The European Union has officially taken a significant step in regulating artificial intelligence (AI) equipped with first compliance deadline for its AI Act that is now in effect. Regulators can now ban AI systems deemed to pose an “unacceptable risk” to users.
The AI Act was approved by the European Parliament in March 2024 and became enforceable in August 2024. It basically aims to create a comprehensive framework for AI regulation across various sectors. It categorizes AI systems into four risk levels namely minimal risk, limited risk, high risk and unacceptable risk. Minimal risk systems are like email spam filters and face no regulatory oversight. Systems falling under “unacceptable risk” category can be entirely banned.
AI technologies that manipulate the decisions of individuals subliminally or exploit their vulnerabilities such as age, disability or socioeconomic status are flagged as “unacceptable.” Other banned practices include use of AI for social scoring, predicting crime based on appearance and creating facial recognition databases from online images or security cameras.
The penalties for using such technologies in the EU are steep and companies violating the rules can face fines of up to €35 million (about $36 million) or 7% of their annual revenue, whichever is higher.
The EU has laid out additional guidelines and is expected to come in early 2025. It will help in clarifying the rules and their application. The EU also expects companies to demonstrate compliance with the AI Act by August 2025.
The regulatory push comes as more than 100 companies such as Amazon, Google and OpenAI have signed the EU AI Pact last September. The Pact is a voluntary agreement to start applying the principles of AI Act even before its official enforcement.
Some companies like Meta and Apple did not sign the Pact, but most are expected to comply with its core rules. However, there are some exceptions and especially for law enforcement agencies which can use certain biometric AI systems in very specific cases such as identifying abduction victims or preventing imminent threats to life.
The landmark step in AI regulation will likely set a global precedent. The approach of EU provides a model for other regions looking to balance innovation with responsibility.
