This is the fact check of the episode as published. Each factual claim was extracted from the script and verified by an AI agent that saw only the claim and the primary document (the PDFs in the corpus). Verdicts: SUPPORTED (the document says it), PARTIAL (supported with a difference, noted), NOT-IN-CORPUS (the source is not among the primary documents on disk), NOT-CHECKABLE (an estimate or a characterisation). Opinions voiced by the hosts are listed but not verified. This report is itself AI output and can be wrong; corrections: jack@jackprior.ai.
These statements stand, but the primary documents on disk do not themselves confirm them: they come from the landscape reference (a working index whose rows are checked against the web), are estimates, or are hedged in the episode as such.
| Line | Speaker | Claim | Verdict | Why it stands | Evidence |
|---|
| 6 | SARAH | The Guiding Principles of Good AI Practice in Drug Development were issued by EMA and FDA together in January 2026. | PARTIAL | Joint attribution is not in the text itself (agency identity is in the PDF logos); accepted as in episode 01 | Title 'Guiding principles of good AI practice in drug development', January 2026; the text never names EMA or FDA (footnote 1 refers to US and EU drug definitions; 'this initial collaborative work') |
| 6 | SARAH | ICH's reflection paper on proposed guideline work for advanced pharmaceutical manufacturing was endorsed by the ICH Assembly on 8 October 2025 and published the following March. | PARTIAL | March publication is landscape-sourced (ICH URL path 2026-03) | Header 'Endorsed by the ICH Assembly on 8 October 2025'; no publication date in the text |
| 14 | SARAH | The joint principles were published on 14 January 2026 (per the landscape). | NOT-CHECKABLE | Hedged in-script ('the landscape has the fourteenth') | Document dated 'January 2026'; the day is the landscape's |
| 14 | SARAH | Per the landscape, the joint principles are the first joint EU and US statement on AI. | NOT-IN-CORPUS | Hedged in-script; the text says 'initial collaborative work' | p.1: 'this initial collaborative work can inform our broader international engagements'; the text never says 'first' or names the parties |
| 26 | SARAH | As of the landscape's last check in August 2026 no ICH topic had been adopted. | NOT-CHECKABLE | Landscape-sourced, spoken as such | Landscape (27 Aug 2026): 'no topic has been formally adopted'; consistent with the paper's 'would begin with a new topic proposal' |
| 46 | SARAH | Continued or ongoing process verification has been the floor for every commercial process in both regions since 2011. | SUPPORTED | Dated by the FDA guidance; the EU term arrived with Annex 15 in 2015, though ongoing review was already expected. | FDA guidance January 2011 Stage 3; Annex 15 (October 2015) 5.28-5.32; earlier EU expectation via Chapter 1 PQR. 'Since 2011' is the FDA date; EU codified the term in 2015. |
| 50 | SARAH | CDER's 2026 guidance agenda lists a planned draft titled AI and ML Quality Considerations in Pharmaceutical Manufacturing; published early 2026, February as understood; not yet published. | NOT-IN-CORPUS | Hedged in-script ('as I understand it') | CDER guidance agenda not on disk; landscape watchlist |
| 54 | SARAH | FDA's January 2025 draft is still a draft as of August 2026 with no date for a final; a possible EMA follow-up on adaptive and generative AI is listed by the landscape. | NOT-IN-CORPUS | Landscape-sourced, spoken as such | Landscape watchlist |
| 56 | SARAH | The Digital Omnibus on AI, in force since late July 2026, deferred high-risk obligations to 2 December 2027 for Annex III use cases and 2 August 2028 for Annex I products; the landscape's EU law section carries the new dates. | NOT-IN-CORPUS | Landscape-sourced and attributed as such; the AI Act text on disk is the 2024 OJ text | Landscape §3.3 (Reg. (EU) 2026/1744); the AI Act text on disk is the 2024 OJ text. The landscape's watchlist row was corrected on 6 Sep 2026 and now carries the new dates |
| 69 | SARAH | The landscape confirms the Airlock launched, its second phase is complete, and it is funded through 2029. | NOT-CHECKABLE | Landscape-sourced, spoken as such | Landscape MHRA row |
| 115 | SARAH | BioPhorum's risk guidance is dated June 2026. | PARTIAL | Document footer May 2026; the June release date is landscape-sourced (same as episode 1 line 60, episode 3 line 9) | Footer on every page: '©BioPhorum Operations Group Ltd \| May 2026'; landscape gives the release as 2 June 2026 |
| 115 | SARAH | The guiding principles were issued jointly by EMA and FDA. | PARTIAL | Joint attribution is not in the text itself (same as line 6 of this episode) | Title 'Guiding principles of good AI practice in drug development', January 2026; the text never names EMA or FDA (footnote 1 refers to US and EU drug definitions; 'this initial collaborative work') |
| Line | Speaker | Claim | Verdict | Evidence |
|---|
| 8 | SARAH | The joint principles are two pages, ten numbered principles, and say they are intended to lay the foundation for developing good practice. | SUPPORTED | p.1: 'These 10 guiding principles are intended to lay the foundation for developing good practice'; principles 1–10 on p.2 |
| 14 | SARAH | The joint principles define AI as system-level technologies used to generate or analyse evidence across the drug product life cycle, listing nonclinical, clinical, post-marketing and manufacturing phases, in the first paragraph. | SUPPORTED | p.1 first paragraph, verbatim |
| 16 | SARAH | The joint principles say the ten principles are intended to lay the foundation for developing good practice and identify areas where international regulators, standards organisations and other collaborative bodies could work: research, educational tools, international harmonisation, and consensus standards, which may help inform regulatory policies and regulatory guidelines in different jurisdictions. | SUPPORTED | p.1: 'lay the foundation for developing good practice'; 'research, creating educational tools and resources, international harmonisation, and consensus standards, which may help inform regulatory policies and regulatory guidelines in different jurisdictions' |
| 18 | SARAH | Principles 1-5: human-centric by design; risk-based approach with proportionate validation, risk mitigation and oversight based on the context of use and determined model risk; adherence to standards naming GxP; clear context of use defined as role and scope for why it is being used; multidisciplinary expertise covering both the AI technology and its context of use integrated throughout the life cycle. | SUPPORTED | p.2 principles 1–5, verbatim |
| 20 | SARAH | Principle 6: data source provenance, processing steps and analytical decisions documented in a detailed, traceable and verifiable manner, in line with GxP requirements. Principle 7: model design and development practices, leveraging fit-for-use data, considering interpretability, explainability and predictive performance. Principle 8: risk-based performance assessments evaluate the complete system including human-AI interactions, using fit-for-use data and metrics appropriate for the intended context of use. | SUPPORTED | p.2 principles 6–8, verbatim |
| 22 | SARAH | Principle 9: risk-based quality management systems throughout the life cycle with scheduled monitoring and periodic re-evaluation, example data drift. Principle 10: clear, essential information in plain language about context of use, performance, limitations, underlying data, updates, and interpretability or explainability. | SUPPORTED | p.2 principles 9–10, verbatim |
| 23 | HOST | The joint principles say good practice and consensus standards must evolve as the use of AI evolves, and name harmonisation as work still to be done. | SUPPORTED | p.1: 'As the use of AI in drug development evolves, so too must good practice and consensus standards'; 'international harmonisation' listed as an area to work on |
| 115 | SARAH | The EMA-FDA guiding principles are two pages. | SUPPORTED | pdfinfo: 2 pages; read in full (episode notes) |
| Line | Speaker | Claim | Verdict | Evidence |
|---|
| 10 | SARAH | The EU AI Act has a sandbox article. | SUPPORTED | Article 57 'AI regulatory sandboxes'; Art. 3(55) definition |
| 56 | SARAH | The AI Act has been in force since 1 August 2024. | SUPPORTED | Art. 113 (twentieth day after OJ 12.7.2024); Art. 97(2) 'from 1 August 2024' |
| 56 | SARAH | Almost nothing in a pharma plant is high-risk under the AI Act unless it is a safety component of a regulated product, a medical device or a machine, that needs third-party conformity assessment. | SUPPORTED | Art. 6(1)(a)-(b): safety component of a product under Annex I legislation requiring third-party conformity assessment; Annex III adds other areas |
| 73 | SARAH | AI Act Article 57: every member state must have at least one AI regulatory sandbox operational by 2 August 2026; a sandbox is a controlled environment facilitating development, training, testing and validation of innovative AI systems for a limited time before market placement under a sandbox plan agreed between provider and competent authority; may include 'testing in real world conditions supervised therein'; the provider leaves with an exit report that market surveillance authorities and notified bodies must take positively into account; an objective is evidence-based regulatory learning. | SUPPORTED | Art. 57(1), (5), (6), (7), (9)(d); quoted phrase verbatim |
| 75 | SARAH | The Article 57 sandbox is supervised by the AI Act competent authority for compliance with the AI Act; the landscape describes it as horizontal, not GMP. | SUPPORTED | Art. 57(6): supervision 'in relation to the obligations and requirements of this Regulation and, where relevant, other Union and national law' |
| 115 | SARAH | Article 10 of the EU AI Act is on data. | SUPPORTED | Article 10 heading: 'Data and data governance' |
| 115 | SARAH | Article 14 of the EU AI Act is on human oversight. | SUPPORTED | Article 14 heading: 'Human oversight' |
| 115 | SARAH | The EU AI Act's high-risk scope mostly does not cover pharmaceutical manufacturing. | SUPPORTED | Annex III lists eight areas (biometrics, critical infrastructure, education, employment, essential services, law enforcement, migration, justice); no entry for pharmaceutical or medicines manufacturing; the only health entry is 5(d) 'emergency healthcare patient triage systems' |
| Line | Speaker | Claim | Verdict | Evidence |
|---|
| 6 | SARAH | ICH's reflection paper on proposed guideline work for advanced pharmaceutical manufacturing was endorsed by the ICH Assembly on 8 October 2025 and published the following March. | PARTIAL | Header 'Endorsed by the ICH Assembly on 8 October 2025'; no publication date in the text |
| 8 | SARAH | The ICH reflection paper's last paragraph says ICH should consider the topics during the annual new topic selection process. | SUPPORTED | p.11, final sentence: 'ICH should consider these topics during the annual ICH new topic selection process' |
| 26 | SARAH | The ICH paper is titled Reflection Paper on Proposed ICH Guideline Work to Facilitate the Adoption of Advanced Pharmaceutical Manufacturing; endorsed 8 October 2025; its last paragraph says ICH should consider these topics during the annual new topic selection process. | SUPPORTED | Cover title and subtitle; header endorsement date; p.11 last sentence |
| 28 | SARAH | The ICH paper's first paragraph says international manufacturers have noted the lack of global regulatory alignment as one reason for not pursuing and adopting advanced manufacturing technologies, and its first example is process modelling including AI-based models. | SUPPORTED | p.1 first paragraph, verbatim (footnote 1: ISPE April 2024) |
| 30 | SARAH | The three topics are process modelling, continuous process verification, and decentralised or distributed manufacturing, in that order, stepwise, possibly three separate guidelines; process models might be considered digital representations of physical manufacturing processes, used for design, scale-up, site transfer, monitoring and control, and can become a critical element of the control strategy. | SUPPORTED | p.2 three topics in that order; p.9 'phased, stepwise approach … (e.g., three separate guidelines)'; p.3 process-model description |
| 32 | SARAH | The ICH paper credits the Points to Consider with the principle that a model's impact guides the extent of regulatory oversight; names a need for global harmonisation on terminology, model risk framework, basis for regulatory oversight and data requirements; says manufacturers require guidance on regulatory notification of model updates considering model risk and maturity of a site's quality system; and says the Points to Consider does not address linking model risk to model validation and lifecycle management activities. | SUPPORTED | p.3, all four elements verbatim |
| 34 | SARAH | The ICH paper says new types of AI models might further challenge the regulatory frameworks for process models; its footnotes point to the EU's consultation on Annex 22 and FDA's January 2025 draft; a new ICH guideline on process models could provide a comprehensive framework with principles applicable to AI models, recognising the Points to Consider did not explicitly foresee these new types of AI models. | SUPPORTED | pp.3-4; footnotes 4 (EU consultation on Chapter 4, Annex 11 and new Annex 22) and 5 (FDA January 2025 draft) attach to the 'regulators are actively developing regional guidelines' sentence, as the script says; 'principles that might be applicable to AI models' |
| 36 | SARAH | On page seven the ICH paper says the Points to Consider approach may not be best suited for AI models used as part of a dynamic control strategy and in continuous manufacturing; regulators and industry recognise the need for a modern risk-based classification of models and lifecycle management approach; AI or continuous learning models can pose a significant challenge because post-approval model verification must be balanced with ongoing changes as new information is generated. | SUPPORTED | p.7, verbatim (footnote 11) |
| 38 | SARAH | ICH Section C: a new guideline on process models would be a preferred first step; it would clarify the Points to Consider on regulatory expectations including 'documentation required in dossiers related to models and model updates over the lifecycle', and revise the Points to Consider for linking model risk to intended use and decision consequence. | SUPPORTED | p.9 Section C, quoted phrase verbatim |
| 40 | SARAH | The ICH paper's questions include expectations for in-process and release testing when a model controls the process; risk-based validation of manufacturing models; and lifecycle maintenance 'for frequently updating process models, e.g., AI models that learn and self-adjust'. | SUPPORTED | p.8 Process Modelling bullets; quoted phrase verbatim |
| 42 | SARAH | The ICH paper says continuous process verification, as defined in Q8, is briefly addressed in the Points to Consider, represents an advanced approach to process validation and has not been widely adopted; needs clarification on product and process understanding, monitoring and control strategy, and dossier information; its second question is whether regional approaches can be aligned. | SUPPORTED | pp.4-5; p.8 second CPV bullet 'Can regional regulatory approaches to continuous process verification be aligned?' |
| 48 | SARAH | The ICH paper says the timing to initiate this effort may be influenced by external factors including regional legislation or policy development, and topic prioritisation will be informed by regulatory experience. | SUPPORTED | p.9 Section C, verbatim ('informed by regulatory experience and scientific knowledge') |
| 66 | SARAH | The ICH reflection paper cites the QIG's preliminary considerations on pharmaceutical process models. | SUPPORTED | p.7 and footnote 12: EMA Preliminary QIG Considerations regarding Pharmaceutical Process Models (22 February 2024) |
| 71 | SARAH | The ICH paper lists an MHRA consultation among the avenues for advanced manufacturing, alongside FDA's FRAME initiative and Emerging Technology Program, CBER's advanced technologies team, EMA's QIG, and Japan's innovative manufacturing team. | SUPPORTED | p.10: FRAME, ETP, CATT, QIG, MHRA Consultation, PMDA IMTT |
| 77 | SARAH | The ICH paper says the lack of harmonisation across regulators' initiatives may discourage manufacturers. | SUPPORTED | p.11, verbatim |
| 108 | SARAH | The ICH paper puts the maturity of a site's quality system next to model risk as what notification will depend on. | SUPPORTED | p.3: 'considering model risk and maturity of a site's quality system' |
| 115 | SARAH | ICH's reflection paper on advanced manufacturing is short. | SUPPORTED | pdfinfo: 11 pages |
| 115 | SARAH | ICH's reflection paper says which words the next ICH topic will use. | PARTIAL | p.3: 'there is currently a need for global harmonisation on multiple aspects (e.g., terminology, model risk framework, basis for regulatory oversight, and data requirements)'; p.9: 'A new ICH guideline on process models ... would be a preferred first step in this effort.' |
| Line | Speaker | Claim | Verdict | Evidence |
|---|
| 52 | SARAH | Annex 22, Annex 11 and Chapter 4 were consulted together 7 July to 7 October 2025; ~1,300 comments on Annex 22; EMA workshop 30 June and 1 July 2026; draft not amended; finals targeted end 2026, effect around 2027 (estimates). [landscape] | NOT-IN-CORPUS | Partly on disk now. EMA minutes, 4 Feb 2026 (EMA/40804/2026), §3, p. 2: 'GMP Annex 22 on AI in manufacturing, which following public consultation received ~1,300 public comments and is undergoing revision. The final document is expected to be published by the end of the year.' — confirms the ~1,300 comments and a final expected by end-2026. EMA event page (EMA-2026-06-Annex22-Workshop-Page.pdf, captured 6 Sep 2026): 'EMA's Good Manufacturing Practice (GMP) / Good Distribution Practice (GDP) Inspectors Working Group is organising a two-day workshop to help shape a risk-based approach to the use of generative artificial intelligence (AI) in medicines manufacturing.' … 'The draft Annex 22 had indicated that dynamic, adaptive and probabilistic models - such as GenAI or LLMs - should not be used in critical GMP applications. EMA is still considering the implications of the stakeholder consultation results.' — confirms the 30 Jun–1 Jul 2026 workshop, EMA's Inspectors Working Group as owner, and that the July 2025 draft still stood with EMA 'still considering' the consultation. The rest: landscape Annex 22 row. |
| 114 | HOST | The season said it covers about forty documents. [series self-reference] | SUPPORTED | Episode 1, line 10: 'a working reference of about forty documents'; show description: 'at least forty regulatory guidances and white papers' |
| 115 | SARAH | The season read twenty-five of the forty-eight documents in the corpus closely; the rest (strategy papers, position statements, device-world texts) were cited rather than read. [episode JSON sources_read; ../AILandscape/docs] | PARTIAL | sources_read across the nine episode JSONs lists 27 distinct docs/ PDFs; docs/ holds 44 PDFs (43 usable, one excluded), 45 counting the excluded industry proposal outside docs/. The device-world texts (PCCP, GMLP, credibility guidance) each appear in the sources_read of two or three episodes |
| 115 | SARAH | Seven of nine episodes needed the FDA January 2025 draft. [episode JSON sources_read] | PARTIAL | sources_read lists the draft in 8 of 9 episodes (all but 03-grading); the eighth is this episode, whose sources_read gained it with the reading list and whose notes say it was cited by reference, not re-opened |
| 115 | SARAH | Six episodes used draft Annex 22. [episode JSON sources_read] | SUPPORTED | sources_read lists the Annex 22 draft in episodes 03, 04, 05, 06, 07 and 08 |
| 115 | SARAH | The documents are linked, in the spoken order, at the end of the show notes. [episode JSON reading_list; scripts/feed.py] | SUPPORTED | reading_list has 17 entries in the spoken order; feed.py renders them as 'Reading list, in order' at the end of the notes. Slot 2 renders as 'Annex 22 — Artificial Intelligence' (PIC/S docview 9715) |