The European Commission has released some excellent draft guidelines to promote faster and better AI innovations, given the first deadline of the EU’s AI Act has come into effect. The stated motivators are removing “unacceptable risk” in AI, prohibiting certain practices under European law, in order to help the industry sustainably grow.
Key Guidelines
The clarification on practices deemed unacceptable are mapped out by potential risks to human values and fundamental rights, including:
- Harmful manipulation using subliminal or deceptive techniques
- Social scoring that could lead to unfavorable treatment of individuals
- Emotion recognition in workplace and educational settings (with some exceptions)
- Real-time remote biometric identification in public spaces for law enforcement (with limited exceptions)
- Untargeted scraping of facial images from the internet to create facial recognition databases
- Biometric categorization systems that infer sensitive characteristics like race or sexual orientation
- Individual criminal risk assessment based solely on profiling
These are basically integrity breach rules, reminiscent of how SB1386 confidentiality breach rules of 2003 unleashed a decade of rapid innovation in technology and expansion of the markets related to identity and encryption.
The recent enactment of SB 1386 and SB 1 suggests California is continuing to lead the nation in efforts to protect consumer rights. This creates unique challenges for national and global companies doing business in California or with California residents.
Enforcement and Penalties
Violations of the EU AI Act face only modest penalties—up to 7% of global annual turnover or €35 million, whichever is greater. It remains to be seen whether AI developers and deployers will prioritize compliance given these financial deterrents, which some may view as merely operational costs. Historically, certain American technology companies have appeared to adopt a “catch-me-if-you-can” strategy, seemingly preferring to pay a lazy tax for doing intentional harm to their users, rather than accepting any nudges to innovate.
The prohibitions are now in effect yet enforcement is likely to be staggered as EU Member States have until August 2 to designate the authorities responsible for overseeing them. The guidelines are also currently published in draft form, while translations are still rolling out for all official EU languages.
Legal Status
The Commission emphasizes the guidelines are non-binding, because authoritative interpretations are reserved for the Court of Justice of the European Union. However, they nonetheless spark innovation through insights into how the Commission interprets prohibitions, along with practical examples to help stakeholders understand their obligations.
This initiative represents another step in the EU’s movement towards the lead of global AI with a sensible regulatory framework that balances functionality and features of technology with basic protection of human rights.
