The bill is the world’s first comprehensive regulation of AI, marking an important step in the EU’s efforts to regulate the rapidly evolving use of AI.
The bill aims to categorize and regulate AI systems through a risk-based approach, ensure their safety and transparency, and protect fundamental rights and democratic institutions. Under the bill, all AI systems, including chatbots, must clearly inform users that they are interacting with an AI system.
Comprehensive regulation and risk-based approach
At the same time, it is the responsibility of AI technology providers to ensure that audio, video, text and images synthesized by AI can be identified as AI-generated. The bill also explicitly prohibits the use of AI systems that may clearly threaten the fundamental rights of users.
Responsibilities and prohibitions for AI technology providers
In addition, the bill establishes special requirements for high-risk AI systems, including rigorous pre-market evaluation and ongoing monitoring.
It is reported that the EU’s Artificial Intelligence Bill is based on the concept of risk prevention, and formulates a set of risk regulation system covering the whole process for artificial intelligence. The Act adopts risk-graded management measures to categorize the risks of AI systems into four types: unacceptable risk, high risk, limited risk and slight risk, and imposes different regulatory measures for different types.
Risk-based classification and regulatory measures
The Artificial Intelligence Bill defines a framework to understand the risks associated with AI and to take decisions or actions based on the data obtained and the use of that data. The bill categorizes the risks associated with artificial intelligence:
- The prohibition of unacceptable risks (e.g., social scoring systems and manipulative AI).
- Most of the text deals with high-risk AI systems that are subject to regulation.
- A smaller section deals with AI systems with limited risk, but with lighter transparency obligations: developers and deployers must ensure that end-users are aware that they are interacting with AI (chatbots and deep forgeries).
- Minimal risk is not regulated (including most AI applications currently available in the EU single market, such as AI video games and spam filters; but this content is changing with the generation of AI).
For example, high-risk AI systems will be subject to strict obligations before they are placed on the market, including transparency requirements, data governance, accuracy, robustness and cybersecurity. For example, generative base models such as GPT must have adequate safeguards in place to ensure they comply with EU law, and disclosures must be made where text is AI-generated.
The bill also explicitly prohibits the use of AI systems that may clearly threaten the fundamental rights of users. Such risky AI systems will be prohibited. Any activity that violates this provision will be subject to severe legal sanctions.
It is worth noting that while certain provisions will be fully applicable within two years, some of them will be implemented earlier. For example, schools and workplaces will not be allowed to use real-time emotion-recognition systems, and AI systems such as chatbots will have to clearly inform users that they are interacting with a machine.
Global impact and compliance requirements
The EU’s influence as the world’s first law to comprehensively regulate AI is not limited to within Europe, but could also have a significant impact on AI companies globally. This includes companies such as Google and OpenAI, which will need to adapt to new compliance requirements and standards, thereby driving the global AI industry to a higher level of standardization.
Therefore, the entry into force of the EU AI Act not only provides a legal framework for the development of AI, but also provides a reference for regulation in other countries and regions.
For more information, welcome to the Latest News official website.
