European proposal for an #AI regulation is here. Will require transparency, so explainable AI? Some uses of AI prohibited. Also the first law that regulates and allows the use of deepfakes. https://ec.europa.eu/newsroom/dae/document.cfm?doc_id=75788

My analysis of the #AI Regulation allowing deepfakes. Many risks exist and will remain. Are societies ready for this transformative technology? https://blog.lukaszolejnik.com/european-union-regulating-ai-allowing-the-use-of-deepfakes/

European #AIRegulation does not in any way impact on lethal autonomous weapons systems. Explicitly placed outside.

Research on the techniques of misusing AI should be accepted. So Cambridge Analytica-style approach is okay? #AIRegulation

Scoring and classifying people's behaviour (to "sort" people using some order) will be a prohibited use of AI. This goes beyond the China's social score. This should prohibit the use of AI in recruitment/employment systems, no exceptions? #AIRegulation

European Union's #AIRegulation not only restricts the use of high-risk AI, but also its OUTPUT. So when high-risk AI is used in a third-country, the output of such computations may be prohibited in Europe?

"High-risk AI systems should ... meet an appropriate level of accuracy, robustness and cybersecurity in accordance with the generally acknowledged state of the art". What does it actually mean today? In 10 years? #AIRegulation

According to the #AIRegulation, high-risk AI systems may recreate potentially existing biases existing in the society?

According to the #AIRegulation, in some cases data collected for unrelated purposes may be used in the development of an AI system. Is it a risk of function creep? #GDPR