Google uses search remedies trial to subpoena OpenAI, Perplexity and Microsoft over their generative AI efforts

By Marty Swant  •  November 27, 2024  •

Ivy Liu

For the remedies phase in the Google search antitrust trial, the giant has subpoenaed three of its biggest AI rivals — OpenAI, Perplexity AI, and Microsoft.

The subpoenas, sent in October, were made public on Monday through legal filings from all four companies. The disclosures came just hours after the Google ad-tech antitrust trial wrapped up closing arguments from lawyers representing Google and the U.S. Department of Justice.

When asked for comment, Google and Microsoft did not immediately respond to Digiday and spokespersons for OpenAI and Perplexity declined to comment. (All three companies are also facing a range of other lawsuits related to issues like AI copyright infringement.)

The new legal filings also shed light on what Google hopes to uncover by gathering fodder from rivals: To convince the court that its search monopoly facing more competition with the emergence of AI-powered search. Many of the materials Google is asking for relate to rivals’ efforts around search and advertising.

Here’s a brief rundown with some examples of what Google’s subpoenas ask each company to provide:

  • OpenAI: Usage data for ChatGPT, distribution agreements with third parties, agreements with Microsoft related to Bing Search API, any agreements with Perplexity, and types of data used to train AI models. Google is also asking for content licensing agreements and board minutes related to OpenAI’s plans for search distribution and any advertising strategies.
  • Perplexity: Download totals, active user data, financial performance, content licensing agreements, info about AI model training data and info about Perplexity’s monetization strategy including advertiser pitch decks and plans to integrate ads around user queries. Google is also asking for any communication between Perplexity and plaintiffs in the case — the DOJ and Colorado’s attorney general — and documents about Perplexity’s view of the competitive search landscape. Google also asked for agreements and communication related to distribution deals including marketing, pre-installations, syndication, negotiation contracts.
  • Microsoft: Agreements between Microsoft and other companies like OpenAI and Perplexity, documents for the data types and quantities used to help train OpenAI’s models, and documents related to using generative AI tools to power search results and advertising services. Google also asked for info about content licensing agreements — including “exclusive content deals” — and materials about ways Microsoft provides search results to third-party AI tools and search tools like ChatGPT and ChatGPT Search. “If new artificial intelligence tools and technologies have provided Bing a novel mode of distribution — and all indications are that they have — Google is entitled to present that evidence,” Google wrote.

The subpoenaed companies differ in their legal responses to the requests. After initially claiming Perplexity never received a subpoena, that company’s lawyers said they now plan to send objections by next month and begin producing documents by early January. OpenAI has agreed to some of the requests — including agreements for Bing Search API, ChatGPT usage and board minutes. However, lawyers objected to other requests like financial performance and training data as being being “extremely burdensome” because of “core trade secrets and confidential information.” Lawyers for Microsoft said the company will provide documents for all but four of the 41 requests, with objections related to confidential agreements with OpenAI, Perplexity, Inflection and the UAE-based AI company G42.

“Microsoft has agreed to conduct a reasonable search for and produce documents with respect to five out of the six requests identified in Google’s Statement,” Microsoft lawyers wrote. “Where Microsoft has indicated it will not produce documents, it has done so based on good-faith objections to relevance, burden, and lack of proportionality. As much as Google wants to cast Microsoft as a party to this Action, it is not one.”

Even if Google is forced to sell Chrome, it could take months or years of legal fighting before it happens. The drawn out process could give Google time to build a new user for its latest standalone app for Gemini, which directly competes with the likes of ChatGPT and Perplexity. However, some ad-tech execs think the startups should get a chance to compete in the AI era of search.

“If [Google] is able to hang on to this distribution chokehold they have over search, that will give them time to continue to improve Gemini and hopefully — at least from [Google’s] perspective — put other competitors out of business,” said Adam Epstein, co-CEO and president of adMarketplace, a search advertising company.

It could take years for startups like Perplexity and OpenAI to gain market share, said Evelyn Mitchell-Wolf, a senior analyst with eMarketer. She also noted it’ll take a while for platforms like Perplexity to scale its advertising business. For now, she thinks it’ll be too difficult to dethrone Google while traditional search is ubiquitous and too easy. However, she thinks Google would have a “nightmare scenario” if it’s forced to sell Chrome.

“If we’re talking about market dynamics and search ads, I don’t expect Perplexity to gain meaningful scale anytime soon since it’ll take time for advertisers to see results,” Mitchell-Wolf said. “Having more options to search is good for consumers, but we’re still in a period when the underlying effects of AI on consumer behavior remains to be seen because we’re still in the big push from these major AI players.”

The DOJ’s proposal for Google to sell Chrome came just a week before closing arguments in the ad-tech antitrust trial, which took place on Monday in federal court ahead of a ruling in the coming months.

While Google’s search trial and ad-tech trial are separate, their potential outcomes remain the subject of conjecture as opposed to clear-cut visions. However, experts have said Google’s dominant search engine has provided a flywheel for much of the data used for ad-targeting while its huge user base gives advertisers a way to reach a high number of potential customers.

Some think the industry is paying closer attention to the search remedies trial scheduled for March. Among the experts following both trials is Ari Paparo, a longtime adtech exec and co-founder of Marketetecture Media, who said the DOJ’s proposal for Chrome went “way past the red line on things and [Google] won’t want that to happen.”

“I’ve asked multiple lawyers, and no one understands what the remedy phase of this [ad-tech] trial is going to be like,” Paparo said. “Is it going to be a separate hearing? Is it going to be a request for documents? Or are they just going to haul off and write a remedy? Nobody knows what the remedy is going to be.”

https://digiday.com/?p=561774

More in Marketing

Note: This article have been indexed to our site. We do not claim legitimacy, ownership or copyright of any of the content above. To see the article at original source Click Here

Related Posts
iPhone subreddit starting protest early, going dark indefinitely thumbnail

iPhone subreddit starting protest early, going dark indefinitely

Following Reddit announcing its new API costs for developers that are killing third-party apps like Apollo, one of the most popular Apple-related subreddits has announced it will be starting its protest early. Going further, the mods have shared that it won’t just be limited to a few days, they plan for the iPhone subreddit to
Read More
WebSpec, a formal framework for browser security analysis, reveals new cookie attack thumbnail

WebSpec, a formal framework for browser security analysis, reveals new cookie attack

Folks at Technische Universität Wien in Austria have devised a formal security framework called WebSpec to analyze browser security. And they've used it to identify multiple logical flaws affecting web browsers, revealing a new cookie-based attack and an unresolved Content Security Policy contradiction. These logical flaws are not necessarily security vulnerabilities, but they can be.…
Read More
Best Pet Supplies and Tips for Newly Adopted Cats and Dogs thumbnail

Best Pet Supplies and Tips for Newly Adopted Cats and Dogs

I spoke with staff from Muttville Senior Dog Rescue in San Francisco—the shelter that WIRED art director Elena Lacey has fostered pups from. They offered a ton of advice and a peek at their take-home paperwork (you can find copies here) that can apply to all new pets, whether they're babies or seniors. The shelter…
Read More
Physicists propose new magnetron technology to control qubits without expensive and complex lasers thumbnail

Physicists propose new magnetron technology to control qubits without expensive and complex lasers

来自俄勒冈大学的两名物理学家,刚刚概述了控制量子计算构建块的新技术,推动此类计算机向着更加准确和潜在效用迈出了重要一步。作为离子实验室的创始人,David Allcock 与 David Wineland 也是新论文的 12 位作者之一。文章基于科罗拉多州博尔德市国家标准与技术研究所(NIST)的一项实验,且两位科学家都曾在这里工作。 资料图(来自:IBM Q System)2018 年的时候,两人来到了俄勒冈大学、并继续在该项目上开展合作。在《自然》杂志上一篇文章中,研究团队详细介绍了在量子计算与模拟中使用俘获量子比特的相关技术。物理学家表示,尽管新技术有助于推动操作方面的改进,但量子计算机仍会产生太多计算错误、使之难以成为一款有效的工具。David Allcock 指出:问题出在量子计算机的逻辑门上 —— 作为用于在计算机中执行基础逻辑功能的工具 —— 其表现着实相当糟糕。它们在大约 1% 的时间里会失败,只需大约 100 次操作,就会带来垃圾数据。目前整个量子计算领域都处于早期阶段,且受制于错误,我们无法在机器上开展冗长的计算、或有价值的模拟。而我们的目标,就是在没有错误的情况下开展上万次操作,然后添加多层检查、以在发生错误时进行修复。研究配图 - 1:实验设置Wineland 表示,被困住的离子,就像是一碗具有一定磁性的弹珠。Allcock 补充道,物理学家们可借助不同的方法对离子施力,比如相当昂贵且复杂的激光器。相比之下,基于磁力方案的逻辑门不仅更加实惠、实用,且能够直接在集成电路上生成。在这项新研究中,他们的主要工作,就是让新技术达到与之前任何人做过的逻辑门一样有效工作的水平。目前以 Google 和 IBM 为代表的商业科技巨头,正好拥有大量工程师可解决这方面的问题。而学术物理学家们,正在试图证明有更好、更基础的技术来攻克难关。研究配图 - 2:纠缠操作的鲁棒性Allcock 指出:“我们已经证明了大家可借助技术上更简单的方式来做到这一点”。Wineland 亦表示,若物理学家和工程师们能够使量子计算机足够可靠、并能够以足够大的容量运行,他们就能够开展其它系统模拟。例如,量子计算机可模拟药物治疗中使用的分子的作用,而无需在实验室中辛苦合成。研究配图 - 3:纠缠态保真度分析即便如此,Wineland 还是相当谦虚地表示:“目前已有一些非常实用、有用的结果,但我们也只是触及到了该领域的一些皮毛”。有关这项研究的详情,已经发表在近日出版的《自然》(Nature)杂志上,原标题为《High-fidelity laser-free universal control of trapped ion qubits》。
Read More
Index Of News