The European Union is opening the consultation period to regulate general-purpose models. The goal is to make these models reliable.
Organizations, rights organizations, rights holders, and academic experts are being asked to provide input on regulation for AI models that can be widely used. The EU designates the task of these models as general-purpose. Examples of this category are the models of OpenAI, Google, and Microsoft; these providers with operations in Europe may also provide input.
August 1 in action
We are on the eve of the passage of the European AI Act. The start of the consultation period seems late, but it is possible because of the gradual rollout of the new set of rules. In other words, the dates by which organizations must comply are scattered.
The part the European Commission is now working on will apply from April 2025. “The consultation is an opportunity for all stakeholders to express their views on the topics covered by the first Code of Practice, which sets out the rules for general purpose AI model providers.”
So this will be done in consultation with experts. These consultation sessions with the field have become the general practice in the development of the AI Act. Because of the sweeping nature and young age of the technology, the legislation was always developed in consultation with professionals in the field. This, incidentally, created a narrow line between consulting and lobbying, where consulting more than once seemed to spill over into lobbying for less stringent regulation.
Tip! Are OpenAI, Microsoft and Google lobbying a way out the AI Act?
First draft in September
Applications for the consultation period are possible until Sept. 10, 2024. In September, the Commission plans to start discussions with interested parties to shape the code further. A first draft of the rules will be made based on the input from the questionnaire that prospective participants are required to fill out upon application.
The consultation will be done in close consultation with the AI Office. This office is charged with exploring possible ways to detect AI risks. Part of the first draft will be based on results already obtained by the AI Office.
Also read: AI Act: OpenAI and Google may not violate copyrights