What the AI Act is and how it works
The Artificial Intelligence Act is an EU regulation, so it applies directly in every Member State without national transposition (European Parliament & Council of the European Union, 2024). It does not regulate "AI" as a technology. It regulates AI systems and general-purpose AI models according to the risk their use poses to health, safety and fundamental rights.
The structure has four layers. A small number of practices are prohibited. High-risk systems, defined through Annex I (products already covered by EU safety legislation) and Annex III (listed use areas such as education, employment and access to essential services), carry the heaviest obligations. Some systems carry only transparency duties under Article 50. Everything else is minimal risk and largely unregulated by the Act, although other law still applies (European Parliament & Council of the European Union, 2024).
Separately, providers of general-purpose AI models, the large models that underpin many chat and coding tools, have their own obligations under Chapter V, which have applied since 2 August 2025 (European Commission, n.d.-c).
The research exclusion: what Article 2(6) and 2(8) actually say
Two provisions matter most for doctoral researchers. Article 2(6) states that the 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" (European Commission, n.d.-a).
Article 2(8) states that the Regulation "does not apply to any research, testing or development activity regarding AI systems or AI models prior to their being placed on the market or put into service". It adds that such activities must comply with applicable Union law and that testing in real-world conditions is not covered by the exclusion (European Commission, n.d.-a).
Commentators have pointed out that these boundaries are hard to apply to modern AI research, where academic and commercial work overlap, models are reused across projects and "real-world testing" is not sharply defined (Meszaros et al., 2026). Treat the exclusion as a condition you need to document, not a blanket exemption for anything done at a university.
When you use AI tools rather than build them
Most PhD students are deployers or simply users of AI tools, not providers. Using a general-purpose chatbot to summarise papers or debug code does not, on its own, make you subject to high-risk obligations. Your obligations come mainly from your university, your funder, your publisher and data protection law.
Two parts of the Act still touch everyday research. Article 50(4) requires deployers who publish deepfakes to disclose that the content is artificially generated or manipulated, with lighter rules for evidently artistic or creative work, and requires disclosure of AI-generated text published to inform the public unless it has undergone human review or editorial control and someone holds editorial responsibility (European Commission, n.d.-b). If your project uses emotion recognition or biometric categorisation on participants, Article 50(3) requires you to inform the people exposed to it (European Commission, n.d.-b).
GDPR, and what changed in 2026
The AI Act sits alongside the GDPR rather than replacing it. Article 2(7) confirms that Union law on personal data, privacy and confidentiality of communications applies to personal data processed in connection with the Act (European Commission, n.d.-a). A research exclusion under the AI Act therefore says nothing about whether your processing of participant data is lawful; that question is answered under the GDPR and national research rules (European Parliament & Council of the European Union, 2016).
In July 2026 the EU adopted the Digital Omnibus on AI, Regulation (EU) 2026/1744, published on 24 July and in force from 27 July 2026 (European Parliament & Council of the European Union, 2026). It postponed the high-risk rules to 2 December 2027 for Annex III systems and 2 August 2028 for systems embedded in Annex I products, set 2 December 2026 as the marking deadline for generative systems already on the market, and added new prohibitions on AI that generates non-consensual intimate imagery or child sexual abuse material (Council of the European Union, 2026; European Commission, n.d.-c). It did not remove the research exclusion. Dates may still be refined by guidance, so check the official timeline before relying on a deadline.
