TOPICS·ARTIFICIAL INTELLIGENCE

EU AI Act

Regulation (EU) 2024/1689

The world's first comprehensive legal framework for artificial intelligence. Classifies AI systems by risk level and assigns obligations proportional to potential harm.

EUIN FORCE14 regulations trackedUpdated 21 April 2026
WHAT IS THIS ABOUT?

If your company builds software that makes predictions, recommendations, or decisions -- and that software touches anyone in Europe -- you should know about the EU AI Act. It is the first law in the world that regulates artificial intelligence across an entire economy. It was passed in 2024, it is already partially in effect, and its biggest deadlines land in 2026 and 2027.

The basic idea is proportional regulation. Not all AI is treated the same. A spam filter or a video game recommendation engine faces almost no rules. But an AI system that helps decide who gets a mortgage, who gets hired, or who gets flagged at a border crossing is considered high-risk and must meet strict requirements: documentation, testing, human oversight, and public registration. A small number of AI uses -- like government social scoring or covert manipulation of vulnerable people -- are banned entirely.

The law also covers large foundation models (think GPT, Gemini, or Claude) separately. Companies that train these models must publish summaries of their training data, comply with copyright rules, and -- if the model is powerful enough -- run safety evaluations and report serious incidents. These rules already apply as of August 2025.

Enforcement is serious. Fines can reach EUR 35 million or 7% of a company's worldwide annual revenue, whichever is higher. That said, the EU has built in protections for startups and small businesses: for SMEs, penalties are calculated using the lower of the two figures rather than the higher. The law is not designed to stop innovation. It is designed to make sure AI that can cause real harm to people is built and used responsibly.

THE ESSENTIALS

The EU AI Act is the first law anywhere in the world that regulates artificial intelligence across an entire economy. It applies to any company that builds, sells, or uses AI systems in the European Union -- regardless of where that company is based. If your AI touches EU residents, this law likely applies to you.

The core idea is simple: the riskier the AI, the stricter the rules. A spam filter has almost no obligations. An AI system that decides who gets a loan or who gets hired faces heavy requirements -- documentation, testing, human oversight, and registration in a public EU database. A handful of AI practices, like mass surveillance scoring or manipulating vulnerable people, are banned outright.

The law does not apply all at once. It phases in over three years, starting with bans on the most dangerous practices (already in effect) and ending with rules for AI embedded in regulated products like medical devices and cars (August 2027). Most companies building or deploying AI need to be ready by August 2026.

Penalties are severe: up to EUR 35 million or 7% of global annual turnover for the worst violations. Even providing incorrect information to regulators can cost up to EUR 7.5 million or 1% of turnover.

WHAT
The world's first comprehensive legal framework for artificial intelligence, classifying AI systems by risk level and assigning obligations accordingly.
WHO
Providers, deployers, importers, and distributors of AI systems placed on the EU market or whose outputs affect EU residents.
WHEN
Phased rollout. Prohibitions from Feb 2025, GPAI rules from Aug 2025, full high-risk requirements from Aug 2026.
PENALTY
Three tiers: up to EUR 35 million or 7% of global turnover for prohibited practices; EUR 15 million or 3% for high-risk and GPAI violations; EUR 7.5 million or 1% for supplying incorrect information to authorities.

The AI Act's obligations phase in over three years. Each phase adds requirements for a new category of AI systems.

AUG 2024
COMPLETED
Entry into force
The AI Act enters into force. AI literacy obligations begin. The 24-month countdown starts.
FEB 2025
COMPLETED
Prohibited AI practices banned
Social scoring, manipulative AI exploiting vulnerabilities, untargeted facial recognition scraping, emotion recognition in workplaces and schools, and certain biometric categorisation systems become illegal.
AUG 2025
COMPLETED
GPAI model rules apply
Providers of general-purpose AI models must provide technical documentation, comply with copyright law, and publish training data summaries. Models with systemic risk face additional evaluation and testing obligations.
AUG 2026
NEXT DEADLINE
High-risk AI requirements
Full obligations for Annex III high-risk AI systems: risk management, conformity assessment, CE marking, EU database registration, post-market monitoring. Transparency rules (Article 50) also apply. National enforcement begins.
AUG 2027
UPCOMING
Product-embedded AI
High-risk AI systems embedded in products already regulated under EU product legislation (Annex I) must comply. Legacy GPAI models placed on market before Aug 2025 must also comply by this date.
DIGITAL OMNIBUS PROPOSAL
On 19 November 2025, the European Commission proposed linking the high-risk AI deadline to the availability of harmonised standards. The Council adopted its negotiating mandate on 13 March 2026 and the Parliament followed on 26 March 2026 (569 votes in favour). Both institutions favour fixed fallback dates: 2 December 2027 for Annex III high-risk systems, and 2 August 2028 for product-embedded AI (Annex I). Trilogue negotiations are underway, with a political agreement targeted for late April 2026. Until formally adopted, organisations should continue to treat August 2026 as the binding deadline.

The AI Act's defining feature is its risk-based approach. Every AI system falls into one of four tiers. Click each tier to see obligations and examples.

Annex III defines eight categories of AI systems that are automatically classified as high-risk. If your system falls into any category below, it is subject to the full set of high-risk obligations.

EXCEPTION
An Annex III system is not high-risk if it does not pose a significant risk of harm -- for example, if it performs only a narrow procedural task, improves the result of a previously completed human activity, detects decision-making patterns without replacing human assessment, or performs a preparatory task. Providers relying on this exception must document their reasoning before placing the system on the market.

The AI Act creates distinct obligations based on your role in the AI value chain. The most common confusion: which role applies to your organisation?

WHEN A DEPLOYER BECOMES A PROVIDER
If you modify a high-risk AI system substantially, put your own name/trademark on it, change its intended purpose, or make a substantial modification to a high-risk system -- you become the provider and inherit the full provider obligations. This includes fine-tuning a foundation model for a high-risk use case.

The AI Act's newest and most debated provisions target GPAI models -- foundation models like large language models that can be adapted for many tasks. Rules took effect August 2025.

ALL GPAI MODELS
Maintain and provide technical documentation
Provide information to downstream AI system providers
Comply with EU copyright law
Publish a sufficiently detailed summary of training data
SYSTEMIC RISK MODELS
Above 1025 FLOPs training compute (rebuttable presumption), or designated by Commission
All obligations above, plus:
Perform model evaluations including adversarial testing
Assess and mitigate systemic risks
Track, document and report serious incidents
Ensure adequate cybersecurity protections
Report energy consumption of the model
GPAI CODE OF PRACTICE
The GPAI Code of Practice was published on 10 July 2025 and endorsed by the AI Board via adequacy decisions on 1 August 2025. It covers three chapters: transparency, copyright, and safety/security. While voluntary, adhering to the code is the primary way to demonstrate compliance. The AI Office can begin enforcement actions from August 2026 -- penalties of up to EUR 15 million or 3% of global annual turnover.

Select your company type for tailored compliance guidance.

KEY OBLIGATIONS
Classify all AI systems by risk level (prohibited, high-risk, limited, minimal)
Implement risk management systems for high-risk AI applications
Maintain technical documentation and logging for high-risk systems
Register high-risk AI systems in the EU database before deployment
Comply with transparency obligations for AI-generated content
YOUR FIRST STEP

Create an AI inventory cataloguing every AI system you develop or deploy, with initial risk classification per the Act's Annex III

01
Risk classification
Classify every AI system you develop or deploy into one of four risk tiers: unacceptable, high, limited, or minimal risk.
02
Prohibited practices
Immediately cease any AI practices banned outright, including social scoring, manipulative techniques, and certain biometric surveillance.
03
High-risk conformity
For high-risk AI, implement risk management systems, data governance, technical documentation, and human oversight mechanisms.
04
Transparency obligations
Ensure users know they are interacting with AI; label AI-generated content (deepfakes, synthetic media) clearly.
05
GPAI model compliance
Providers of general-purpose AI models must provide technical documentation, comply with copyright law, and publish training data summaries.
06
Fundamental rights assessment
Deployers of high-risk AI in public and private sectors must assess impacts on fundamental rights before deployment.
07
Post-market monitoring
Maintain continuous monitoring of high-risk AI systems in operation and report serious incidents to authorities.
08
Quality management system
Establish documented quality management procedures covering the entire AI system lifecycle from design to retirement.
AI REGULATIONS14
EU14
US0
COURT RULINGS0
VIEW ALL →
JUR.TITLESTATUSLINKS
EURegulation (EU) 2024/1689 of the European Parliament and of the Council of 13 June 2024 laying down harmonised rules on artificial intelligence and amending Regulations (EC) No 300/2008, (EU) No 167/2013, (EU) No 168/2013, (EU) 2018/858, (EU) 2018/1139 and (EU) 2019/2144 and Directives 2014/90/EU, (EU) 2016/797 and (EU) 2020/1828 (Artificial Intelligence Act) (Text with EEA relevance)Adopted4
EURegulation (EU) 2024/2748 of the European Parliament and of the Council of 9 October 2024 amending Regulations (EU) No 305/2011, (EU) 2016/424, (EU) 2016/425, (EU) 2016/426, (EU) 2023/988 and (EU) 2023/1230 as regards emergency procedures for the conformity assessment, presumption of conformity, adoption of common specifications and market surveillance due to an internal market emergency (Text with EEA relevance)Adopted3
EUCouncil Decision (CFSP) 2025/2539 of 11 December 2025 amending Decision (CFSP) 2022/2269 on Union support for the implementation of a project ‘Promoting Responsible Innovation in Artificial Intelligence for Peace and Security’Adopted1
EUCorrigendum to Regulation (EU) 2024/2748 of the European Parliament and of the Council of 9 October 2024 amending Regulations (EU) No 305/2011, (EU) 2016/424, (EU) 2016/425, (EU) 2016/426, (EU) 2023/988 and (EU) 2023/1230 as regards emergency procedures for the conformity assessment, presumption of conformity, adoption of common specifications and market surveillance due to an internal market emergency (OJ L, 2024/2748, 8.11.2024)Adopted1
EUCommission Implementing Regulation (EU) 2025/454 of 7 March 2025 laying down the rules for the application of Regulation (EU) 2024/1689 of the European Parliament and of the Council as regards the establishment of a scientific panel of independent experts in the field of artificial intelligenceAdopted1
EUCouncil Regulation (EU) 2024/1732 of 17 June 2024 amending Regulation (EU) 2021/1173 as regards a EuroHPC initiative for start-ups in order to boost European leadership in trustworthy artificial intelligenceAdopted1
EUCouncil Decision (CFSP) 2022/2269 of 18 November 2022 on Union support for the implementation of a project ‘Promoting Responsible Innovation in Artificial Intelligence for Peace and Security’Adopted1
EUCouncil Decision (EU) 2025/2350 of 13 November 2025 on the position to be taken on behalf of the European Union within the Committee of Ministers of the Council of Europe on the Draft Recommendation on equality and artificial intelligenceAdopted0
EUCouncil Decision (EU) 2024/2218 of 28 August 2024 on the signing, on behalf of the European Union, of the Council of Europe Framework Convention on Artificial Intelligence and Human Rights, Democracy and the Rule of LawAdopted0
EUCommission Decision of 24 January 2024 establishing the European Artificial Intelligence OfficeAdopted0
VIEW ALL →
DATEJUR.TITLESTATUS
Dec 11, 2025EUCouncil Decision (CFSP) 2025/2539 of 11 December 2025 amending Decision (CFSP) 2022/2269 on Union support for the implementation of a project ‘Promoting Responsible Innovation in Artificial Intelligence for Peace and Security’Adopted
Nov 13, 2025EUCouncil Decision (EU) 2025/2350 of 13 November 2025 on the position to be taken on behalf of the European Union within the Committee of Ministers of the Council of Europe on the Draft Recommendation on equality and artificial intelligenceAdopted
Apr 22, 2025EUCorrigendum to Regulation (EU) 2024/2748 of the European Parliament and of the Council of 9 October 2024 amending Regulations (EU) No 305/2011, (EU) 2016/424, (EU) 2016/425, (EU) 2016/426, (EU) 2023/988 and (EU) 2023/1230 as regards emergency procedures for the conformity assessment, presumption of conformity, adoption of common specifications and market surveillance due to an internal market emergency (OJ L, 2024/2748, 8.11.2024)Adopted
Mar 7, 2025EUCommission Implementing Regulation (EU) 2025/454 of 7 March 2025 laying down the rules for the application of Regulation (EU) 2024/1689 of the European Parliament and of the Council as regards the establishment of a scientific panel of independent experts in the field of artificial intelligenceAdopted
Oct 9, 2024EURegulation (EU) 2024/2748 of the European Parliament and of the Council of 9 October 2024 amending Regulations (EU) No 305/2011, (EU) 2016/424, (EU) 2016/425, (EU) 2016/426, (EU) 2023/988 and (EU) 2023/1230 as regards emergency procedures for the conformity assessment, presumption of conformity, adoption of common specifications and market surveillance due to an internal market emergency (Text with EEA relevance)Adopted
Aug 28, 2024EUCouncil Decision (EU) 2024/2218 of 28 August 2024 on the signing, on behalf of the European Union, of the Council of Europe Framework Convention on Artificial Intelligence and Human Rights, Democracy and the Rule of LawAdopted
Jun 17, 2024EUCouncil Regulation (EU) 2024/1732 of 17 June 2024 amending Regulation (EU) 2021/1173 as regards a EuroHPC initiative for start-ups in order to boost European leadership in trustworthy artificial intelligenceAdopted
Jun 13, 2024EURegulation (EU) 2024/1689 of the European Parliament and of the Council of 13 June 2024 laying down harmonised rules on artificial intelligence and amending Regulations (EC) No 300/2008, (EU) No 167/2013, (EU) No 168/2013, (EU) 2018/858, (EU) 2018/1139 and (EU) 2019/2144 and Directives 2014/90/EU, (EU) 2016/797 and (EU) 2020/1828 (Artificial Intelligence Act) (Text with EEA relevance)Adopted
Jan 24, 2024EUCommission Decision of 24 January 2024 establishing the European Artificial Intelligence OfficeAdopted
Nov 21, 2022EUCouncil Decision (EU) 2022/2349 of 21 November 2022 authorising the opening of negotiations on behalf of the European Union for a Council of Europe convention on artificial intelligence, human rights, democracy and the rule of lawAdopted
Nov 18, 2022EUCouncil Decision (CFSP) 2022/2269 of 18 November 2022 on Union support for the implementation of a project ‘Promoting Responsible Innovation in Artificial Intelligence for Peace and Security’Adopted
Jun 18, 2020EUEuropean Parliament decision of 18 June 2020 on setting up a special committee on artificial intelligence in a digital age, and defining its responsibilities, numerical strength and term of office (2020/2684(RSO))Adopted
Oct 24, 2019EUCommission Directive 2019/1832 of 24 October 2019 amending Annexes I, II and III to Council Directive 89/656/EEC as regards purely technical adjustmentsAdopted
Apr 24, 2019EUCorrigendum to Council Directive 90/270/EEC of 29 May 1990 on the minimum safety and health requirements for work with display screen equipment (fifth individual Directive within the meaning of Article 16 (1) of Directive 87/391/EEC) (OJ L 156, 21.6.1990)Adopted