The AI Act, a significant development in AI regulation, has been passed by an overwhelming majority. It has been hailed as one of the most important developments in global AI regulation and is expected to set the standard for years to come. However, immediate clarity on its implementation may not be forthcoming due to the complex European system.
The next step involves members of the European Parliament negotiating details with the Council of the European Union and the European Commission before the draft rules become legislation. This process will result in a compromise between three different drafts from each institution, which currently vary significantly. It is estimated that it will take approximately two years for these laws to be fully implemented.
The recent vote accomplished approving the European Parliament’s position for final negotiations. The AI Act follows a risk-based approach, imposing restrictions based on perceived dangers posed by AI applications. Businesses will also need to submit their own risk assessments regarding their use of AI.
Major implications of this legislation include:
Ban on emotion-recognition AI: The draft text from the European Parliament prohibits the use of AI attempting to recognize people’s emotions in policing, schools, and workplaces. The accuracy and bias issues surrounding facial detection and analysis have been criticized; however, it has not been banned in drafts from other institutions, indicating potential political disputes.
Ban on real-time biometrics and predictive policing: This aspect may face significant opposition as EU bodies determine how and whether to enforce a ban on real-time biometric technologies in public spaces. Policing groups argue that such technologies are crucial for modern policing, with certain countries planning to increase their use of facial recognition.
Ban on social scoring: Public agencies using data about individuals’ social behavior to make generalizations and profiles would be prohibited from doing so under this legislation. However, social scoring has multifaceted implications beyond simply mirroring practices associated with authoritarian governments like China; it is commonly used in mortgage approvals, insurance rates determination, hiring, and advertising.
New restrictions for generative AI: The draft bill is the first to propose regulations for generative AI, including banning the use of copyrighted material in training large language models such as OpenAI’s GPT-4. European lawmakers have already scrutinized OpenAI concerning data privacy and copyright concerns. Additionally, the bill requires clear labeling of AI-generated content.
New restrictions on recommendation algorithms on social media: This draft categorizes recommender systems as “high risk,” subjecting them to increased scrutiny and potentially holding tech companies more liable for user-generated content’s impact.
Margrethe Vestager, executive vice president of the EU Commission, has highlighted concerns regarding the risks associated with AI. These risks include trust in information, vulnerability to social manipulation by malicious actors, and mass surveillance.