European Union Sets Out to Regulate Generative AI with a Three-Tiered Approach

NNicholas October 18, 2023 7:01 AM

The European Union (EU) is planning a three-tiered strategy to regulate generative AI models, with each tier corresponding to the power and capabilities of the AI systems. This groundbreaking proposal, part of the AI Act, aims to control the swift advancement of this technology and would make the EU the first Western government to enforce mandatory regulations on AI.

The EU's proposed three-tiered approach to AI regulation

In a move to harness the rapidly advancing technology of AI, the European Union is contemplating a three-tiered regulation strategy for generative AI models and systems. According to a Bloomberg-exclusive proposal, rules would be established for different foundation models, which are adaptable AI systems that can perform a variety of tasks. The most potent technology in this field would also be subject to extra external testing.

The proposed AI Act, which is the cornerstone of this legislation, stipulates that AI systems used to predict crimes or sort job applications must undergo risk assessments among other checks. This is a part of the EU's initiative to ensure ethical and responsible use of AI in sensitive areas such as crime prevention and employment.

Classification of AI systems under the proposed rules

The proposed rules break down AI systems into three categories. The first includes all foundation models which form the base for other AI systems. The second category refers to 'very capable' systems, characterized by extensive computing power used to train their large language models. The final category, general purpose AI systems at scale, encompasses the most widely used AI tools, indicating their broad application and impact.

Generative AI software, which has seen a surge in popularity since OpenAI's ChatGPT was released, has the striking ability to respond to basic prompts with text, images, and videos. This is made possible by its large language models, raising both excitement and concerns about its capabilities.

Challenges in EU's regulation of generative AI

Despite the progress made, the EU is yet to resolve several critical issues related to the regulation of generative AI. The nuances of how to implement these regulations and the potential banning of live facial scanning in crowds are among the challenges yet to be addressed. This reflects the complexities and ethical considerations inherent to the regulation of AI.

More articles

Also read

Here are some interesting articles on other sites from our network.