigation findings.
DECISION: INVESTIGATE — hard GMP rule violated, investigation mandatory REASON: OOS without assignable cause; retest cannot override original result NEXT ACTION: Conduct Phase I laboratory investigation to identify assignable cause
chatqp — /qp mode — Annex 16 QP reliance
Query type: A (HIGH, keyword) | Batch decision | Domain: general | Sources: 8 [Sources: eu_gmp_annex16, v4_an16, ICH_Q11, ich_Q6B...] ───────────────────────────────────────────────────────────────── Irish QP (MIA holder) receives siRNA batch certified by German QP at approved EMA site ─────────────────────────────────────────────────────────────────
## REGULATORY POSITION
[REGULATORY — eu_gmp_annex16.pdf] EU GMP Annex 16 §1.7: A QP in the EU/EEA may rely on written confirmation from a QP at the manufacturing site that each batch was produced and tested in accordance with GMP and the marketing authorisation. A formal Quality Technical Agreement must be in place.
## QP RECOMMENDED POSITION
RELEASE — batch meets all requirements, QP can certify
The Irish QP may release under Annex 16 §1.7 QP reliance, provided: 1. Batch Production Record reviewed and compliant 2. Certificate of Analysis reviewed — all parameters within specification 3. No open critical deviations flagged by the German QP 4. Quality Technical Agreement with the German site is current 5. German site is on the approved supplier list under this MIA
## CONFIDENCE [High]
Annex 16 reliance framework is well established. Regulatory basis is clear. Based on information provided; assumes no undisclosed deviations.
DECISION: RELEASE — batch meets all requirements, QP can certify REASON: Annex 16 QP reliance applicable; German QP certification and QTA in place NEXT ACTION: Proceed with batch certification — confirm QTA is current
chatqp — /qp mode — unqualified impurity above threshold
Query type: A (HIGH, keyword) | Batch decision | Domain: analytical/toxicology | Sources: 8 [Sources: ICH_Q3A, ich_Q6B, ich_M7, draft-guideline-oligonucleotide...] ───────────────────────────────────────────────────────────────── Unknown impurity 0.18% (qualification threshold 0.15%), no toxicology data available ─────────────────────────────────────────────────────────────────
## REGULATORY POSITION
[REGULATORY — ICH_Q3A(R2).pdf] ICH Q3A §4.2: Any impurity above the qualification threshold must be qualified. Qualification requires toxicological data or a structural alert assessment (ICH M7 QSAR / expert review).
[REGULATORY — ICH_Q3A(R2).pdf] ICH Q3A §4.2: An unidentified impurity above the qualification threshold cannot be released. Absence of toxicology data is not a documentation gap — it is a patient safety gap.
## QP RECOMMENDED POSITION
REJECT
The batch fails its implied specification. An unqualified unknown impurity at 0.18% cannot be certified. This is not a soft flag — no documented justification permits the QP to certify this batch in its current state.
Path forward (future batches): (1) Identify the impurity — chromatographic and MS characterisation (2) ICH M7 structural alert assessment for genotoxic potential (3) If no alert: qualify per ICH Q3A §4.2 toxicology data package (4) Amend specification to include an appropriate limit for this impurity
## CONFIDENCE [High]
ICH Q3A §4.2 is unambiguous. Based on information provided; assumes no undisclosed toxicology data exists.
DECISION: REJECT — specification failure, batch cannot be certified REASON: Unknown impurity 0.18% above qualification threshold; no toxicology data NEXT ACTION: Identify impurity and conduct ICH M7 structural alert assessment

The corpus

Regulatory depth at every layer.

305 regulatory documents, manually classified by domain and authority. Rank-weighted so EU GMP and ICH always lead. Nothing from the open web. Nothing unverified.

Domain-aware retrieval ensures each question type pulls the right sources. OOS queries retrieve FDA OOS Guidance and EU GMP Chapter 6 — not sterility annexes.

EudraLex Vol. 4 ICH Q7–Q14 FDA Guidance Annex 16 Draft Annex 22 GMP Annexes Oligo-specific Microbiology Toxicology Data Integrity
305
Source documents
5
Quality domains
0
Cloud dependencies
∞
Queries, locally

Built by a practitioner

Developed by a pharmaceutical quality and regulatory affairs professional with hands-on experience in GMP manufacturing, analytical chemistry, microbiology, and oligonucleotide drug substance quality — across clinical and commercial environments in Ireland.

ChatQP was built because existing tools treat pharmaceutical quality questions as information retrieval problems. They are not. They are decision problems with specific regulatory constraints, authority hierarchies, and professional accountability. The system reflects that distinction in how it reasons.

"

The gap between knowing the regulation and making the decision is exactly where QPs earn their certification.