The EU Commission is escalating its scrutiny of Microsoft over generative AI in Bing, with election risks at the center of the dispute. The Commission wants internal documents and data on how Microsoft is handling problems such as false information, deepfakes and automated manipulation in Bing’s AI features.
The pressure comes as generative AI becomes part of search, where users may treat answers as factual even when the systems make errors. For Microsoft, the issue is not only technical performance. It is now a compliance question under the Digital Services Act.
Why Bing is under EU scrutiny
Bing is treated as a "very large online search engine" under the Digital Services Act, which means it must meet the platform obligations attached to that status. The Commission says alleged DSA breaches could create risks for political discourse and elections.
The specific concern is the use of generative AI in Bing and Bing Copilot. According to the source article, adding generative AI to Bing Copilot has not helped Microsoft gain ground in the search market, but it has produced bizarre situations and misinformation for users, including around elections.
The EU sees several types of risk in these systems. The Commission has pointed to hallucinations, deepfakes and automated manipulation of services that could mislead voters. In search, these risks matter because the product sits close to how people look for information, compare claims and form an initial view of public events.
What the Commission is demanding
The Commission first made a request on March 14 about risks linked to Bing’s generative AI capabilities and the image generator. Microsoft now has until May 27 to provide the requested information.
The latest step is legally binding. The Commission is asking for internal documents and data that Microsoft had not previously provided. The purpose is to assess Bing’s risk mitigation measures, as required by the DSA.
The timing is important because the Commission has identified generative AI as a risk for the upcoming European Parliament elections in June. The issue is not simply whether an AI chatbot makes mistakes in ordinary search sessions. The larger question is whether a search engine’s AI functions can distort information around elections and political debate.
- The request covers risks from generative AI in Bing.
- The Commission is focused on false information, deepfakes, hallucinations and automated manipulation.
- Microsoft must provide internal documents and data not previously submitted.
- The deadline named in the source is May 27.
The financial stakes for Microsoft
If Microsoft does not comply within the specified period, the Commission may impose fines of up to 1% of the provider’s total annual turnover or worldwide turnover. It may also impose periodic penalty payments of up to 5% of the provider’s average daily turnover or worldwide annual turnover.
The DSA also allows the EU to impose fines of up to 1% of total annual turnover for the dissemination of false, incomplete or misleading information. The source notes that Microsoft has annual revenues of more than $200 billion, which means such a fine could amount to several billion dollars.
That makes the case significant beyond Bing’s market position. Even if generative AI has not shifted the search market in Microsoft’s favor, the compliance risk attached to AI search can still be large. A search engine that adds generative answers may also take on new obligations to prove that it understands and mitigates the risks those answers introduce.
AlgorithmWatch’s warning on election answers
AlgorithmWatch has raised concerns about Bing Copilot in election research. According to Clara Helming, Senior Policy and Advocacy Manager at AlgorithmWatch, Copilot has named false candidates, misrepresented poll results and invented scandals about politicians in that context.
Helming also said the error rate remained constant for several weeks. AlgorithmWatch says it informed Microsoft but saw no improvement. With European and German state elections coming up, the organization urged quick action.
These examples show why election-related AI errors are different from ordinary product glitches. A made-up candidate, a wrong poll result or an invented political scandal can create a false impression at the exact moment when users are seeking reliable information. When that happens inside a search product, the error may appear more authoritative than it would in a casual chatbot exchange.
Why Google is also in the picture
The source article argues that the case should also be a warning to Google. Google offers an alternative to Bing and ChatGPT with the Gemini chatbot and has been aggressively integrating generative AI into Google Search through Search Generative Experience, or SGE.
The Commission has not yet issued specific warnings to Google, according to the source. SGE is currently being rolled out increasingly in the US and also in many other countries outside the EU.
The broader issue is that chatbot-style search changes the relationship between search engines, users and website content. The source notes concerns that Microsoft has misused Bing’s suggestive power for unlabeled ads, and that Bing, SGE and Perplexity use unauthorized website content to answer questions directly, undermining website owners.
The article also points to OpenAI, Google and reportedly Apple as companies influencing the future media landscape through murky backroom deals with select publishers. Those concerns sit alongside the DSA case: AI search is no longer only a product feature. It is becoming a policy, media and competition question.
For now, Microsoft faces the clearest immediate deadline. By May 27, it must provide the information the Commission is demanding or risk penalties tied to its turnover. The outcome will show how forcefully the EU is prepared to apply the DSA to generative AI inside search.