The European Union has officially began to implement its AI Act, marking the first compliance time limit on 2 February. This regulation is considered a “unacceptable risk” for individuals or society to European Union officials.
The AI Act approved in March 2024, from August to August and from August, AI classifies risks into four levels: minimal, limited, high and unacceptable. While minimal and limited-risk AI (such as spam filters and chatbots) faces slightly oversight, high-risk systems, such as healthcare, are heavy regulated. The most severe category, unacceptable-risk AI, is now prohibited in the European Union.
Some of the restricted AI applications include social scoring, future policing based on presence, real -time biometric monitoring in public places, emotion detection in workplaces or schools and AI systems that manipulate or cheat individuals. Companies violating these rules can face a fine of €35 million (~ $ 36 million) or 7% of their annual revenue, which is also high.
However, some exceptions exist. Law enforcement agencies may use biometric AI for targeted discoveries, such as finding missing individuals or preventing adjacent hazards, provided that they get a proper authority. Similarly, AI system can be allowed for workplace or educational emotion analysis when appropriate for medical or safety reasons.
While more than 100 companies including Amazon, Google and Openai have voluntarily promised to comply with the EU AI Treaty, prominent players like Apple, Meta, and Mistral AI have not joined. Despite this, all companies working in the European Union will eventually have to follow the new rules.
The next major compliance time limit is in August, when the enforcement mechanisms and punishment will be officially effective. Since regulatory officials have continued to refine the guidelines, businesses have to navigate how the AI Act moves with existing laws such as GDPR, further increases the challenges.
Thanks for reading..
