The European Parliament announced on December 9 that it will introduce the Artificial Intelligence Act, which is a bill that strictly restricts the use of AI and has penalties. Once formally approved by the European Parliament and the European Council, it will become law with the force of law.
For the first time, the bill divides AI into four types: high, medium, low-risk and exempt. It also sets very strict regulatory requirements for the development and use of high-risk AI systems, including: risk assessment, privacy and data protection, transparency, security, fairness, and accountability. At the same time, the requirements for risk assessment, transparency and security are also put forward for medium-risk AI systems.
Under this law, the use of biometric classification systems for sensitive characteristics, social scoring systems based on social or personal characteristics, manipulative behavior, Xi and emotion recognition at work will be severely restricted. However, prior judicial authorization may be waived. These regulations have a significant impact on the financial system, among other things.
The bill pre-emptively imposes legal requirements on artificial general intelligence (GPAI). These systems and their models must comply with the principle of transparency. The law is relatively friendly for open-source AI systems. At the same time, MEPs are demanding the protection of SMEs and innovation.
However, the bill also provides specific penalties, which can result in fines for non-compliance. The amount of the fine can range from €7.5 million to €35 million, or 1Anywhere from 5% to 7%, depending on the violation and the size of the company.
The EU's Artificial Intelligence Act could have a profound impact on the AI industry. On the one hand, it provides the first paradigm for comprehensive regulation of AI, especially by setting clear rules such as risk, transparency, etc., to provide guardrails to promote AI innovation, but on the other hand, it may also impose huge compliance costs for AI companies and have an impact on global AI investment and adoption.