Pharma AI glossary
Generative AI
AI that produces new content, text, images, code, structures, rather than only classifying or predicting.
Definition
AI that produces new content, text, images, code, structures, rather than only classifying or predicting. The productivity revolution of 2023–2026 and the compliance headache of the same years: excluded from critical GMP applications in draft Annex 22, embraced (with human verification) for drafting and summarization everywhere else. The strategic point for pharma: the highest-value near-term uses are documentation-adjacent (drafting, summarizing, searching), which is precisely where verification discipline decides whether it's an asset or an incident.
Creation, not just classification
Generative AI produces new content (text, images, molecular structures, synthetic data) rather than only labeling inputs. In quality, it is only as GxP-ready as the harness and workflow around it: who may run it, what it may read or write, and what gets logged before a record is official.
GxP boundary
Annex 22 draft explicitly excludes generative AI and LLMs from critical GMP uses as written. That does not ban generative AI elsewhere (early research, non-GMP content) but draws a bright line for manufacturing decisions. Literacy is knowing which side of the line a use case sits on before validation spend begins.
Frequently asked questions
Is generative AI the same as an LLM?
LLMs are one generative modality; generative AI also includes image models, diffusion models, and structure generators. The GxP question is always: does this output directly influence a regulated decision?
Where is generative AI allowed in pharma today?
Broadly in non-GMP and discovery contexts with appropriate governance. In GMP, common patterns are documentation-adjacent: investigation and CAPA drafts, batch record summaries, change-control packages, and training content, each with human verification before the official record. Critical manufacturing decisions, release, and deterministic inspection under draft Annex 22 still favor traditional ML or locked models, not open-ended generation.
What GMP uses are usually out of scope for generative AI?
As draft Annex 22 is written: critical applications requiring static, deterministic models, direct batch release decisions, and probabilistic generative outputs without qualified human gates. Literacy means classifying the use case before buying or validating a tool.
Related terms
Benchmark your AI literacy
Free 20-question assessment, eight modules, and instant scoring, built for pharma quality and manufacturing teams.
Open the Literacy CenterBrowse all terms
See the full 71-term glossary index or search interactively in the Literacy Center.
Back to glossary hub →Educational content only, not legal or regulatory advice. Regulatory guidance cited here includes drafts (FDA AI credibility guidance; EU GMP Annex 22) as of August 2026; verify against final texts before relying on them in submissions. Company-reported figures (Merck CSR timings, Sanofi results) are labeled where used. MIT's ~95% pilot figure carries its own caveat (~150 interviews, contested definitions, not peer-reviewed). Re-check sources on module finalization.