AI ASSURANCE · SPEQ SYNTHESIS

AI systems in GxP — the assurance registry

Regulators treat an AI system by what it decides, not what it is called. Each class below is framed for assurance — its Context of Use, the human oversight it demands, the AI-specific risks, and the qualification, evaluation, and monitoring it needs. The framing is SPEQ synthesis; the standards and frameworks it maps to are cited facts.

Generative & LLM

Regulatory RAG Copilothigh risk

A retrieval-augmented copilot that answers regulatory and SOP questions from a grounded corpus with cited sources.

Oversight: human in the loop
GxP GenAI Assistanthigh risk

A general GenAI copilot embedded in quality and regulatory workflows to draft, summarize, and reformat content.

Oversight: human in the loop

Agentic

Autonomous Quality Agentcritical risk

An agentic AI that takes actions in QMS workflows — triaging deviations, drafting CAPAs, and calling tools.

Oversight: human approval required

Computer Vision

ML Automated Visual Inspectionhigh risk

Computer-vision models that detect defects on the manufacturing line, informing accept, reject, and release decisions.

Oversight: human in the loop

Predictive ML

Predictive Process Modelhigh risk

ML predictive and soft-sensor models that inform process control, release decisions, or continued process verification.

Oversight: human in the loop
Pharmacovigilance Signal MLhigh risk

ML that detects safety signals and triages individual case safety reports in pharmacovigilance and drug safety.

Oversight: human in the loop
Document Intelligence & Extractionhigh risk

AI that extracts and classifies content from documents for batch-record review, complaint triage, and data entry.

Oversight: human in the loop
Clinical Trial AI Decision Supporthigh risk

AI for clinical trial operations — patient matching, risk-based monitoring, and operational decision support.

Oversight: human in the loop

Embedded Vendor AI

Vendor-Embedded AI Featureshigh risk

Third-party AI features built into validated GxP systems such as LIMS, eQMS, and MES copilots.

Oversight: human in the loop
ACROSS SPEQ’S AI ESTATE
HubAI in GxPWhere to start: the AI system registry, the principles, and the assurance toolchain.PrinciplesGood AI PracticeThe ten guiding principles FDA and EMA issued jointly in January 2026, read through a GxP lens.Standard · draftEU GMP Annex 22 — Artificial IntelligenceThe draft annex published for consultation on 7 July 2025: what it scopes in, and what it leaves out.StartupsBuilding an AI imaging productA founder’s journey from research result to regulated evidence, stage by stage.AdapterNVIDIA BioNeMo governance adapterHow a governed run of a pinned biology model is recorded, and what a change costs.