Judge Mehta is zeroing in on the possibility Google could only share those long-tail queries. The plaintiffs say the approach might make sense, but that it’s not always obvious how to define a long-tail query. Mehta objects to that, saying it seems like something competitors should figure out. “They don’t have access to the same data ... that’s the problem,” he says — not that they can’t figure out what they are.
Mehta has been pushing on ways Google could share its search data without including significant amounts of user data or kneecapping Google and still give rival search engines enough to improve their products. He’s pressing the government on whether witnesses really focused on comprehensive user data, or if they were more interested in Google’s index. He cites an amicus brief from Brave, which suggested limiting the shared data to unusual and hard-to-get “long-tail” results that can supercharge search quality. Also, how often should Google have to provide fresh results, and how much should it cost competitors to access them? “We’re trying to kickstart competitors, we’re not trying on Day 1 to put them on equal footing with Google,” he says.
“I understand this is now forward-looking, but I don’t know that” there’s an argument that would “allow the court morphing the market” this way, he says. “What case can you look to that has that kind of quality to it that can address that kind of emerging market and put it in this box of qualified competitor?” Judges’ questions don’t always indicate how they’ll rule, but that might bode ill for some of the companies Google is most worried about — generative AI providers — getting access to valuable data.
Mehta asks the question to the government. “I don’t think they’re interested in competing as a search engine qua search engine,” he says, elaborating. They “want to have better grounding in search” to have the best model in the world; is that sufficiently related? “It feels like to me the definition ought to be changed” of a general search engine if so, he says. The government disagrees, saying AI companies are clearly aiming to compete in the same areas. Mehta points out a contradiction: the government wanted to exclude a bunch of other search-engine-like services to establish Google had no meaningful search competitors during the liability trial, and now it wants to add new ones during remedies. “I’m not sure it quite fits,” Mehta says.
We’re still arguing about why data sharing is important, and whether Google’s “quality gap” of information compared with competitors is both significant and unfairly obtained. Data sharing and AI are emerging as two of the biggest elements of these arguments, which isn’t surprising: the former is something Google sees as potentially catastrophic, and the latter is where it sees the future of search.
We’re taking a 15-minute break before getting back to arguments — after a little chitchat, it’s near-silence on the teleconference line.
The DOJ’s lawyer is back up for rebuttal — arguing that Google might say its remedies are extreme, but a truly extreme remedy would be something like requiring it to sell search. The goal is denying Google the fruits of the monopoly it’s established and making sure there are “no practices” left to help it unfairly maintain it, he says.
“I don’t believe you have a factual basis to ... suggest there was a causal connection” between Google’s multibillion-dollar deal and Apple not developing a competing search engine, says Schmidtlein — and Google is still going to get the most traffic on Apple’s Safari browser even if it’s competing for placement on a non-exclusive basis. Yahoo catches a stray here: “Yahoo can’t even keep its own users” using its search engine, he says.
Mehta asks: what kind of legal framework should he be using to think about generative AI product remedies? “GenAI products are not in the relevant market ... there is no evidence that generative AI products were harmed” by any of Google’s conduct in this case, Schmidtlein says — they didn’t even exist during the period covered in the original trial. “The notion that these AI rivals have been harmed at all is... whatever,” he says. But Google is willing to address the question about forward-looking impact by avoiding bundling or exclusive deals like the ones it used for search. “You can rest assured that at least in that sense it can be addressed.”
Google’s attorney starts by emphasizing all the privacy and competitive problems he sees with the DOJ’s overall proposal, but concedes that some limited data syndication could be feasible. “None of this reverse-engineering” products or “feeding all of our data into an LLM”, Schmidtlein says, but the syndication could help improve search results marginally. (You may recall that Sundar Pichai claims unlimited data sharing could essentially kill Google search.)