The EU Artificial Intelligence Act (AI Act), the world’s first comprehensive law governing AI, has been officially enacted! This landmark legislation is poised to shape the future of AI in Europe and beyond, ensuring safety, transparency, and accountability.
Who Does the AI Act Apply To?
- Everyone who wants to develop, distribute, and import AI systems in the EU.
- AI systems that affect users located in the EU territory.
Determining Risk Levels and Obligations
The new rules establish obligations for providers and users depending on the level of risk associated with the AI system. Providers must determine the risk level of their AI systems according to the categories set out in the AI Act:
- Prohibited AI Systems
- High-Risk AI Systems
- General Purpose AI Systems (GPAI)
- Limited Risk AI Systems
Prohibited AI Systems
AI systems posing unacceptable risks to people’s rights are banned under the AI Act. These include:
- Real-time remote biometric identification using sensitive characteristics
- Indiscriminate scraping of facial images for facial recognition databases
- Emotion recognition
- Social scoring by governments
- Predictive policing
High-Risk AI Systems
High-risk AI systems are those deemed to pose significant risks to human fundamental rights, safety, health, environment, democracy, and the rule of law. These systems are used in critical sectors such as:
- Energy and transportation
- Education
- Healthcare
- Banking and insurance
- Law enforcement
- Migration
- Justice and democratic processes
Requirements for High-Risk AI Systems:
- Assess and mitigate risks
- Maintain transparency and accuracy
- Incorporate human oversight
- Enable consumers to submit complaints regarding improper AI use and receive explanations for AI decisions affecting their rights
General Purpose AI Systems (GPAI)
General Purpose AI Systems (GPAI) have specific transparency requirements, including:
- Compliance with EU copyright law
- Technical documentation for training, testing, and evaluation
- Integration information for downstream systems
Additional requirements for powerful GPAI models:
- Clearly label artificial or manipulated content (e.g., “deepfakes”)
- Conduct model evaluations
- Assess and mitigate high risks
- Implement cybersecurity measures
Note: Free and open license GPAI models are exempt from most requirements, except for EU copyright compliance and providing training data summaries.
Limited Risk AI Systems
Limited risk AI systems must comply with transparency requirements to enable users to make informed decisions. The AI Act ensures users know when they are interacting with AI or AI-generated content.
Financial Penalties
Non-compliance with the AI Act can result in substantial financial penalties:
- High-risk systems or prohibited practices: up to €35 million or 7% of global turnover
- Other non-compliance cases: up to €15 million or 3% of global turnover
- Providing incorrect/misleading information: up to €7.5 million or 1% of global turnover
Legal Support
Legal professionals can assist with:
- Preparing conformity assessments
- Creating legal documentation packages for full AI Act compliance
- Assisting with implementing voluntary codes of practice for lower-risk AI systems
Key Dates to Remember:
- 12th July 2024: Publication in the OJEU
- 1st August 2024: Act enters into force
- 2nd February 2025: Bans on prohibited practices begin
- 2nd May 2025: Preparation of codes of practice for GPAI providers
- 2nd August 2025: Compliance deadline for new GPAI
- 2nd August 2026: Compliance deadline for high-risk AI systems
- 2nd August 2027: Compliance for GPAIM and high-risk AI systems on the market before 1st August 2025
- 2nd August 2028: Compliance for AI systems in large-scale information systems and high-risk AI systems used by public authorities
Conclusion
This landmark legislation is set to ensure the safe, transparent, and accountable development and use of AI in Europe and beyond. As the AI landscape evolves, the AI Act will play a crucial role in safeguarding fundamental rights while fostering innovation.
Is your organization ready to comply with the new regulations? Preparing now will ensure a smooth transition and help you stay ahead in the rapidly evolving AI ecosystem.
Article written by Milanova & Partners.