The European Union (EU) released the world's first comprehensive set of laws governing artificial intelligence (AI) on March 13th 2024. The groundbreaking legislation is a significant development as the EU aims to mitigate the potential risks of AI’s ever-expanding capabilities and influence over our lives.
The Act is geared towards establishing an environment in which AI can be trusted, which essentially means ensuring that systems are safe, unbiased, and accountable. In practice this means that AI developers will be required to follow specific guidelines, including following data management laws and establishing clear ways to ensure effective and transparent human oversight.
The act is based on a framework in which AI applications are classified based on their perceived risk level. "Unacceptable risk" applications, like social scoring systems are made illegal, while “high-risk" applications, such as AI-powered recruitment tools, will be subject to stricter regulations to ensure fairness and user safety. Lower-risk applications such as spam filters will be granted more freedom.
On the issue of generative AI, whose risks for misuse include the spread of misinformation and deepfakes, the act states that generative AI providers will be required to label their output and disclose images, video, or audio content that have been manipulated or generated to the extent that it would constiute a deepfake. Additionally, Developers of generative AI models may also have to provide information and cooperate their models are used in high-risk applications. The current draft also states that the use of generative AI to support illegal activity is prohibited.
This development is a strong sign of intent from the EU,and may represent a significant development towards curbing the spread of online disinformation. However, the act is only a start, and the legislation will only have a credible impact if it is combined with effective enforcement, at both the organizational and technical level.
This is nonetheless a landmark moment, setting the precedent for global AI regulation in a rapidly developing environment. This may prove to be essential in a critical year for democracies worldwide.
Comments