The EU plans to tighten rules for the most powerful generative AI models.

Oct 18
As part of a ground-breaking initiative to control rapidly developing technology, the European Union is exploring a three-tier strategy for regulating generative AI models and systems.

These three levels will establish the guidelines for various base models, or flexible artificial intelligence systems, and will necessitate extra outside testing for the most advanced technologies.
The EU is about to impose binding regulations on artificial intelligence, making it the first Western authority to do so. The Artificial Intelligence Act, which is a proposed piece of legislation, would include standards for risk assessments for systems that forecast crime or filter employment applications.

All foundation models will fall within the first group as per the suggested rules. The amount of computing power used to train the massive language models—algorithms that employ enormous amounts of data to generate AI capabilities—in the second tier of "highly capable" systems will vary. The hypothesis suggests that these models might be above the current state of the art and not fully understood.

The most widely used AI tools will be included in the third group, which is referred to as large-scale general-purpose AI systems and will be determined by the total number of users.

The demand for generative AI tools has soared with the introduction of OpenAI ChatGPT last year, and a large portion of the tech industry is hurrying to create their versions.

Based on their extensive language models, generative AI software can often respond to simple queries with an astounding level of competence using text, graphics, and video.
How precisely it will govern generative artificial intelligence and whether or not to outright forbid live face scanning in crowds are just two of the major problems the EU still needs to resolve about its approach to the technology. Plans from the EU parliament and council have come under fire because they could forbid smaller businesses from competing with the industry's major digital titans.

The new AI Authority will be in charge of overseeing additional regulations for massive general-purpose AI systems and extremely effective core models. The paper states that it has the authority to make document requests, coordinate compliance audits, compile a list of certified red testers, launch investigations, and, as a final option, suspend the model.

Created with