THE TECHNOLOGY BLIND SPOT

On the morning of November 5, 2025, Bloomberg’s Mark Gurman published a number that should have made every antitrust lawyer in Washington set down their coffee. Apple, his sources said, was finalizing a deal to pay Google roughly $1 billion a year for a custom Gemini model, 1.2 trillion parameters, built to run the rebuilt Siri. Thirty days later, Judge Amit Mehta signed the final judgment in United States v. Google, the monopolization case the government had spent five years building around a payment that ran the other way: the $20 billion Google paid Apple in 2022 alone to remain the iPhone’s default search engine.

Hold the two payments side by side. For two decades, Google paid Apple for access to users. Now Apple pays Google for access to intelligence. A second wire opened in the opposite direction while the court was still settling the language of the order written to govern the first.

That reversal is the quiet verdict on the biggest antitrust case in a generation, and no judge issued it. The final judgment polices default placements, licensing terms, and data access in a product category, the general search engine, that the defendant is dissolving in public view. This is not an argument that Judge Mehta got the law wrong. It is an argument that his judgment teaches a drafting lesson every attorney who papers a technology deal needs this week: a remedy aimed at a product category belongs to the defendant, because the defendant decides what category its conduct lives in next.

The Case That AI Rewrote

Start with how the category dissolved the case itself. In August 2024, after a ten-week trial, Judge Mehta ruled that Google had illegally maintained a monopoly in general search services, largely through the default payments. “Google is a monopolist, and it has acted as one to maintain its monopoly,” he wrote. The Justice Department asked for structure: divest Chrome, contingently divest Android, ban the payments, open the data.

Thirteen months later, the remedies opinion gave the government almost none of that. No Chrome divestiture. No Android divestiture. No payment ban. Google must share portions of its search index and user-interaction data with qualified competitors, must strip exclusivity conditions from its default deals, and must answer to a technical committee for six years. Alphabet’s stock rose 8 percent the day the opinion published. Wall Street can read a remedies opinion.

The explanation sits in a single sentence of the 230-page decision: “The emergence of GenAI changed the course of this case.” At the liability trial, no witness testified that generative AI posed a near-term threat to general search engines. Two years later, the same court counted tens of millions of people using “GenAI chatbots, like ChatGPT, Perplexity, and Claude, to gather information that they previously sought through internet search.” By one count, the remedies opinion mentions ChatGPT 28 times and OpenAI 30. The defendant’s most persuasive exhibit was its competitors’ products. OpenAI built Google’s best defense and collected nothing for the work.

Testimony at the remedies trial previewed the migration. In April 2025, Nick Turley, the head of ChatGPT at OpenAI, took the stand and said his company would be interested in buying Chrome if the court forced a sale. The witness box in a search monopolization case had become a stage for AI companies to bid on the monopolist’s assets. Judge Mehta heard that testimony and drew the reasonable conclusion that the wall between the two markets had already failed.

Sit with the full sequence, because its two halves do not fit together. The court declined to restructure a monopolist on the theory that new AI entrants might discipline it. Then it wrote remedies calibrated to the market those entrants were pulling apart.

The Inversion

The judgment’s distribution provisions govern how Google buys placement: covered default agreements must expire after one year and must expressly permit the browser or device maker to promote rival search services and rival GenAI products. Read that last clause again, because it shows the court saw the adjacent category coming and drafted for it. Then notice what every one of those provisions assumes: Google is the party paying for position. The Gemini deal inverts the assumption. Apple pays Google. A custom model licensed to run inside Apple’s assistant is not a default placement Google purchased; it is intelligence Apple bought. Yet the strategic function is identical to the one the court spent five years examining. Google’s intelligence sits in front of every iPhone user, and money changes hands to keep it there. Same two companies. Same function. Reversed wire. No coverage.

To be precise about what reversed: Google still pays Apple for the Safari search default, now on the judgment’s one-year terms. The Gemini deal does not replace that payment. It succeeds it. The old wire buys the default position in an interface both companies are winding down. The new wire buys the equivalent position in the interface replacing it, and the judgment reaches only the first. A court that spent half a decade pricing the value of a default now supervises the less valuable of the two defaults on the table.

Evidence of where the value went arrived on July 22, 2026, when Sundar Pichai told analysts that AI Mode, Google’s conversational search surface, had passed one billion monthly users and had merged with AI Overviews into “one seamless Search experience.” Search advertising grew 17 percent, to $63.3 billion for the quarter. He added a sentence that deserves more attention than the user number: engineering and hardware work had pushed the cost of an AI Mode response to “its lowest level since launch.” The ten blue links are dissolving into an AI answer surface, and Google is telling investors the replacement gets cheaper to serve every quarter. A German court has already ruled that this surface is Google’s own speech, with liability to match. [See When Google Became the Answer, It Inherited the Liability, The Technology Blind Spot (2026).]

Users are following the interface. Pew Research Center measured the shift in 2025: when an AI summary appeared, Google users clicked a traditional result link on 8 percent of visits, against 15 percent when no summary appeared. Publishers experience that arithmetic as vanishing referral traffic. The judgment’s data and syndication remedies assume a web of destinations that search engines send people to. The AI surface keeps the people and dissolves the destinations, which means the remedies subsidize access to a commons the defendant’s own product is draining.

Late Tuesday night, the Justice Department answered. Its appellate team filed a 144-page brief asking the D.C. Circuit to affirm the liability finding and to vacate Judge Mehta’s refusal to ban the distribution payments. The brief concedes the clock problem in one line: “While the court might ‘revisit’ a payment ban later, correcting this error now is important.” Emerging GenAI products pose “a potential threat to Google’s dominance,” the government warns, and the “competitive window may close” while Google spends monopoly profits “to squelch the threat.” Google’s response is due September 29. Oral argument follows in the term that opens this fall. By the time a three-judge panel hears argument about default search payments, the newer payment between the two companies at the center of the case will have been flowing in the opposite direction for nearly a year.

The Strongest Defense of the Judgment

The best case for Judge Mehta’s approach deserves its full weight. He did not miss the AI transition. He drafted for it. The data-sharing remedy defines qualified competitors broadly enough to arm AI entrants, the companies most likely to need a search index they cannot afford to build. The six-year technical committee gives the court standing machinery to adjust as the market moves. And the opinion reserves its trigger explicitly: the court stands “prepared to revisit a payment ban (or lesser remedy) if competition is not substantially restored.” On this reading, the judgment is adaptive equity, built humble because the court knew the market would not hold still. A member of Mehta’s chambers could read this paragraph and confirm every word of it.

But the flaw is not in the machinery. It is in what the machinery watches. Every trigger in the judgment monitors the old market: search defaults, search data, search syndication. The revisit clause fires if competition in general search fails to recover. Nothing in the judgment watches whether the competitive struggle has moved to markets the trial never mapped: custom model licensing, assistant integration, the cost curve of inference. The Apple deal shows the migration in practice. Google did not evade the judgment. It outgrew the nouns the judgment uses.

There is precedent for this exact failure mode, and every lawyer on the docket knows it. The 2001 Microsoft consent decree policed operating system licensing and browser bundling. It expired in 2011, supervised to the end, technically successful, and strategically beside the point, because the fight had moved to phones and search, two markets the decree never named. One of the companies that grew up in that decree’s shadow was Google. The company now running the sequel studied the original.

The Microsoft comparison has a limit worth naming. Scholars credit that decree with holding the door open for the next generation of competitors, Google among them, so adaptive relief can matter even when it polices yesterday’s category. The comparison holds for the drafting lesson. It does not support a claim that Mehta’s judgment accomplishes nothing, and the data-sharing provisions may yet arm the competitor that matters. The point is narrower and harder: whatever the judgment accomplishes, it will accomplish in a market its own defendant is winding down.

This Is a Drafting Problem, and You Have It Too

Remove Google from the story entirely. A settlement that bars your client’s former partner from “operating a competing retail website” does not reach the partner’s app. A noncompete defined by “the software product known as X” expires the day the counterparty renames the product and ships it as something else. A protective order drafted for “documents” meets an AI system that ingests, transforms, and regenerates. In each case the defect matches the one sitting in the Google judgment: the operative terms bind a noun the counterparty controls rather than conduct the counterparty performs.

Your vendor stack carries the same seam. The AI addendum your firm negotiated last year defines the tool by product name and version, and the indemnities, data-use limits, and confidentiality commitments all attach to that noun. When the vendor retires the product and migrates your data into its new agentic platform, as vendors routinely do after acquisitions, whether your negotiated protections travel with it depends entirely on whether you bound the vendor’s conduct or its branding.

Andy Grove had a name for what the judgment collided with. A strategic inflection point, he wrote in Only the Paranoid Survive, is the moment the forces acting on a business change so fundamentally that the old categories mislead everyone still using them. The final judgment in United States v. Google speaks the old categories. The wire transfers between Cupertino and Mountain View speak the new ones.

Before Thursday

The Google docket will resolve itself without your help. Your own definitions sections will not. This week, pull the operative definitions from the most recent technology-touching agreement you negotiated: a settlement, a stipulated injunction, a vendor contract, a noncompete. Read each defined term and ask one question: does this bind what the counterparty does, or what the counterparty currently calls it? Mark every term that names a product, a platform, or a category. Each one is a term the counterparty can exit by shipping something new. Then redraft the single most consequential one as conduct: not “the Platform,” but “any system, however named or delivered, that performs the function.” The counterparty’s lawyers will resist that language, and the shape of their resistance will tell you exactly what they plan to ship next.

If you litigate with or against technology companies, add one calendar entry: the D.C. Circuit argument this fall. Not for the outcome. For the transcript, because the panel’s questions will reveal whether the next generation of technology remedies gets written in nouns or in verbs, and every tech settlement you draft after that opinion issues will live downstream of the answer.

Gurman’s sources handed him a number, and the number told the story the 230-page judgment could not: by the time the court finished describing the market, the defendant had finished leaving it. The Justice Department will spend the fall asking three judges to strengthen the description. Somewhere in your own files sits a definitions section with the same defect, binding a noun some counterparty has already scheduled for retirement. Which one?

About the Author

JD Morris is Co-Founder and COO of LexAxiom, an Agentic AI platform for the business of law. Over a 25-year career, he has built and scaled enterprise technology products across Dell, EMC, VMware, and Cisco, including the first exabyte eDiscovery platform. He holds dual MBAs from Columbia Business School (Finance) and UC Berkeley Haas (Marketing), a Master of Legal Studies in Cybersecurity Law from Texas A&M, and a Master of Engineering from George Washington University. He writes The Technology Blind Spot on the intersection of emerging technology and law. Connect with him on LinkedIn at www.linkedin.com/in/jdavidmorris, on X at @JDMorris_LTech, or on Bluesky at @JDMorris-ltech.bsky.social.

References

1. United States v. Google LLC, 747 F. Supp. 3d 1 (D.D.C. 2024).

2. United States v. Google LLC, 803 F. Supp. 3d 18 (D.D.C. 2025) (remedies opinion).

3. United States v. Google LLC, No. 20-cv-3010 (APM) (D.D.C. Dec. 5, 2025) (final judgment).

4. Brief for the United States, United States v. Google LLC, Nos. 26-5023, 26-5047, 26-5049 (D.C. Cir. filed July 28, 2026).

5. Mark Gurman, Apple Plans to Use 1.2 Trillion Parameter Google Gemini Model to Power New Siri, Bloomberg (Nov. 5, 2025), https://www.bloomberg.com/news/articles/2025-11-05/apple-plans-to-use-1-2-trillion-parameter-google-gemini-model-to-power-new-siri.

6. OpenAI Would Buy Google’s Chrome Browser, ChatGPT Chief Says, Bloomberg (Apr. 22, 2025), https://www.bloomberg.com/news/articles/2025-04-22/openai-would-buy-google-s-chrome-browser-chatgpt-chief-says.

7. Cristiano Lima-Strong, How AI Upended a Historic Antitrust Case Against Google, TechPolicy.Press (Sept. 3, 2025), https://www.techpolicy.press/how-ai-upended-a-historic-antitrust-case-against-google/.

8. Ryan Knappenberger, DOJ Urges DC Circuit to Beef Up Remedies in Google Monopoly Ruling, Courthouse News Serv. (July 29, 2026), https://www.courthousenews.com/doj-urges-dc-circuit-to-beef-up-remedies-in-google-monopoly-ruling/.

9. Sundar Pichai, Alphabet Earnings Call Q2 2026: Sundar Pichai Remarks, Google: The Keyword (July 22, 2026), https://blog.google/company-news/inside-google/message-ceo/alphabet-earnings-q2-2026/.

10. Alphabet Inc., Alphabet Announces Second Quarter 2026 Results (July 22, 2026), https://s206.q4cdn.com/479360582/files/doc_financials/2026/q2/2026q2-alphabet-earnings-release.pdf.

11. Google Stock Jumps 8% After Search Giant Avoids Worst-Case Penalties in Antitrust Case, CNBC (Sept. 2, 2025), https://www.cnbc.com/2025/09/02/google-antitrust-search-ruling.html.

12. Google’s Payments to Apple Reached $20 Billion in 2022, Antitrust Court Documents Show, Gulf News (May 2, 2024), https://gulfnews.com/technology/companies/googles-payments-to-apple-reached-20-billion-in-2022-antitrust-court-documents-show-1.1714660457987.

13. Andrew S. Grove, Only the Paranoid Survive (1996).

14. Landgericht München I [Regional Court of Munich I], May 28, 2026, Az. 26 O 869/26 (Ger.).

15. Athena Chapekis & Anna Lieb, Google Users Are Less Likely to Click on Links When an AI Summary Appears in the Results, Pew Rsch. Ctr. (July 22, 2025), https://www.pewresearch.org/short-reads/2025/07/22/google-users-are-less-likely-to-click-on-links-when-an-ai-summary-appears-in-the-results/.