NEWS
The EU is hoping is to emulate the success of the GDPR, the EU’s data privacy law, by establishing a new EU Artificial Intelligence (AI) Act that could also set the “gold standard” for AI legislation around the world.
Central to the EU AI Act is the categorizing of AI applications into four groups based on risk:
- Unacceptable risk: AI solutions (such as social scoring) which present an unacceptable threat to people’s safety, livelihoods or rights would be banned outright
- High risk: AI solutions considered high risk (such as those controlling critical infrastructure) would be subject to strict controls, both before being put on the market and throughout their lifecycle
- Limited risk: such solutions would have to meet transparency requirements so that people are aware that they are interacting with AI and can make an informed decision whether to continue or step back. ChatGPT would fall in this category
- Minimal risk: the Act foresees the free use of minimal risk AI and the EU believes the vast majority of current AI systems fall into this category
On 14th June, the EU Parliament voted in favour of the AI Act, paving the way for the final round of negotiations. If all goes smoothly, this could see the AI Act finalized by the end of 2023 and coming into force two years later, thereby becoming the world’s first comprehensive AI law. However, whether it will have the maturity and robustness to stand the test of time in such a dynamic and evolving domain remains to be seen. A new “gold standard” or more an iron “ball and chain” for the EU, we will see.
