EU AI Act and GDPR: Omnibus Dates and Privacy Rules (2027)
Independently fact-checked against primary sources (last audited November 5, 2026). · Law checked current as of November 5, 2026. · 48 primary sources cited on this page. How we verify our legal content

As of November 4, 2026, the EU AI Act (Regulation (EU) 2024/1689) is in force and has applied since 2 August 2026 with some exceptions, but its main high-risk rules are not yet in application: Regulation (EU) 2026/1744 set them for 2 December 2027 (Annex III) and 2 August 2028 (Annex I). The GDPR keeps applying alongside it. Article 2(7) says the AI Act "shall not affect" the GDPR, "[w]ithout prejudice to Articles 4a and 59" of the Act.
Status as of November 4, 2026: Regulation (EU) 2026/1744, the Digital Omnibus on AI, was published on 24 July 2026 and has been in force since 27 July 2026. Two new Article 5 points apply from 2 December 2026. The separate GDPR proposal, procedure 2025/0360(COD), is not adopted: Parliament's procedure file showed "Awaiting committee decision" when we read it on November 4, 2026.
This page covers the EU-level AI Act and how it meets the GDPR. It does not cover national implementing laws in each Member State. For the data-protection rules that predate the AI Act, see our guide to EU data privacy laws and the what is GDPR explainer.
What Is the EU AI Act?
Regulation (EU) 2024/1689 is a risk-based regulation of AI systems and general-purpose AI models. The European Commission describes it as "the first-ever comprehensive legal framework on AI worldwide." It sorts AI into four levels: unacceptable risk (banned), high risk (strict duties), transparency risk (disclosure duties), and minimal or no risk (no rules specific to that level, though the Article 4 AI literacy duty covers AI systems generally).
Its reach is not limited to EU companies. Article 2(1) covers providers placing AI systems or general-purpose AI models on the EU market wherever they are established, deployers established or located in the EU, and providers and deployers in a third country where the system's output is used in the EU.
The Act entered into force on 1 August 2024. Application then rolls out in phases under Article 113, as amended in July 2026.
EU AI Act Timeline After the Digital Omnibus (Regulation (EU) 2026/1744)
Regulation (EU) 2026/1744 was signed on 8 July 2026, published in the Official Journal on 24 July 2026, and entered into force on 27 July 2026, under its Article 4 rule of "the third day following that of its publication." Here is the calendar as the amended Act now reads.
| Date | What applies | Provision | Status on November 4, 2026 |
|---|---|---|---|
| 1 August 2024 | AI Act enters into force | Art. 113 | In force |
| 2 February 2025 | General provisions and the original Article 5 prohibitions | Art. 113(a) | Applies |
| 2 August 2025 | General-purpose AI rules, governance, penalties (except Art. 101) | Art. 113(b) | Applies |
| 27 July 2026 | Articles 102 to 110 | Art. 113(d) | Applies |
| 2 August 2026 | General application date, including Article 50 transparency | Art. 113 | Applies |
| 2 December 2026 | New Art. 5(1)(ba), (bb), (1a), (1b) | Art. 113(a) | Scheduled |
| 2 December 2026 | Art. 50(2) for synthetic-content systems on the market before 2 August 2026 | Art. 111(4) | Scheduled |
| 2 August 2027 | General-purpose AI models on the market before 2 August 2025 | Art. 111(3) | Scheduled |
| 2 August 2027 | At least one regulatory sandbox per Member State | Art. 57(1) | Scheduled |
| 2 December 2027 | High-risk rules, Art. 6(2) and Annex III | Art. 113(c)(i) | Scheduled |
| 2 August 2028 | High-risk rules, Art. 6(1) and Annex I | Art. 113(c)(ii) | Scheduled |
| 2 August 2030 | High-risk systems intended for use by public authorities | Art. 111(2) | Scheduled |
| 31 December 2030 | AI components of large-scale IT systems in Annex X, placed before 2 August 2027 | Art. 111(1) | Scheduled |
What Already Applies
AI literacy (Article 4): The Omnibus rewrote this duty. Providers and deployers now "shall take measures to support the development of AI literacy" of their staff and others operating AI for them, and the duty "does not require providers or deployers to guarantee any specific level of AI literacy of any individual." The Commission and Member States support those efforts under Article 4(2).
Article 5 prohibitions: The original eight have applied since 2 February 2025. The Commission says that from 2 August 2026 the AI Office and Member State authorities are responsible for implementing, supervising and enforcing the Act.
General-purpose AI: Provider duties have applied since 2 August 2025. Article 113(b) excepts Article 101 from that date, so the Commission's power to fine these providers falls to the general date of 2 August 2026.
Article 50 transparency: These duties have applied since 2 August 2026, and Article 113 gives them no later date. The one transition is Article 111(4), covered in the Article 50 section below.
What Is Not Yet in Application
High-risk systems: Chapter III, Sections 1, 2 and 3 (except Article 6(5)) apply from 2 December 2027 for Annex III systems and 2 August 2028 for Annex I systems. That is where Article 10 data governance, Article 26 deployer duties and Article 27 fundamental rights impact assessments sit. Any source giving 2 August 2026 as the start of all high-risk rules is out of date.
Systems already on the market: Under Article 111(2), the Act reaches operators of high-risk systems placed on the market before those dates only if the systems are subject to significant changes in their designs from that date. Providers and deployers of high-risk systems intended for use by public authorities must in any case comply by 2 August 2030.
Regulated products: Annex I, Section A still lists products such as toys, lifts, radio equipment and medical devices. The Omnibus deleted machinery from Section A and added the Machinery Regulation, (EU) 2023/1230, to Section B. For high-risk AI systems related to Section B products, Article 2(2) applies only Article 6(1), Article 60a and Articles 102 to 112 of the Act, plus Articles 57 to 59 in so far as the high-risk requirements have been integrated into the sectoral law.
The AI Act Does Not Replace the GDPR
Article 2(7) is the operative rule: Union law on personal data, privacy and confidentiality of communications applies to personal data processed in connection with the Act, and, "[w]ithout prejudice to Articles 4a and 59," the Act "shall not affect" the GDPR, Regulation (EU) 2018/1725, Directive 2002/58/EC or Directive (EU) 2016/680.
Recital 10 says it in plainer terms: the Act "does not seek to affect the application of existing Union law governing the processing of personal data." It adds that data subjects keep their rights, "including the rights related to solely automated individual decision-making, including profiling."
Two Overlapping Regimes
An AI system that processes personal data needs a lawful basis under Article 6 GDPR, and a condition under Article 9 GDPR if it processes special-category data. It remains subject to purpose limitation and data minimisation under Article 5(1)(b) and (c) GDPR.
Meeting the GDPR does not settle the AI Act either. A deployer may have a lawful basis for processing and still be using a practice Article 5 of the AI Act prohibits. The GDPR governs the processing of personal data, while the AI Act governs the AI system itself by its risk level.
Article 4a and Article 59 are the two carve-outs Article 2(7) names. Article 4a is covered in the Article 10 section below. Article 59 concerns further processing of personal data in AI regulatory sandboxes.
Article 5 Prohibited Practices and Privacy
Article 5 lists practices banned outright. The Commission counts nine: eight that took effect in February 2025, and a ninth, added by the Omnibus, that the Act sets out in two points, (ba) and (bb). Several of the eight involve biometric data or profiling, which is why they matter for GDPR teams.
Social scoring (Art. 5(1)(c)): Bans AI systems that evaluate or classify people over a period of time based on social behaviour or known, inferred or predicted personal or personality characteristics, where the score leads to detrimental treatment in unrelated social contexts, or treatment that is unjustified or disproportionate to their behaviour or its gravity. The text is not limited to public authorities.
Criminal risk prediction (Art. 5(1)(d)): Bans assessing or predicting a person's risk of committing a criminal offence "based solely on the profiling of a natural person or on assessing their personality traits and characteristics." It does not apply to AI that supports a human assessment already based on objective and verifiable facts directly linked to a criminal activity.
Untargeted facial scraping (Art. 5(1)(e)): Bans AI systems that create or expand facial recognition databases through untargeted scraping of facial images from the internet or CCTV footage. Article 9 GDPR separately treats biometric data processed to uniquely identify a person as special-category data.
Emotion recognition at work and in education (Art. 5(1)(f)): Bans AI systems that infer emotions of a person in workplaces and education institutions, "except where the use of the AI system is intended to be put in place or into the market for medical or safety reasons." Outside those two settings this point does not apply, and Annex III lists emotion recognition as high-risk where its use is permitted by law.
Biometric categorisation (Art. 5(1)(g)): Bans systems that categorise individuals from biometric data to deduce or infer race, political opinions, trade union membership, religious or philosophical beliefs, sex life or sexual orientation. It does not cover labelling or filtering of lawfully acquired biometric datasets, or categorising biometric data in the area of law enforcement.
Real-time remote biometric identification (Art. 5(1)(h)): Bans use in publicly accessible spaces for law enforcement unless strictly necessary for one of three objectives: the targeted search for specific victims of abduction, trafficking or sexual exploitation and the search for missing persons; preventing a specific, substantial and imminent threat to life or physical safety, or a genuine and present or foreseeable terrorist threat; or locating or identifying a suspect of an Annex II offence punishable by a maximum of at least four years.
Each use needs prior authorisation from a judicial authority or an independent administrative authority whose decision is binding. In a duly justified urgency, use may start without it if authorisation is requested at the latest within 24 hours (Art. 5(3)). Each use must also be notified to the market surveillance authority and the national data protection authority (Art. 5(4)).
Manipulation and exploitation (Art. 5(1)(a) and (b)): Bans subliminal, purposefully manipulative or deceptive techniques, and the exploitation of vulnerabilities due to age, disability or a specific social or economic situation, where they materially distort behaviour in a way that causes or is reasonably likely to cause significant harm.
The Commission published non-binding guidelines on these prohibitions on 4 February 2025.
New From 2 December 2026: Article 5(1)(ba) and (bb)
These are scheduled, not yet applicable. Point (ba) covers AI systems that generate or manipulate realistic images, videos, audio or similar material of an identifiable person's intimate parts, or of an identifiable person engaged in sexually explicit activities, "without that person's freely-given, specific, informed, unambiguous and explicit consent." Point (bb) covers material or performance within Article 2, points (c) and (e), of Directive 2011/93/EU, "except where a 'without right' defence applies under national law."
Article 5(1a) narrows both. Placing such a system on the market or putting it into service is prohibited only where that generation is the system's intended purpose, or where its design makes the result a "reasonably foreseeable and reproducible outcome, without requiring significant technical modification" and it lacks reasonable and adequate safeguards. Use is prohibited only where the deployer uses the system for the purpose of generating such material.
Article 5(1b) adds that, for point (ba), manipulation that does not increase the exposure of depicted intimate parts or alter the nature of depicted sexually explicit activities is not "manipulation."
Article 10 Data Governance and Article 4a (High-Risk Rules From 2 December 2027)
Article 10 is the AI Act's most data-specific rule for high-risk systems. It is scheduled, not yet applicable. Providers' training, validation and testing data sets must be subject to data governance practices appropriate for the system's intended purpose.
What Article 10 Requires
Those practices concern, among other things, design choices, data collection and origin (including, for personal data, the original purpose of collection), preparation steps such as labelling and cleaning, examination for possible biases, measures to detect and mitigate them, and data gaps that prevent compliance (Art. 10(2)).
Data sets must be "relevant, sufficiently representative, and to the best extent possible, free of errors and complete in view of the intended purpose" (Art. 10(3)). That is a qualified standard.
The link to the GDPR is direct. Purpose limitation asks why personal data was collected, and Article 10(2)(b) makes the original purpose of collection part of the provider's data governance. Data minimisation limits personal data to what is necessary, while Article 10(3) asks for data sets that are sufficiently representative, so a provider handling personal data has both standards to meet.
Article 4a: Special-Category Data for Bias Correction
Article 4a is new, added by Regulation (EU) 2026/1744. It sits in Chapter I, which has applied since 2 February 2025, so it took effect when the amending regulation entered into force on 27 July 2026.
"To the extent strictly necessary" for bias detection and correction in high-risk systems, under Article 10(2)(f) and (g), providers "may exceptionally process special categories of personal data," subject to appropriate safeguards. This applies in addition to the GDPR, and all of the following conditions must be met (Art. 4a(1)):
- bias detection and correction cannot be effectively fulfilled by processing other data, including synthetic or anonymised data;
- the data is subject to technical limits on re-use and to state-of-the-art security and privacy-preserving measures, including pseudonymisation;
- the data is secured, with strict controls and documentation of access, so that only authorised persons with confidentiality obligations reach it;
- the data is not transmitted, transferred or otherwise accessed by other parties;
- the data is deleted once the bias is corrected or the retention period ends, whichever comes first;
- the records of processing activities state why the processing was strictly necessary and why other data could not achieve the objective.
Article 4a(2) opens the same possibility to providers and deployers of other AI systems and models and to deployers of high-risk systems. The processing must be strictly necessary for bias of the kinds it lists, all the same conditions apply, and the paragraph creates no duty to run such bias correction.
Which Systems Are High-Risk Under Annex III
Annex III lists these areas: biometrics (remote biometric identification, biometric categorisation and emotion recognition, where permitted by law, excluding verification that only confirms a person is who they claim to be); critical infrastructure safety components; education and vocational training; employment and workers' management; access to essential private and public services and benefits; law enforcement; migration, asylum and border control; and the administration of justice and democratic processes.
Under point 5, that includes creditworthiness evaluation and credit scoring (except detecting financial fraud), and risk assessment and pricing for life and health insurance. Under point 4, it includes recruitment and selection, including analysing and filtering job applications.
The Fundamental Rights Impact Assessment: Article 27 and the GDPR DPIA
Article 27 applies from 2 December 2027 under Article 113(c)(i), so the general deployer duty is scheduled, not yet in application. One link already exists: under Article 5(2), law enforcement use of real-time remote biometric identification may be authorised only if the authority has completed a fundamental rights impact assessment (FRIA) as provided for in Article 27.
Who Must Do a FRIA
The duty falls on deployers, not providers. It covers deployers of Article 6(2) high-risk systems that are bodies governed by public law or private entities providing public services, and deployers of the Annex III, point 5(b) and (c) systems (credit and life or health insurance). Systems in Annex III, point 2 (critical infrastructure) are excluded. Recital 96 names education, healthcare, social services, housing and the administration of justice as areas linked to public-interest tasks.
What It Must Cover
Article 27(1) requires a description of the deployer's processes, the period and frequency of use, the categories of people and groups likely to be affected, the specific risks of harm to them, human oversight measures, and the steps to be taken if those risks materialise, including governance and complaint arrangements.
It applies to the first use of the system and must be updated if an element changes (Art. 27(2)). The deployer then notifies the market surveillance authority using the AI Office's template, with a possible exemption in Article 46(1) cases (Art. 27(3) and (5)).
Deployers that are public authorities, EU bodies, or persons acting on their behalf must also register their use in the EU database before using an Annex III system, other than point 2 systems (Art. 49(3)). That entry includes a summary of the FRIA findings and, where applicable, of the data protection impact assessment (Annex VIII, Section C).
How a FRIA Differs From a GDPR DPIA
A data protection impact assessment (DPIA) under Article 35 GDPR is required where processing "is likely to result in a high risk to the rights and freedoms of natural persons," and the controller carries it out before the processing. A FRIA is an assessment of the impact on fundamental rights of using a specific high-risk AI system, and only the deployers listed above owe one.
A deployer may owe both. Article 27(4), as amended by the Omnibus, lets it include cross-references to the relevant sections of its DPIA, or relevant parts of the DPIA, where an Article 27 obligation is already met through the DPIA. Deployers also use the information providers supply under Article 13 to carry out their DPIA (Art. 26(9)).
Automated Decisions: GDPR Article 22 and AI Act Article 86
Article 22 GDPR gives a data subject the right not to be subject to a decision based solely on automated processing, including profiling, that produces legal effects or similarly significantly affects them, subject to the exceptions in Article 22(2).
Article 86 of the AI Act is a different right. A person subject to a deployer's decision taken on the basis of the output of an Annex III high-risk system (other than point 2), with legal or similarly significant effects they consider adverse to their health, safety or fundamental rights, may obtain "clear and meaningful explanations of the role of the AI system in the decision-making procedure and the main elements of the decision taken." Its text does not use the GDPR's word "solely."
Article 86 sits in Chapter IX, so Article 113 gives it no date of its own and the general date of 2 August 2026 applies. It covers high-risk systems listed in Annex III, and the rules that classify those systems apply from 2 December 2027. The text does not say how the right works before then.
Exceptions or restrictions may follow from Union or national law (Art. 86(2)), and the right applies only to the extent it is not otherwise provided for under Union law (Art. 86(3)).
Article 50 Transparency Duties and Personal Data
Since 2 August 2026, providers must design systems that interact directly with people so that those people are informed, unless that is obvious to a reasonably well-informed, observant and circumspect person (Art. 50(1)). Providers of systems generating synthetic audio, image, video or text must mark outputs in a machine-readable format, except for assistive standard editing or where the system does not substantially alter the input (Art. 50(2)).
For systems placed on the market before 2 August 2026, Article 111(4) says providers "shall take the necessary steps in order to comply with Article 50(2) by 2 December 2026." That transition covers Article 50(2) only.
Deployers of emotion recognition or biometric categorisation systems must inform the people exposed, and "shall process the personal data" in line with the GDPR and related instruments (Art. 50(3)). Deployers must disclose deepfakes and AI-generated text published to inform the public on matters of public interest, unless the text has undergone human review or editorial control and someone holds editorial responsibility (Art. 50(4)). Each paragraph carries its own law-enforcement limit, and Article 50(4) has a lighter rule for evidently artistic or satirical work.
The Commission published guidelines on these duties on 20 July 2026. The final voluntary Code of Practice on Transparency of AI-generated Content was published in June 2026, and the Commission concluded on 8 July 2026 that it adequately covers Article 50(2), (4) and (5). The Commission says adherence is voluntary, while the Article 50 requirements are legal obligations, and that adherence "does not constitute conclusive evidence of compliance" with them.
General-Purpose AI: Duties Since August 2025 and GDPR Training Data
General-purpose AI (GPAI) models can perform a wide range of tasks and are becoming the basis for many AI systems, in the Commission's description. Articles 53 and 55 set the duties. Providers of models placed on the market before 2 August 2025 must take the necessary steps to comply by 2 August 2027 (Art. 111(3)).
All GPAI Providers (Article 53)
Providers must keep technical documentation for the AI Office and national authorities on request (Annex XI), give downstream AI-system providers the information in Annex XII, put in place a policy to comply with EU copyright law, including rights reservations under Article 4(3) of Directive (EU) 2019/790, and publish a sufficiently detailed summary of training content on the AI Office's template.
The two documentation duties do not apply to models released under a free and open-source licence with public weights, architecture and usage information, unless the model has systemic risk (Art. 53(2)).
The Commission published guidelines on the scope of these duties on 18 July 2025. The General-Purpose AI Code of Practice was published on 10 July 2025, and the Commission and AI Board have confirmed it is an adequate voluntary tool.
Systemic-Risk Models (Articles 51 and 55)
A model is presumed to have high-impact capabilities when training computation exceeds 10^25 floating point operations (Art. 51(2)). The Commission can also designate a model under Annex XIII and can amend the thresholds by delegated act (Art. 51(1)(b) and (3)).
These providers must evaluate models, including adversarial testing, assess and mitigate systemic risks at Union level, track and report serious incidents, and ensure adequate cybersecurity (Art. 55(1)).
GDPR and AI Models: EDPB Opinion 28/2024
The European Data Protection Board adopted Opinion 28/2024 on 17 December 2024. It answers an Article 64(2) GDPR request from the Irish supervisory authority and gives general considerations for supervisory authorities.
On anonymity, the opinion says AI models trained with personal data "cannot, in all cases, be considered anonymous." Claims of anonymity are assessed case by case, and both the likelihood of extracting personal data from the model and the likelihood of obtaining it from queries should be insignificant.
On lawful basis, it recalls that there is no hierarchy between GDPR legal bases and that controllers identify the appropriate one. Relying on legitimate interest means a three-step test: identifying the interest, testing necessity, and checking that it is not overridden by data subjects' interests or rights. How unlawful development-stage processing affects later use of a model is, in the scenarios it describes, assessed case by case.
Who Enforces the AI Act and the GDPR Together
National Authorities and Data Protection Authorities
Each Member State must establish or designate at least one notifying authority and at least one market surveillance authority (Art. 70(1)). National data protection authorities keep their GDPR tasks and powers, as recital 10 confirms.
Article 74(8) brings data protection authorities directly into AI Act supervision. For Annex III, point 1 biometric systems used for law enforcement, border management and justice and democracy, and for points 6, 7 and 8 systems, Member States designate either the competent data protection supervisory authorities or another authority designated on the conditions of Articles 41 to 44 of Directive (EU) 2016/680.
National set-up is uneven. The Commission's list, last updated 7 September 2026, shows 9 Member States with an authority named, 12 marked as pending final adoption, and 6 blank (Austria, Belgium, Bulgaria, Croatia, Slovakia and Hungary). Member States were to designate authorities by 2 August 2025.
EU Level: AI Office and EDPS
The AI Office, within the Commission, holds enforcement powers over GPAI models. As amended by the Omnibus, Article 75(1) makes it exclusively competent for two groups of AI systems: those based on a GPAI model where model and system come from the same provider or undertaking, and those that are or are integrated into a designated very large online platform or search engine.
The first group has four exceptions: AI related to Annex I products, Annex III point 2 systems, certain systems of law enforcement, border and financial bodies under Article 74(6), and Annex III point 8 systems as regards the administration of justice. The exclusive competence applies to providers, and to deployers only when they are the provider or part of its undertaking.
The European Data Protection Supervisor acts as market surveillance authority for EU institutions, bodies, offices and agencies, except the Court of Justice acting in its judicial capacity (Art. 74(9)).
EU AI Act Fines Compared With GDPR Fines
Member States apply the three fine tiers in Article 99, which is part of Chapter XII and has applied since 2 August 2025. Fines on GPAI providers are separate: under Article 101, the Commission may impose up to 3% of annual total worldwide turnover or EUR 15,000,000, whichever is higher, where a provider intentionally or negligently commits a listed infringement.
| Tier | Maximum | Covers |
|---|---|---|
| AI Act Art. 99(3) | EUR 35,000,000 or 7% | Breach of Article 5 prohibitions |
| AI Act Art. 99(4) | EUR 15,000,000 or 3% | Listed operator duties, including Arts. 16, 22, 23, 24, 25(2) and (4), 26 and 50, and notified-body duties |
| AI Act Art. 99(5) | EUR 7,500,000 or 1% | Incorrect, incomplete or misleading information to notified bodies or authorities |
| AI Act Art. 101 | EUR 15,000,000 or 3% | GPAI providers, imposed by the Commission |
| GDPR Art. 83(5) | EUR 20,000,000 or 4% | Listed GDPR infringements, including basic processing principles |
Percentages are of total worldwide annual turnover for the preceding financial year where the offender is an undertaking, and in each row above the higher figure is the cap. These are ceilings, and Article 99(7) requires all relevant circumstances to be taken into account in each case.
For SMEs, including start-ups, each Article 99 fine is capped at the lower of the two figures (Art. 99(6)). The Omnibus extended that lower-of rule to small mid-cap enterprises, for paragraphs 4 and 5 only (Art. 99(6a)).
Conduct that breaches Article 5 can also involve processing that infringes the GDPR, and the two regimes have separate fine provisions.
What Changes in the EU AI Act in 2027
These dates are as the amended text reads on November 4, 2026.
Before the year starts, 2 December 2026: Article 5(1)(ba) and (bb) apply, and the Article 111(4) transition for legacy synthetic-content systems ends.
2 August 2027: Providers of GPAI models placed on the market before 2 August 2025 must have taken the necessary steps to comply (Art. 111(3)). Article 57(1) also requires Member States to ensure that their competent authorities establish at least one AI regulatory sandbox at national level, operational by that date. A joint sandbox, or participation in an existing one with equivalent national coverage, can also meet it.
2 December 2027: Chapter III, Sections 1 to 3 apply to Annex III high-risk systems (Art. 113(c)(i)). That brings Article 10 data governance, Article 26 deployer duties and Article 27 FRIAs into application.
Next, 2 August 2028: The same rules apply to Annex I systems (Art. 113(c)(ii)).
Pending as of November 4, 2026
The GDPR limb of the Digital Omnibus: Procedure 2025/0360(COD), COM(2025)837, is a separate proposal from Regulation (EU) 2026/1744. Parliament's summary says it "clarifies how 'legitimate interests' may apply" to processing personal data for developing and operating AI systems, subject to safeguards and the right to object, and adds a narrowly framed derogation for residual special-category data in datasets. The procedure file showed "Awaiting committee decision" when read on November 4, 2026. None of this is law.
Joint Commission and EDPB guidelines: The Commission's overview, last updated 31 July 2026, lists joint guidelines on the interplay of the AI Act and EU data protection law among those still being worked on, along with guidelines carrying a FRIA template. We did not find them on the EDPB's guidelines list or in a search of the EDPB and Commission sites on November 4, 2026; a later release would not show here.
Standards: The Commission announced on 19 October 2026 that the reference to the first harmonised standard, EN 18286:2026 on quality management systems, was published in the Official Journal through Implementing Decision (EU) 2026/2274. Providers implementing it benefit from a presumption of conformity with the Article 17 requirements it covers.
Practical Checklist for AI and GDPR Compliance
These are the checks the two laws point to. They are general information, not a compliance determination for any system.
- Map systems by risk tier. Identify anything that may fall within Article 5, including the points that apply from 2 December 2026.
- Record the GDPR basis separately. For each AI use involving personal data, record the lawful basis and any Article 9 condition. AI Act classification does not answer that question.
- Check Article 50 status. Confirm which systems interact with people, generate synthetic content, or use emotion recognition or biometric categorisation. Legacy synthetic-content systems face 2 December 2026.
- GPAI providers: check training summaries, copyright policies and technical documentation against Article 53, whether Article 55 applies, and whether a model falls under the 2 August 2027 date.
- High-risk providers and deployers: plan against 2 December 2027 (Annex III) and 2 August 2028 (Annex I), including Article 10 records and any reliance on Article 4a.
- FRIA deployers: public bodies, private providers of public services, and credit or life and health insurance deployers can plan the FRIA alongside the DPIA, since Article 27(4) allows cross-references.
- Test each regime independently. A system outside the high-risk list can still involve processing that requires a DPIA under Article 35 GDPR.
Change Log
- November 4, 2026: Rewrote the page against the amended text of the AI Act. Added the status line, a full date table, Articles 50 and 86, Article 4a's conditions, the exact limits on the new Article 5 points, Article 75's exceptions, the pending GDPR proposal, and the 2027 outlook. Corrected the Article 10(3) data standard, the AI literacy duty as rewritten in July 2026, and the treatment of machinery.
- October 27, 2026: Updated the timeline for transparency duties and Commission fines applying from 2 August 2026 and the two new Article 5 points applying from 2 December 2026, replaced Article 10(5) with Article 4a, and corrected FRIA scope, GPAI duties and fine tiers.
- September 9, 2026: Updated dates for Regulation (EU) 2026/1744: high-risk rules moved to 2 December 2027 for Annex III and 2 August 2028 for Annex I products, with the 2 August 2026 transparency date unchanged.
Related guides
- GDPR International Data Transfers: Chapter V Rules (2026)
- GDPR Right to Be Forgotten (Article 17) Explained
- Does GDPR Apply to US Companies? A Compliance Guide
- EU Data Privacy Laws: GDPR, AI Act & the 2025-2026 Digital Reforms
- What Is GDPR? Complete Guide to EU Data Protection (2026)
Frequently Asked Questions
Does the EU AI Act replace the GDPR?
No. Article 2(7) says that, without prejudice to Articles 4a and 59, the AI Act does not affect the GDPR, and recital 10 says it does not seek to affect existing Union law on personal data. AI systems that process personal data remain subject to the GDPR, and meeting the GDPR does not settle the AI Act.
When does the EU AI Act take effect?
It entered into force on 1 August 2024 and has applied since 2 August 2026, with exceptions. The original prohibitions applied from 2 February 2025 and general-purpose AI rules from 2 August 2025. The new Article 5 points apply from 2 December 2026. High-risk rules apply from 2 December 2027 for Annex III systems and 2 August 2028 for Annex I systems.
What is Regulation (EU) 2026/1744, the Digital Omnibus on AI?
It is the July 2026 regulation amending the AI Act. It was published on 24 July 2026 and entered into force on 27 July 2026. It set new high-risk dates, added two Article 5 points, added Article 4a on bias correction, rewrote Article 4 on AI literacy, extended some SME treatment to small mid-caps, and amended the AI Office's exclusive competence in Article 75.
Did the Digital Omnibus delay the EU AI Act high-risk rules?
Yes, the Commission describes the timelines as extended. Under amended Article 113(c), Chapter III, Sections 1 to 3 apply from 2 December 2027 for Annex III systems and 2 August 2028 for Annex I systems. As of November 4, 2026, those rules are scheduled and not yet in application.
Did the Digital Omnibus delay Article 50 transparency?
No general delay. Article 50 has applied since 2 August 2026. Article 111(4) gives providers of synthetic-content systems placed on the market before 2 August 2026 until 2 December 2026 to comply with Article 50(2) only.
Which Article 5 prohibited practices already apply?
Eight have applied since 2 February 2025: harmful manipulation; exploiting vulnerabilities; social scoring; criminal risk prediction based solely on profiling or personality traits; untargeted facial-image scraping; inferring emotions in workplaces and education institutions, except for medical or safety reasons; biometric categorisation to infer listed characteristics; and real-time remote biometric identification in publicly accessible spaces for law enforcement, unless strictly necessary for listed objectives. Points (ba) and (bb) apply from 2 December 2026.
What is the difference between a FRIA and a DPIA?
A DPIA under Article 35 GDPR concerns processing likely to result in a high risk to people's rights and freedoms, and the controller does it before the processing. A FRIA under Article 27 of the AI Act assesses the impact on fundamental rights of using a high-risk AI system, and only certain deployers owe one, from 2 December 2027. Article 27(4) lets a deployer cross-reference its DPIA where an Article 27 obligation is already met through it.
Can AI training use sensitive personal data under the AI Act?
Only in a narrow case. Article 4a lets providers of high-risk systems exceptionally process special-category data to the extent strictly necessary for bias detection and correction, on listed conditions, including that other data will not do, that the data is not accessed by other parties, and that it is deleted once the bias is corrected or the retention period ends. Article 4a(2) extends this, on the same conditions, to others. It is not a general permission for AI training, and the GDPR still applies.
What does Article 10 of the AI Act require for training data?
From 2 December 2027 for Annex III systems and 2 August 2028 for Annex I systems, providers of high-risk systems must subject training, validation and testing data sets to data governance practices, including data origin, bias examination and gap analysis. Data sets must be relevant, sufficiently representative and, to the best extent possible, free of errors and complete in view of the intended purpose.
How do EU AI Act fines compare with GDPR fines?
The AI Act's top tier is up to EUR 35,000,000 or 7% of total worldwide annual turnover, whichever is higher, for Article 5 breaches. The GDPR's top tier is up to EUR 20,000,000 or 4%, whichever is higher. The AI Act also has tiers of EUR 15,000,000 or 3% and EUR 7,500,000 or 1%. For SMEs, each of those Article 99 caps is the lower of the two figures.
Do general-purpose AI providers have to comply with the EU AI Act now?
Yes for models placed on the market on or after 2 August 2025. Providers of models placed on the market before that date must take the necessary steps to comply by 2 August 2027 under Article 111(3). Since 2 August 2026 the Commission can fine GPAI providers up to 3% of worldwide turnover or EUR 15,000,000, whichever is higher. GDPR duties for personal data in training apply alongside these rules.
Who enforces the AI Act for law enforcement and justice systems?
Under Article 74(8), Member States designate as market surveillance authority either the competent data protection supervisory authorities or another authority meeting the conditions of Articles 41 to 44 of Directive (EU) 2016/680, for the Annex III areas it lists. The European Data Protection Supervisor covers EU institutions, except the Court of Justice acting in its judicial capacity.
Has the EU changed the GDPR for AI training?
Not as of November 4, 2026. The Digital Omnibus proposal 2025/0360(COD) addresses legitimate interests for AI, but Parliament's procedure file showed 'Awaiting committee decision' on that date. It is a proposal, not law.
Updates
We rewrote this tracker against the amended AI Act, adding a full date table, Articles 50 and 86, the exact limits on the new Article 5 points and Article 4a, and the pending GDPR proposal, and correcting the Article 10(3) standard and the AI literacy duty.
Independently fact-checked against the cited primary sources; governing law re-checked for recent changes
We updated the AI Act timeline to October 2026, including the transparency duties and Commission fines for general-purpose AI providers that have applied since 2 August 2026 and the two new prohibitions that apply from 2 December 2026, replaced the deleted Article 10(5) with the new Article 4a, and corrected the scope, notification and registration of fundamental rights impact assessments, the general-purpose AI duties and fine tiers, the real-time biometric identification exceptions, the social scoring ban and the role of data protection authorities.
AI Act dates updated for the July 2026 Digital Omnibus (Regulation (EU) 2026/1744): high-risk obligations now apply 2 December 2027 for Annex III systems and 2 August 2028 for AI built into regulated products, while the 2 August 2026 transparency date is unchanged.
Sources and References
- AI Act Art. 2 (scope; relationship to the GDPR), as amended by Regulation (EU) 2026/1744(eur-lex.europa.eu).gov
- Regulation (EU) 2024/1689 (EU AI Act), Official Journal(eur-lex.europa.eu).gov
- Regulation (EU) 2026/1744 (Digital Omnibus on AI), Official Journal(eur-lex.europa.eu).gov
- European Parliament Legislative Observatory, procedure 2025/0360(COD), Digital Omnibus (read November 4, 2026)(oeil.europarl.europa.eu).gov
- European Commission: AI Act regulatory framework(digital-strategy.ec.europa.eu).gov
- AI Act Art. 113 (entry into force and application), as amended by Regulation (EU) 2026/1744(eur-lex.europa.eu).gov
- European Parliament Legislative Observatory, procedure 2025/0359(COD), Digital Omnibus on AI(oeil.europarl.europa.eu).gov
- European Commission: AI Omnibus enters into force (27 July 2026)(digital-strategy.ec.europa.eu).gov
- AI Act Art. 111 (systems already on the market), as amended by Regulation (EU) 2026/1744(eur-lex.europa.eu).gov
- AI Act Art. 4 (AI literacy), as amended by Regulation (EU) 2026/1744(eur-lex.europa.eu).gov
- AI Act Annex I (Union harmonisation legislation), as amended by Regulation (EU) 2026/1744(eur-lex.europa.eu).gov
- AI Act recital 10(eur-lex.europa.eu).gov
- Regulation (EU) 2016/679 (GDPR), text of 4 May 2016, stored copy retrieved 12 August 2026(eur-lex.europa.eu).gov
- AI Act Art. 5 (prohibited practices), as amended by Regulation (EU) 2026/1744(eur-lex.europa.eu).gov
- AI Act Annex III (high-risk areas)(eur-lex.europa.eu).gov
- European Commission: guidelines on prohibited AI practices (4 February 2025)(digital-strategy.ec.europa.eu).gov
- AI Act Art. 10 (data and data governance), as amended by Regulation (EU) 2026/1744(eur-lex.europa.eu).gov
- AI Act Art. 4a (special categories of personal data for bias detection and correction), added by Regulation (EU) 2026/1744(eur-lex.europa.eu).gov
- AI Act Art. 27 (fundamental rights impact assessment), as amended by Regulation (EU) 2026/1744(eur-lex.europa.eu).gov
- AI Act recital 96(eur-lex.europa.eu).gov
- AI Act Art. 49 (registration)(eur-lex.europa.eu).gov
- AI Act Annex VIII (registration information), as amended by Regulation (EU) 2026/1744(eur-lex.europa.eu).gov
- AI Act Art. 26 (deployer obligations)(eur-lex.europa.eu).gov
- AI Act Art. 86 (right to explanation)(eur-lex.europa.eu).gov
- AI Act Art. 50 (transparency), as amended by Regulation (EU) 2026/1744(eur-lex.europa.eu).gov
- European Commission: guidelines on transparency obligations (20 July 2026)(digital-strategy.ec.europa.eu).gov
- European Commission: Code of Practice on Transparency of AI-generated Content(digital-strategy.ec.europa.eu).gov
- European Commission: opinion on the assessment of the Code of Practice on Transparency of AI-generated Content (9 July 2026)(digital-strategy.ec.europa.eu).gov
- AI Act Art. 53 (general-purpose AI providers)(eur-lex.europa.eu).gov
- European Commission: guidelines on the scope of obligations for GPAI providers (18 July 2025)(digital-strategy.ec.europa.eu).gov
- European Commission: General-Purpose AI Code of Practice (published 10 July 2025)(digital-strategy.ec.europa.eu).gov
- AI Act Art. 51 (systemic-risk classification)(eur-lex.europa.eu).gov
- AI Act Art. 55 (systemic-risk provider duties)(eur-lex.europa.eu).gov
- EDPB Opinion 28/2024 on AI models (adopted 17 December 2024)(www.edpb.europa.eu).gov
- EDPB: Opinion 28/2024 document page(www.edpb.europa.eu).gov
- AI Act Art. 70 (national competent authorities), as amended by Regulation (EU) 2026/1744(eur-lex.europa.eu).gov
- AI Act Art. 74 (market surveillance)(eur-lex.europa.eu).gov
- European Commission: market surveillance authorities under the AI Act (updated 7 September 2026)(digital-strategy.ec.europa.eu).gov
- European Commission: AI Act governance and enforcement(digital-strategy.ec.europa.eu).gov
- AI Act Art. 75 (AI Office competence), as amended by Regulation (EU) 2026/1744(eur-lex.europa.eu).gov
- AI Act Art. 99 (penalties), as amended by Regulation (EU) 2026/1744(eur-lex.europa.eu).gov
- AI Act Art. 101 (fines for GPAI providers)(eur-lex.europa.eu).gov
- European Parliament Legislative Observatory: summary of COM(2025)837(oeil.europarl.europa.eu).gov
- European Commission: overview of AI Act guidelines (last updated 31 July 2026)(digital-strategy.ec.europa.eu).gov
- European Commission: first harmonised standard supporting the AI Act (19 October 2026)(digital-strategy.ec.europa.eu).gov
- EDPB: minutes of the 100th plenary meeting, 17 December 2024(www.edpb.europa.eu).gov
- EDPB: guidelines, recommendations and best practices list (read November 4, 2026)(www.edpb.europa.eu).gov
- AI Act Art. 57 (AI regulatory sandboxes), as amended by Regulation (EU) 2026/1744(eur-lex.europa.eu).gov