The EU AI Act sets clear duties for AI providers based on risk, from banned practices to obligations for high-risk systems and general-purpose AI later on. Published 12 July 2024 and in force from 1 August 2024, prohibitions begin 2 February 2025, followed by general-purpose rules on 2 August 2025. Throughout the page, you’ll see how readiness gaps, compliance cost expectations, and key requirements like risk management, data governance, documentation, and CE marking connect to enforcement and penalties.
Key Takeaways
- 40% of EU companies unaware of AI Act per PwC survey.
- 65% of firms expect compliance costs increase per Deloitte.
- Only 10% of SMEs feel prepared per EY poll.
- High-risk AI providers must ensure risk management system.
- Data governance for high-risk AI requires quality datasets.
- Technical documentation must be kept for 10 years.
- Fines up to €35 million or 7% global annual turnover for prohibited AI.
- Fines up to €15 million or 3% turnover for other violations.
- Fines up to €7.5 million or 1.5% for supplying incorrect info.
- AI Act influences 20+ global regulations.
- China referenced EU AI Act in its rules.
- US states passed 50+ AI bills inspired by EU Act.
- The EU AI Act was published in the Official Journal of the EU on 12 July 2024.
- The EU AI Act entered into force on 1 August 2024.
- Prohibitions under the AI Act apply from 2 February 2025.
Most EU firms are unprepared for the EU AI Act’s rising compliance costs, strict timelines, and major penalties.
Related reading
01 · Category
Business Impact17 stats
01
40% of EU companies unaware of AI Act per PwC survey.
02
65% of firms expect compliance costs increase per Deloitte.
03
Only 10% of SMEs feel prepared per EY poll.
04
AI Act could add €31B annual compliance costs per Arthur D Little.
05
76% of executives concerned about fines per KPMG.
06
55% plan to increase AI governance budgets.
07
82% of non-EU firms see extraterritorial impact.
08
Projected 25% slowdown in high-risk AI deployment.
09
90% of chatbots will need transparency labels.
10
€10B EU funding for AI innovation 2021-2027.
11
37% of EU businesses use AI per Eurostat 2023.
12
68% of large enterprises use AI vs 8% SMEs.
13
45% of firms report readiness gap per Boston Consulting.
14
92% of EU citizens support AI rules per Eurobarometer.
15
AI Act aligns with GDPR for data protection.
16
Expected 15% growth in AI compliance jobs.
17
70% of US tech firms plan EU-specific compliance teams.
Interpretation
Business Impact Interpretation
For the Business Impact angle, the message is clear: while 76% of executives are worried about fines, major costs are already anticipated, with compliance expenses rising for 65% of firms and potentially totaling €31B annually, leaving only 10% of SMEs feeling prepared.
02 · Category
Compliance Obligations15 stats
01
High-risk AI providers must ensure risk management system.
02
Data governance for high-risk AI requires quality datasets.
03
Technical documentation must be kept for 10 years.
04
CE marking required for high-risk AI on market.
05
Transparency for GPAI: technical docs and summaries public.
06
User instructions must disclose AI interaction for limited risk.
07
Register of high-risk AI systems to be public.
08
Conformity assessment before market placement for high-risk.
09
Incident reporting within 15 days for high-risk AI.
10
Human oversight required to minimize risks.
11
Cybersecurity standards mandatory for high-risk systems.
12
GPAI models training compute >10^25 FLOPs are systemic.
13
Model evaluation, testing, monitoring for systemic GPAI.
14
Codes of Practice to be developed within 9 months.
15
Accuracy, robustness, cybersecurity for GPAI obligations.
Interpretation
Compliance Obligations Interpretation
For compliance obligations, the EU AI Act places long term accountability on providers and users, including 10 year technical documentation retention for high risk AI and ongoing market readiness steps like CE marking, with transparency requirements for GPAI and clear user disclosures even for limited risk interactions.
03 · Category
Enforcement Penalties14 stats
01
Fines up to €35 million or 7% global annual turnover for prohibited AI.
02
Fines up to €15 million or 3% turnover for other violations.
03
Fines up to €7.5 million or 1.5% for supplying incorrect info.
04
European AI Office established for enforcement.
05
National authorities handle market surveillance.
06
AI Board coordinates at EU level with 1 member per state.
07
Database for prohibited AI practices managed by Commission.
08
Market surveillance max harmonized under AI Act.
09
Appeals process for classification decisions.
10
Corrective measures include withdrawal from market.
11
72-hour notice for law enforcement biometric use.
12
Annual reports on enforcement by Member States.
13
Scientific Panel of independent experts for advice.
14
Advisory Forum with stakeholders for AI Office.
Interpretation
Enforcement Penalties Interpretation
For the Enforcement Penalties under the EU AI Act, the top enforcement risk can reach €35 million or 7% of global annual turnover for prohibited AI, showing that the regime uses sharply escalating fines backed by an EU level enforcement structure.
More related reading
04 · Category
Global Influence17 stats
01
AI Act influences 20+ global regulations.
02
China referenced EU AI Act in its rules.
03
US states passed 50+ AI bills inspired by EU Act.
04
Brazil's AI bill mirrors risk-based approach.
05
Singapore updated AI governance using EU model.
06
60% of G20 countries adopting similar frameworks.
07
UK's AI Safety Summit referenced EU Act.
08
Canada's AIDA delayed to align with EU.
09
Japan amended AI guidelines post-EU Act.
10
South Korea's AI Act effective 2026 like EU.
11
Australia consulting on EU-style risk framework.
12
85% global AI market affected by EU rules.
13
Non-EU firms 50% of GPAI notifications expected.
14
UN AI resolution mentions EU Act as model.
15
12 international standards bodies harmonizing with AI Act.
16
30% increase in global AI ethics searches post-Act.
17
75% of multinationals cite AI Act in ESG reports.
Interpretation
Global Influence Interpretation
Under the Global Influence angle, the EU AI Act has become a blueprint that helped shape 20 plus international regulations and drove 60% of G20 countries toward similar risk-based frameworks, with China, the US, Brazil, and Singapore all reflecting its impact in their own AI rules.
05 · Category
Legislative Timeline10 stats
01
The EU AI Act was published in the Official Journal of the EU on 12 July 2024.
02
The EU AI Act entered into force on 1 August 2024.
03
Prohibitions under the AI Act apply from 2 February 2025.
04
General-purpose AI rules apply from 2 August 2025.
05
High-risk AI systems obligations apply from 2 August 2027.
06
The AI Act contains 113 articles.
07
The regulation includes 151 recitals.
08
The AI Act was provisionally agreed on 8 December 2023.
09
Final adoption by European Parliament on 13 March 2024.
10
Council formal adoption on 21 May 2024.
Interpretation
Legislative Timeline Interpretation
From a legislative timeline standpoint, the EU AI Act moves through staggered milestones over more than three years, starting with publication on 12 July 2024 and entering into force on 1 August 2024, with major obligations ramping up from 2 February 2025 to 2 August 2025 and high risk requirements by 2 August 2027.
06 · Category
Risk Classifications14 stats
01
The AI Act defines 5 prohibited AI practices.
02
Unacceptable risk AI systems are banned entirely.
03
High-risk AI systems are listed in Annex I with 8 areas.
04
Annex III lists 34 product groups under high-risk.
05
Limited risk AI requires transparency obligations.
06
Minimal risk AI covers 99% of current AI uses with no obligations.
07
General-purpose AI models with systemic risk have extra rules.
08
Remote biometric identification in public spaces is prohibited except exceptions.
09
AI systems manipulating human behavior are unacceptable risk.
10
High-risk AI in biometrics has specific conformity assessment.
11
15% of AI systems expected to be high-risk per EC estimates.
12
Emotion recognition AI in workplaces is high-risk.
13
AI for critical infrastructure management is high-risk.
14
High-risk AI in education and vocational training covered.
Interpretation
Risk Classifications Interpretation
From the Risk Classifications perspective, the AI Act starts by banning 5 prohibited practices and then categorizes the rest into tightly defined tiers, with high-risk systems mapped to 8 areas across 34 product groups while limited risk adds transparency duties and minimal risk covers about 99% of today’s AI uses with no obligations.
Reference
Cite This Report
This report is designed to be cited. We maintain stable URLs and versioned verification dates. Copy the format appropriate for your publication below.
APA
Sophie Moreland. (2026, February 24). EU AI Act Statistics. Gitnux. https://gitnux.org/eu-ai-act-statistics
MLA
Sophie Moreland. "EU AI Act Statistics." Gitnux, 24 Feb 2026, https://gitnux.org/eu-ai-act-statistics.
Chicago
Sophie Moreland. 2026. "EU AI Act Statistics." Gitnux. https://gitnux.org/eu-ai-act-statistics.
Sources & references
35 datasets cited across this report · attribution is report-level