The European Union has issued a warning to Microsoft, indicating the company could face fines amounting to as much as 1% of its global annual turnover under the Digital Services Act (DSA). This follows Microsoft’s failure to adequately respond to a Request for Information (RFI) pertaining to its generative AI tools.
In March, the EU sought information from Microsoft and other major tech firms concerning the systemic risks associated with generative AI technologies. The European Commission noted on Friday that Microsoft had not provided all the requested documents. However, an updated version of the Commission’s press release removed an earlier assertion that Microsoft had not replied at all, instead stating that the EU is escalating enforcement actions “following an initial request for information.”
The Commission has mandated Microsoft to submit the requested data by May 27 or face enforcement measures. Under the DSA, fines can reach up to 6% of a company’s global annual revenue, but providing incorrect, incomplete, or misleading information in response to a formal RFI could lead to a standalone penalty of 1%. This could mean fines of up to several billion dollars for Microsoft, which reported a revenue of $211.92 billion for the fiscal year ending June 30, 2023.
Microsoft’s obligations under the DSA, especially regarding systemic risks, fall under the direct supervision of the European Commission. This warning serves as part of the Commission’s substantial enforcement toolkit, which could inflict more severe consequences than any reputational damage from non-compliance.
The Commission specified that it is lacking information related to risks from Bing’s generative AI features, notably the AI assistant “Copilot in Bing” and the image generation tool “Image Creator by Designer.” The EU expressed particular concern about potential threats these tools pose to civic discourse and electoral integrity.
Should Microsoft fail to comply by the May 27 deadline, the Commission could impose “periodic penalties” of up to 5% of the company’s average daily income or global annual turnover. Bing was classified as a “very large online search engine” (VLOSE) under the DSA in April 2023, imposing additional responsibilities related to mitigating systemic risks, including disinformation.
The requirement for larger platforms under the DSA to mitigate disinformation extends directly to generative AI technologies. Despite their rapid integration into mainstream platforms, generative AI tools remain problematic due to large language models (LLMs) inherently fabricating information and presenting it as fact. Similarly, AI-powered image generation tools can produce racially biased or harmful output, such as misleading deepfakes. The EU’s upcoming elections have heightened concerns in Brussels regarding AI-driven political disinformation.
The Commission’s statement highlighted suspicions that Bing might have breached the DSA concerning risks linked to generative AI, including “hallucinations,” the viral spread of deepfakes, and automated manipulation of services that could mislead voters.
“Under the DSA, designated services like Bing must conduct adequate risk assessments and implement respective risk mitigation measures (Articles 34 and 35 of the DSA). Generative AI is among the risks flagged by the Commission in its guidelines on protecting the integrity of electoral processes, especially for the upcoming European Parliament elections in June,” the press release noted.
In response to the EU’s actions, a Microsoft spokesperson emphasized the company’s commitment to creating safe online experiences and collaborating with regulators on this critical issue.
“We have actively cooperated with the European Commission on this voluntary request for information and are committed to responding to their inquiries. We continue to enhance our digital safety and compliance measures in line with the DSA,” the spokesperson stated. They added, “Across our diverse online services, we strive to measure and mitigate potential risks. This includes various proactive steps to ready our tools for the 2024 elections and protect voters, candidates, campaigns, and election authorities. We will also maintain collaboration with industry peers through the Tech Accord to Combat Deceptive Use of AI in 2024 Elections.”
This article includes Microsoft’s comments and reflects the adjustments made in the Commission’s press release at 15:30 CET, which replaced the earlier claim of receiving no response from Microsoft with an indication of increased enforcement following the initial information request.
