Answers · By Hugh Donatello, Atlas Bio · Updated 2026-10-29

EU AI Act and Pharma: Does It Cover Drug R&D?

Does the EU AI Act apply to AI used in drug discovery R&D? Sometimes. Article 2(6) of the AI Act, Regulation (EU) 2024/1689, excludes AI specifically developed and put into service for the sole purpose of scientific research and development. Recital 25 keeps any other AI system used for research in scope, so a commercial or general-purpose tool is not exempt just because the work is research. Typical discovery uses such as target identification and molecule design are not listed as high-risk, but AI literacy and transparency duties can still apply.

Who the AI Act applies to

Article 2(1) applies the Act to providers that place AI systems or general-purpose AI models on the EU market or put AI systems into service in the EU, wherever the provider is established, and to deployers of AI systems established or located in the EU. It also reaches providers and deployers outside the EU where the output of the AI system is used in the EU. A deployer, under Article 3(4), is any person or organisation using an AI system under its authority, other than for personal non-professional activity.

For a biotech or pharma company this usually means two roles. When it licenses or subscribes to an AI tool and uses it in its own work, it is a deployer of that tool. When it develops an AI system and puts it into service under its own name, including for its own use (Article 3(11) defines putting into service as supply for first use directly to the deployer or for own use in the Union for its intended purpose), it can also be a provider. Which role you hold for each tool decides which obligations could apply.

The research exclusion in Article 2(6) and its limits

Article 2(6) reads: 'This Regulation does not apply to AI systems or AI models, including their output, specifically developed and put into service for the sole purpose of scientific research and development.' Two conditions sit in that sentence. The system must be specifically developed for scientific research and development, and it must be put into service for that sole purpose.

Recital 25 explains the intent: the Act should respect freedom of science and not undermine research. It then sets the boundary: 'without prejudice to the exclusion of AI systems specifically developed and put into service for the sole purpose of scientific research and development, any other AI system that may be used for the conduct of any research and development activity should remain subject to the provisions of this Regulation.' In practice, a model a discovery team builds and runs only for its own research is the strongest candidate for the exclusion. A commercial platform, a general-purpose chatbot or a vendor tool that was developed for a wider market and is used by a research team does not become exempt because of how it is used.

Article 2(8) is a separate exclusion. The Act does not apply to research, testing or development activity regarding AI systems or models before they are placed on the market or put into service, but testing in real world conditions is not covered. Recital 25 adds that once a system that is in scope is placed on the market or put into service as a result of that research, the Act applies to it.

Article 3 does not define 'scientific research', and the text does not say whether commercially funded drug discovery counts. A broad reading that covers all AI used in medicines R&D is an interpretation, not settled law. If you rely on Article 2(6), record why each system qualifies, and review that record when a tool is reused outside research, shared with partners, or placed on the market. A model built for discovery that is later used in a clinical trial, for example to choose or stratify patients, should be reassessed: it now influences decisions about trial participants and may feed a regulatory submission, and it may qualify as a medical device or in vitro diagnostic.

Article 2(12) separately excludes AI systems released under free and open-source licences, unless they are placed on the market or put into service as high-risk systems, as prohibited practices under Article 5, or as systems covered by Article 50.

Is drug discovery AI high-risk?

Article 6 sets two routes to high-risk status. Under Article 6(1), an AI system is high-risk if it is a safety component of a product, or is itself a product, covered by the Union harmonisation legislation in Annex I, and that product needs a third-party conformity assessment. Under Article 6(2), the AI systems listed in Annex III are high-risk, subject to the Article 6(3) derogation for systems that do not pose a significant risk of harm, for example because they perform a narrow procedural task. An Annex III system that profiles natural persons is always high-risk.

Annex I includes the Medical Devices Regulation (EU) 2017/745 and the In Vitro Diagnostic Medical Devices Regulation (EU) 2017/746. It does not list EU medicines legislation. Annex III lists eight areas: biometrics, critical infrastructure, education, employment, access to essential services (including risk assessment and pricing for life and health insurance, and emergency patient triage), law enforcement, migration, and the administration of justice and democratic processes. Target identification, molecule design or assay analysis is not among them. The table shows where common uses tend to fall.

AI use in a drug discovery organisationLikely AI Act positionWhat decides it
In-house model built and used only for target or molecule researchCandidate for the Article 2(6) exclusionSpecifically developed and put into service for the sole purpose of scientific research; Article 3 does not define the term
Commercial or general-purpose AI tool used by a research teamIn scope; usually not high-riskRecital 25 keeps other AI systems used for research in scope; discovery is not an Annex III area
AI software that is, or is a safety component of, a medical device or IVD needing a notified bodyHigh-risk under Article 6(1)Annex I lists Regulations (EU) 2017/745 and 2017/746; only products that need third-party (notified body) conformity assessment meet Article 6(1); from 2 August 2028 the requirements must be met before the product is placed on the market or put into service
AI used in a clinical trial to select, stratify or allocate patients, or a biomarker algorithm developed into a companion diagnosticDepends on device qualification; timing of obligations depends on market statusHigh-risk under Article 6(1) if the software qualifies as a medical device or IVD needing a notified body (see MDCG 2019-11). From 2 August 2028 its AI Act requirements must be met before it is placed on the market or put into service, as part of the device conformity assessment, so plan the documentation during development, including the trial. During the trial an investigational or performance-study device is not yet placed on the market; Article 2(8) excludes pre-market testing but not testing in real world conditions, and how that applies to in-trial use is unsettled. Otherwise usually not high-risk, and whether Article 2(6) can still apply is unsettled. Clinical trials are not an Annex III area
AI used to screen job applicants or evaluate staffHigh-risk under Annex III point 4Employment is an Annex III area whatever the employer's sector; the Article 6(3) derogation is not available where the system profiles people; rules apply from 2 December 2027
Chatbot or text generator used by staff or the publicArticle 50 transparency duties can applyDuties fall mainly on the provider, so on you if you build it; deployers have duties for deep fakes and published public-interest text. Applies from 2 August 2026

Key dates after the AI Omnibus

The AI Act entered into force on 1 August 2024. The Commission proposed the AI Omnibus on 19 November 2025, a political agreement was reached on 7 May 2026, and the amending regulation entered into force on 27 July 2026. Article 113 as amended, with the transitional rules in Article 111, sets the dates below.

DateWhat appliesProvision
2 February 2025Chapters I and II, including the prohibited practices in Article 5 and the AI literacy duty in Article 4Article 113(a)
2 August 2025Obligations for providers of general-purpose AI models, and the governance rulesArticle 113(b)
2 August 2026The Act applies generally, including the transparency rules in Article 50Article 113
2 December 2026New prohibitions added by the Omnibus in Article 5(1)(ba) and (bb), covering AI that generates non-consensual sexually explicit or intimate content or child sexual abuse material; machine-readable marking deadline for generative systems already on the marketArticle 113(a); Article 111(4)
2 August 2027Deadline for general-purpose AI models placed on the market before 2 August 2025Article 111(3)
2 December 2027High-risk obligations for Annex III systemsArticle 113(c)(i)
2 August 2028High-risk obligations for Annex I product-related systems, such as AI in medical devices that need a notified bodyArticle 113(c)(ii)

Obligations that can apply even when nothing is high-risk

Article 4 requires providers and deployers to take measures to support the AI literacy of their staff and others operating or using AI systems on their behalf, taking into account their knowledge, experience and the context of use. As amended, it states that the duty does not require anyone to guarantee a specific level of AI literacy for any individual. For a research organisation using AI tools that are in scope, this is the duty most likely to apply today.

Article 50 applies from 2 August 2026. Providers must design AI systems that interact directly with people so those people know they are dealing with AI, unless that is obvious, and providers of systems that generate synthetic audio, image, video or text must mark the output in a machine-readable format. Deployers of deep fakes must disclose them. Deployers who publish AI-generated text to inform the public on matters of public interest must disclose that it is AI-generated, unless the text has gone through human review or editorial control and a person holds editorial responsibility for its publication.

Obligations on general-purpose AI models fall on the model provider. That does not mean a company building on such a model has none. A company that integrates a model into its own AI system is the provider of that system, a downstream provider under Article 3(68). A company that modifies or fine-tunes a general-purpose model can itself take on provider obligations; Recital 109 says these are limited to the modification or fine-tuning. Under Article 25(1), a deployer that changes the intended purpose of an AI system, including a general-purpose AI system, so that it becomes high-risk is treated as the provider of a high-risk system.

Article 5 prohibitions apply to every organisation within the Act's scope. One that matters for any employer is Article 5(1)(f), which bans AI systems that infer the emotions of a person in the workplace or in education, except for medical or safety reasons.

Other EU rules, and how the medicines regulator views discovery AI

The AI Act is not the only EU text that matters. The GDPR, Regulation (EU) 2016/679, applies on its own terms, so an AI system excluded under Article 2(6) is still subject to data protection law when it processes personal data such as genomic or patient data; Article 2(7) confirms that, apart from Articles 4a and 59, the AI Act does not affect it. A clinical trial that uses AI is also governed by the Clinical Trials Regulation (EU) No 536/2014.

The European Medicines Agency's Reflection paper on the use of artificial intelligence in the medicinal product lifecycle, adopted by the CHMP in September 2024, says AI in drug discovery can be of low regulatory impact if non-optimal performance only affects the developer. It adds that where results contribute to the body of evidence presented for regulatory review, the principles for non-clinical development should be followed, and models and datasets should be reviewed for bias. In other words, the evidential bar rises when discovery outputs move into a submission, whether or not the AI Act applies.

What to do now

Start with an inventory. For each AI system in your research and development work, note whether you built it or bought it, whether you are its provider or deployer, whether its output is used in the EU, and whether you claim the Article 2(6) exclusion and why. Flag anything touching medical devices, diagnostics, patient selection in clinical trials (which can make the software a medical device or IVD), hiring or staff evaluation, because those are the routes to high-risk status, and confirm nothing infers staff emotions, which Article 5(1)(f) prohibits. Put AI literacy measures in place for staff who use in-scope tools, and check Article 50 duties for any chatbot or content generator you provide. This page summarises the regulation and Commission material; it is not legal advice, and scoping decisions belong with your legal and regulatory functions.

Working out which of your AI systems sit inside the Act starts with knowing what you use and how. Atlas Bio's Secure AI for Life Sciences begins with an AI risk assessment that maps current AI use and data exposure, a starting point for the inventory described above. Status, October 2026. In use in Atlas Bio's own agent platform today: a separate verification step that checks agent work, and a hold that stops a task until a person decides. Built and tested on synthetic data, not yet used with a client: the AI risk assessment tooling (AI discovery, data-flow and vendor-terms mapping, and a NIST AI RMF gap report) and a record of what data an AI used, what it produced and who reviewed it. In development: a customer-controlled AI environment in your own Azure tenant, an AI gateway, agent permissions, evidence linkage and notifications. Each engagement states which controls are built. Decisions on regulatory scope and compliance stay with your legal and regulatory functions.

Is all pharmaceutical R&D exempt from the EU AI Act?

No. Article 2(6) excludes AI systems and models specifically developed and put into service for the sole purpose of scientific research and development. Recital 25 says other AI systems used to carry out research remain subject to the Act, so a commercial or general-purpose tool used in R&D is not exempt because of how it is used.

Is an AI model for molecule design or target identification high-risk?

Usually not. High-risk status comes from Annex I products that need third-party conformity assessment, such as medical devices and IVDs, or from the Annex III areas. EU medicines legislation is not in Annex I and drug discovery is not an Annex III area.

When do the high-risk rules apply after the AI Omnibus?

From 2 December 2027 for Annex III systems and from 2 August 2028 for Annex I product-related systems, under Article 113 as amended by the AI Omnibus, which entered into force on 27 July 2026.

Does the AI Act apply to a US company running drug discovery AI?

It can. Article 2(1)(c) covers providers and deployers established outside the EU where the output produced by the AI system is used in the EU, and Article 2(1)(a) covers providers placing systems on the EU market wherever they are established. The Article 2(6) research exclusion can apply to such a company on the same terms as to an EU one.

What does the AI literacy duty require?

Article 4 requires providers and deployers to take measures to support the AI literacy of staff and others using AI systems on their behalf, suited to their knowledge and the context of use. The duty has applied since 2 February 2025. Since the AI Omnibus took effect on 27 July 2026, Article 4 states that it does not require anyone to guarantee a specific level of AI literacy for any individual.

  1. AI Act Service Desk (European Commission): Article 2 - Scope (consolidated text as at 27 July 2026)
  2. AI Act Service Desk: Recital 25 - scientific research and development
  3. AI Act Service Desk: Article 3 - Definitions
  4. AI Act Service Desk: Article 6 - Classification rules for high-risk AI systems
  5. AI Act Service Desk: Annex I - List of Union harmonisation legislation
  6. AI Act Service Desk: Annex III - High-risk AI systems referred to in Article 6(2)
  7. AI Act Service Desk: Article 4 - AI literacy
  8. AI Act Service Desk: Article 50 - Transparency obligations
  9. AI Act Service Desk: Article 5 - Prohibited AI practices
  10. AI Act Service Desk: Article 25 - Responsibilities along the AI value chain
  11. AI Act Service Desk: Recital 109 - general-purpose AI model obligations and fine-tuning
  12. AI Act Service Desk: Article 111 - AI systems already placed on the market
  13. AI Act Service Desk: Article 113 - Entry into force and application
  14. European Commission: AI Act - regulatory framework and application timeline, including the AI Omnibus
  15. Medical Device Coordination Group: MDCG 2019-11 rev.1 - Guidance on qualification and classification of software under the MDR and IVDR (June 2025)
  16. European Medicines Agency: Reflection paper on the use of artificial intelligence in the medicinal product lifecycle (September 2024)

Talk to Atlas Bio.

Sponsors, researchers and clinicians: tell us the decision you are trying to make and we will show what the platform can and cannot answer.

Contact Us →