KEY HIGHLIGHTS
- The EU AI Act classifies AI systems based on risk levels.
- AI developers must provide necessary documentation to improve transparency.
- Provisions imposing ban on prohibiting AI systems to come in force in next six months.
INTRODUCTION
The European Union (EU) has taken a significant step in shaping the future of artificial intelligence (AI) by introducing the EU AI Act on 13th March 2024. The regulation, agreed in consultations with member states in December 2023, was endorsed with 523 votes in favour, 46 against and 49 abstentions. During the plenary debate on Tuesday, the Internal Market Committee co-rapporteur- Brando Benifei (S&D, Italy) said: “We finally have the world’s first binding law on artificial intelligence, to reduce risks, create opportunities, combat discrimination, and bring transparency. He acknowledged the commitment of EU, highlighting the significant features of the act, highlighting the provisions prohibiting unacceptable AI practices within Europe. Further, he also mentioned about “The AI Office” to be set up in the future to assist organisations, in fulfilling the obligations of the Act before its provisions come into force. Civil Liberties Committee co-rapporteur Dragos Tudorache (Renew, Romania) said: “The EU has delivered. We have linked the concept of artificial intelligence to the fundamental values that form the basis of our societies”. However, he emphasised that much work need to be done to implement the Act effectively. This comprehensive legislation aims to strike a balance between fostering innovation and safeguarding fundamental rights. In this blog, we delve into the intricacies of the AI Act, explore its implications, and provide actionable insights for organizations preparing for compliance.
RISK-BASED CLASSIFICATION
The foundation of the AI Act lies in its risk-based classification of AI System. AI systems are categorized based on their potential impact on health, safety, and fundamental rights. The Act classified AI systems into following categories-
Prohibited Systems: These AI systems pose an unacceptable risk and are outright prohibited. Examples include social scoring systems and manipulative AI.
High-Risk Systems: These systems require comprehensive compliance measures before being placed on the market. They cover critical areas such as healthcare, transportation, and energy.
Limited Risk AI Systems: These systems fall into a smaller category and are subject to lighter transparency obligations. Examples include chatbots and deepfakes.
Minimal Risk AI Systems: Minimal risk AI systems are unregulated. They include the majority of AI applications currently available on the EU single market.
KEY PROVISIONS UNDER THE ACT
Transparency Requirements: The EU AI Act imposes transparency requirements on specific AI systems.For instance, when there is a clear risk of manipulation (such as with chatbots), users must be informed that they are interacting with an AI system.
Conformity Assessments (CAs): Prior to placing an AI system on the EU market, CAs must be conducted.Additionally, if a high-risk AI system undergoes substantial modifications, a new CA is necessary. Importers of AI systems must verify that the foreign provider has already completed the appropriate CA procedure.
Fundamental Rights Impact Assessments (FRIA):
Prior to deploying high-risk AI systems, deplorers and public bodies must assess the potential impact on fundamental rights. If a data protection impact assessment (DPIA) is necessary, the FRIA should be conducted alongside it.
Distinct Provisions for Generative AI Systems:
Providers of AI Systems capable of generating synthetic audio, image, video, or text content must ensure that the generated content is marked and labelled properly. Act emphasises that detectability as artificially generated or manipulated content is crucial for transparency and accountability.
IMPLEMENTATION TIMELINE
The AI Act will come into force in 20 days after publication in the Official Journal. Whereas, the Act will be fully applicable 24 months after its entry into force, excluding the provisions imposing bans on prohibited AI systems, which will be active in next six months. Further, rules for Generative AI Systems will apply after 12 months and obligations for high-risk systems in 36 months.
References –
Artificial Intelligence Act: MEPs adopt landmark law
MEPs approve world’s first comprehensive AI law
EU adopts landmark AI law: Everything you need to know