The European Union has issued a global ban on Microsoft under the region's online governance regime, the Digital Services Act (DSA), for failing to respond to legally binding information requests focused on: It warned that it could be fined up to 1% of its annual turnover. Its generation AI tool.
The EU asked Microsoft and a number of other tech giants for information on the systemic risks posed by their generative AI tools back in March, but on Friday the European Commission provided some of the documents Microsoft had requested. announced that it did not.
The company has until May 27 to provide the requested data and risk enforcement. Fines under the DSA can extend up to 6% of global revenue, but if the information provided in response to a formal request for information (RFI) is inaccurate, incomplete, or misleading, may be subject to a 1% fine. In Microsoft's case, this could result in fines of up to billions of dollars.
The systemic risk obligations of large platforms under the DSA will be overseen by the Commission itself. And today's written warning to Microsoft for failing to provide requested information puts it on top of its toolbox of strong enforcement options, and for failing to provide data as requested. The cost to the company could be much higher than the reputational damage that comes from being reprimanded.
The commission said Microsoft was missing information related to the risks posed by the generative AI capabilities of its search engine Bing, particularly its AI assistant Copilot in Bing and image generation tool Image Creator by He emphasizes “Designer”.
The EU said it was particularly concerned about the risks this tool poses to public debate and the electoral process.
Microsoft has until May 27th to provide the missing information or face a 1% fine. If the company fails to submit the data by then, the European Commission could also impose a “periodic fine” of up to 5% of average daily income or global annual revenue, which could affect Microsoft's could potentially reach billions of dollars. .
The company's search engine Bing was designated as a so-called Very Large Online Search Engine (VLOSE) under the EU's DSA in April 2023. This means additional obligations related to mitigating systemic risks such as disinformation.
Despite obvious flaws, such as the tendency of large-scale language models (LLMs) to fabricate information, DSA is at the forefront of large technology companies incorporating GenAI into their mainstream platforms. We are tasked with mitigating disinformation for platforms of scale, and we are squarely focused on generative AI technology. While presenting it as a fact.
AI-powered image generation tools have also been shown to produce racially biased and other potentially harmful outputs, including misleading deepfakes. Meanwhile, with the EU due for elections next month, political disinformation using AI is the focus of attention in Brussels.
“The request for information raises suspicions that Bing may have violated the DSA regarding risks associated with generative AI, including so-called ‘hallucinations,’ the spread of deepfakes, and automated operation of services that may be misleading. “voters,'' the commission said in a press release.
“Under the DSA, designated services, including Bing, must conduct appropriate risk assessments and adopt respective risk mitigation measures (Articles 34 and 35 of the DSA). This is one of the risks identified by the European Commission in its guidelines on the integrity of electoral processes in the run-up to the upcoming European Parliament elections.
Microsoft has been asked to comment on the EU's complaints.