European Union Draft Regulations on Artificial Intelligence
Following on its now famous and influential regulation on data protection (GDPR), the European Union is now tackling the societal and ethical implications of AI with a set of draft regulations published on April 21. Unsurprisingly, industry players are already complaining that these regulations will be too burdensome, and argue instead for "self-regulation," which generally means no real regulation at all. On the other hand, some privacy and civil rights advocacy groups are criticizing the draft for being too lenient on the private sector. Brussels must be thinking that If you get attacked from both sides, you must be doing something right.
The EU approach is to classify AI applications into four tiers, based on the risk of abuse. The most risky (such as face recognition used for racial profiling) would be banned, the least risky would be allowed with minimal safeguards, and the intermediate tiers would require various levels of testing and labeling.