Google finds its artificial intelligence strategy under intense legal examination this week in a Washington D.C. federal court. As the remedies phase of its landmark search antitrust trial commenced on April 21st, the focus quickly shifted to Google’s current conduct in the AI market.
Court testimony brought to light that the tech giant has been channeling “enormous sums of money” monthly to Samsung Electronics Co. since January 2025, securing the preinstallation of its Gemini AI on the manufacturer’s popular mobile devices.
The Department of Justice is framing this arrangement as a direct extension of the monopolistic behavior Google employed in the search market – practices for which Judge Amit P. Mehta already found Google liable last August.
Government lawyers contend Google is deploying the same playbook, leveraging its existing dominance unfairly. The DOJ further expressed concern that Google could use profits from its search business to fund these potentially exclusionary AI deals, disadvantaging competitors unable to match such spending. Google’s past default deals across all partners were reported to cost $26.3 billion in 2021 alone.
Echoes Of Past Practices In New AI Deals
Details of the Google-Samsung Gemini pact emerged via testimony from Google Vice President Peter Fitzgerald. Since January, Google provides fixed monthly payments per device alongside a cut of advertising revenue generated within the Gemini app. T
he contract was initially set for at least two years, and court documents indicated it might span until 2028. This setup exists despite Judge Mehta’s prior ruling that Google’s earlier payments to Samsung – reportedly $8 billion between 2020-2023 for Search, Play Store, and Assistant defaults – were illegal.
While Fitzgerald testified the Gemini agreement permits Samsung to install rival AI services and that Google removed search exclusivity in April 2025, the DOJ pointed to other agreements. A deal with Motorola to preload Gemini onto Razr devices, requiring home screen placement, was described by reporting based on court documents as “nearly identical” in structure to the previously condemned search arrangements.
The Justice Department is consequently seeking significant interventions. Beyond demanding a ban on exclusive default placement deals for both search and AI products, proposals include potentially forcing Google to sell its Chrome browser and even divesting parts of its lucrative advertising business. “We’re here to restore competition to these markets,” stated DOJ lawyer David Dahlquist, aiming for remedies that would “allow that block of ice to thaw.” He emphasized that “The antitrust laws are designed to adapt to advances in technology, the oil companies and railroads of yesterday are the internet and search engines of today.”
Gemini’s Role And Rapid Evolution
Gemini is clearly central to Google’s future plans. The company is actively replacing Google Assistant with Gemini as the primary AI on most Android devices, amplifying the importance of deals ensuring its presence on hardware from manufacturers like Samsung.
The latest model underpinning these efforts, Gemini 2.5 Pro, saw a remarkably fast rollout. Launched to paying subscribers on March 25th, an experimental version became freely available to all Gemini web app users starting March 29th. Google attributed this acceleration on X to wanting “to get our most intelligent model into more people’s hands asap.”
This version of Gemini boasts advanced capabilities, including a large 1 million token context window (meaning it can process large amounts of information like entire books at once) and strong performance in multimodal tasks (handling text, images, and potentially other inputs together).
Analysis suggests it uniquely understands PDF visual structures, enabling more precise citations, though Google’s own documentation notes limitations in exact spatial reasoning. Benchmarks show strength in areas like mathematics but also indicate it trails competitors like OpenAI’s GPT-4.5 in factual recall and Anthropic’s Claude 3.7 Sonnet in autonomous coding.
Google continues to build out the Gemini platform, making features like Deep Research and customizable Gems free in March, and integrating AI deeply into Workspace apps like Drive, Docs, Sheets, and Gmail. The company is also testing “Scheduled Actions” for task automation. This rapid development, however, faced some criticism regarding the timing and detail of its safety reports for the widely released 2.5 Pro model.
Crowded Field And Google’s Rebuttal
The market for AI assistants is highly contested. An internal Google presentation confirmed that rivals OpenAI, Microsoft, and Meta had approached Samsung regarding AI placement on Galaxy devices. Furthermore, testimony revealed Motorola was engaging with other AI firms, “including Perplexity,” according to Google VP Fitzgerald.
Google vigorously contests the DOJ’s narrative. Lead counsel John Schmidtlein called the proposed remedies “extreme” and “fundamentally flawed,” asserting they were “untethered from the liability findings.” He argued Google achieved its success through “hard work and ingenuity” and won its position “fair and square,” suggesting that AI competitors “would like handouts as well even though they are competing just fine.” Google insists users prefer its services for their quality, not due to anticompetitive contracts – an argument similar to one Judge Mehta rejected regarding the earlier search deals.
The remedies phase of the trial is expected to conclude within approximately two weeks. Judge Mehta has indicated he anticipates delivering a ruling on the penalties by late summer or early fall 2025, deciding how to address Google’s search monopoly in the context of the burgeoning AI landscape.
