EU Mulls Ban On Using AI For Mass Surveillance & Social Credit Scores

The European Union is considering a ban on the use of AI for a slew of use cases, including mass surveillance and social credit scores, according to a leaked draft proposal by the European Commission, first accessed by Politico. An official statement is expected next week. 

One of the proposals in the draft recommends European Commission to ban certain use-cases of AI and limit its use for other applications if they don’t meet certain standards. The document recommends a ban on the use of AI for mass surveillance or developing a social credit score system.

The draft proposal also seeks special authorisation to use ‘remote biometric identification systems’ and demands an explicit notification to people when they interact with AI systems, ‘unless it’s obvious’. It also calls for oversight on “high-risk” AI systems that pose a direct threat to someone’s safety, like self-driving cars or systems that directly impact livelihood like AI systems used for hiring, assigning recidivism scores, or allocation of personal loans.

Member states of the EU would be required to set assessment boards to test and validate the high-risk AI systems. The draft proposal also calls for a ‘European Artificial Intelligence Board’ with representation from all member states to help the European Commission identify which systems can be classified as high-risk.

Companies that don’t comply could be fined up to €20 million or 4% of their turnover.

While the US and China have focused their attention on developing powerful AI systems, they have fallen short in setting up an airtight regulatory framework to ensure individual safety and rights. The EU has set up a powerful GDPR (General Data Protection Regulation) to address such issues. This draft proposal is also in line with EU’s "human-centric" approach towards developing AI.

However, the leaked draft has drawn flak from policy wonks, calling for improvements in terms of the ambiguity of the language used. The experts demand more clarity on what constitutes AI and the definition of what’s detrimental or high-risk in AI use cases.