EU Demands Microsoft Share Bing AI Risk Data Or Face Fines

May 18, 2024

Recently, the European Union has become more careful about the generative artificial intelligence (AI) technologies that big tech companies like Microsoft use. This is because these technologies could be dangerous during the EU polls, which are coming up from June 6th to June 9th across the 27-member bloc.

EU’s Action Against Microsoft’s Bing

The European Commission is looking into adding creative AI to Microsoft’s Bing search engine. They are mainly looking at “Copilot in Bing” and “Image Creator by Designer.” Based on what the Commission says, these traits may help spread false information and change people’s minds during elections. The biggest worries are “hallucinations,” which are false information made by AI, and the spread of deepfakes like a virus.

Compliance Deadline and Potential Penalties

The date for Microsoft to turn in the necessary internal documents and data is May 27. If the company doesn’t, it could be fined. These fines could be up to 1% of their yearly income or global turnover, plus fines every day that could add up to 5% of their average daily global turnover.

Regulatory Framework: The Digital Services Act (DSA)

The investigation into Microsoft is part of the larger Digital Services Act (DSA) framework. The DSA is a new piece of EU legislation that aims to regulate the operations of digital platforms to protect users from harmful material and false information. Tech giants have to show that they follow DSA, and the European Commission says Microsoft is currently at risk of not doing so.

Microsoft’s Stance and Cooperation

The European Commission and Microsoft have agreed to work together fully. A Microsoft representative said that the company was working to meet the Commission’s request for information and went over the steps they are taking to stay in line with the DSA framework.

More About Generative Artificial Intelligence

Generative AI is a term for algorithms that take learned inputs like text, images, and sounds and make new data outputs. Ian Goodfellow created Generative Adversarial Networks (GANs) in 2014, and they are very important for making realistic media. The advanced model VQ-VAE-2 makes picture quality and resolution much better. Using prompts, OpenAI’s GPT-3 is very good at writing text that sounds like it was written by a person. Google’s BERT engine helps language processing understand context better. AI Dungeon 2 makes big text-based games with GPT-3. Generative AI can also help predict how proteins will fold, which could speed up medical breakthroughs. The ethical consequences include the chance that it will be abused to make deepfakes or spread false information.

« Previous Post Next Post »