EU passes AI Act

*European Commission passes landmark AI Act* This act will regulate AI in Europe and was passed last week by the European Commission. It categorizes the use of AI by risk level and regulates it based on that.

High level overview:

❇️ High risk is defined as use cases that can affect safety, critical infrastructure, employment, law enforcement, democratic processes. Basically anything that has high likelihood of bad data creating negative outcomes. Healthcare falls into this category and has strict regulations. For example, using AI for #robotic surgery, #biometric data, and use in clinical settings. It can be used but has to demonstrate that there is high quality data, traceability of data, robustness, security. It also, interestingly, requires human oversight to minimise risk (something that most of us using AI clinically do anyway). Of note, using biometric data from cameras and tracing emotions is prohibited overall. So use in tracking etc.

❇️ Limited or no risk is less regulated. It is stuff like chatbots, social media etc. It has to be transparent that AI was used but there is less regulations. This is new and will likely inform other countries and regions on how to regulate AI. In the case of #medicine, which has potential to create bad outcomes if we use bad data and algorithms, these regulations make sense.

I'm sure someone will argue that this will stifle innovation. I don't agree at least not in #healthcare. Medicine is still risk averse and everyone is cautious. While, the USA tends to act more wild west (a technical term) in terms of innovation than Europe, this is something that can help us too. Getting some guidelines will actually make clinicians feel safer using it.