06/15/2023 at 12:15 p.m. by Claus Ludewig – The EU Parliament has agreed on rules for AI tools, paving the way for the formulation of the law. ChatGPT and Co. must therefore meet transparency requirements and are banned in certain areas.
Since the beginning of the year everyone has been talking about ChatGPT. Finally, the AI tool is accessible to the general public after previously having limited availability. Short text entries can be used to generate letters, write code, etc. The “AI godfather” Dr. Geoffrey Hinton recently stated that it is the responsibility of politicians that AI “is developed with a lot of thought about how to prevent it from going rogue”. After all, there are always reports about alternative facts. The European Parliament has now announced that rules for safe and transparent AI are being negotiated.
More transparency and ban on AI in certain areas
On Wednesday, June 14th, the European Parliament adopted a legislative proposal regulating artificial intelligence. As a result, talks can now start with EU member states to shape the final form of the law. According to the EU, the regulations are based on the level of potential danger. It should be forbidden to use AI tools for the following services within the EU:
No AI tools for “social scoring”, as is common in China, for example. No biometric systems that make it possible to identify people in real time or subsequently in public places using cameras etc. No predictive police work with AI tools. It’s all about profiling potential criminals. No emotion-recognition systems in the workplace, educational institutions, border patrols, and law enforcement. No random reading of facial images from the Internet or from surveillance cameras to create facial recognition databases.
All AI systems that endanger the health, safety and fundamental rights of people or the environment are considered high-risk applications and must comply with strict regulations. Also included in this list are AI tools that influence voting behavior. All providers of other AI tools, e.g. B. Open AI with ChatGPT, will have to assess the risks to health, safety, fundamental human rights/environment, democracy and the rule of law in the future. Before publishing in the EU, providers must upload their tools to an EU database and meet transparency requirements, e.g. B. Clearly label that content is AI-generated and create summaries of which proprietary data was used for training purposes. If you want to test ChatGPT-4 for free, you can use the new Microsoft Bing, provided you have a Microsoft account. In connection with Microsoft 365, future AI features with the AI copilot will be used in Outlook and Co. This is how e-mails can be formulated.
Also worth reading: “AI godfather” Geoffrey Hinton warns of the dangers of ChatGPT and Co.
Collection of regulations for ChatGPT and Co. in the EU:
In order to contain the potential dangers of AI tools, the EU Parliament has developed regulations for secure and transparent AI. As a result, talks can now start with the EU member states to shape the final form of the law. According to the EU Parliament, the Use of AI tools to recognize people or emotions is prohibited. In addition, there must be no “social scoring” in the EU. Providers of AI tools, such as Open AI, must meet transparency requirements in order to continue offering ChatGPT & Co. in the EU. For example, it must be flagged if content is AI-generated and the provider must create summaries of which proprietary data has been used for training purposes.