Today, the European Parliament agreed on proposed rules on artificial intelligence, which should subsequently be formally adopted by EU member states. The new rules aim to make it easier to identify AI-generated content, including deep fakes, and to completely ban the use of AI in biometric surveillance, emotion recognition, and criminal behavior prediction.
The new rules stipulate that AI tools such as OpenAI’s ChatGPT must indicate that content is generated by artificial intelligence and take some responsibility for ensuring that users are aware of whether an image is fake or real. This looks like a daunting task because once an image is created, it’s hard to limit how it can be shared by the user, but maybe AI companies will have to deal with this shortly.
If the new rules are adopted by the European Parliament in their current form, artificial intelligence models will have to provide “detailed summaries” of the author data used in training. In particular, for OpenAI, this would mean the need to disclose the training data for their GPT-3 and GPT-4 scale models, which are not currently available for review. There are large datasets used to train AI models that already make this data available, such as LAION-5B.
Certain uses of AI will also be completely prohibited, especially those that may violate the privacy rights of EU citizens:
Remote biometric identification systems in real-time and “after the fact” in public places
Biometric categorization systems using specific characteristics (e.g. gender, race, ethnicity, citizenship, religion, political orientation)Advertisement
Predictive policing systems (based on profiling, location, or past criminal behavior)
Emotion recognition systems in law enforcement, border management, at work, and educational institutions
Non-specific collection of information from faces from the Internet or video surveillance to create face recognition databases (violation of human rights and privacy rights)
These rules have not yet been passed into law. Before that happens, Member States can submit proposals and the process will start today.
It appears that the EU intends to make sure that it controls the use of artificial intelligence and its potential applications to the extent possible.