Can the New York Times Save Journalism From Our AI Overlords?
In his July 28, 2026 WIRED interview, A.G. Sulzberger put the stakes in blunt terms: platforms “have taken it all”, including journalism, books, and music, and built massive AI products on top of others’ work. That line frames the legal and strategic posture the Times is pursuing right now: sue, litigate, and force a public debate about who pays for the raw materials of AI.
Where the litigation stands
The New York Times sued OpenAI and Microsoft in late 2023, alleging the companies used Times journalism to train large language models without permission. The paper filed an amended complaint on June 25, 2026 that narrowed some claims (it dropped a secondary‑infringement theory against OpenAI and modified a claim against Microsoft). Those filings are public and show the case is active in the U.S. District Court for the Southern District of New York.
Sulzberger told WIRED the Times has spent “well over $20 million” on the litigation and that the case, filed roughly two and a half years ago, remains in discovery. Treat those figures and the discovery timeline as the company’s statements until they are corroborated in company disclosures or the court docket.
“They’ve just taken it. They’ve just taken it all. Without permission and without any compensation, because they thought they could.”, A.G. Sulzberger
What Sulzberger means by the “original sin of AI”
His shorthand is a simple economics claim: modern generative AI products require four inputs: coders, compute, electricity, and content. In Sulzberger’s view the unpermitted taking of content, like journalism, books, and music, is the foundational wrong. Enforcing copyright, he says, is the practical defense of journalism’s business model and civic role: “Your rights only hold if you seek to enforce them.”
That idea drives the Times’ willingness to litigate. It also explains the paper’s public message: the case is framed as “anti‑theft, ” not anti‑technology. Sulzberger said, “It is not anti‑tech. It is anti‑theft.”
What the evidence says about traffic and referral economics
Publishers have lost meaningful referral traffic as search engines surface AI-generated summaries or “overviews” directly in results. Columbia Journalism Review, citing Similarweb analytics and Pew Research Center studies, estimated about a 15% decline in worldwide search referral traffic tied to AI summaries. Pew’s July 2025 analysis found links embedded in those AI summaries were clicked roughly 1% of the time. Those figures are estimates across publishers, not an audit of the Times’ traffic.
Sulzberger pointed to scale as the Times’ defense. He said the company has “surpassed 13 million subscribers” and is “aiming for 15, ” and that the newsroom has “basically doubled” to “2, 400” journalists. The Times also described major foreign deployments, with reporters in more than 150 countries last year and “more than 70 journalists” in Ukraine. Treat these as the company’s statements until they are confirmed in investor filings or newsroom fact sheets.
How the Times uses AI internally, and where it draws lines
The Times draws a clear line between training models on journalism without permission, which is the conduct at issue in the lawsuit, and using AI as an internal reporting tool. Sulzberger and newsroom leaders describe practical uses like pattern detection, image triage, translation, and other filtering tasks that speed reporters’ work. As one Times reporter who worked on an AI‑supported, Pulitzer‑winning project put it, “You have to think of AI as a filter.”
“Humans are responsible for everything we produce, period, ” Sulzberger said, an explicit insistence that AI remain assistive rather than authorial in the newsroom. That workflow ethic has two payoffs: better editorial control and, arguably, fewer legal risks than publishing AI‑generated content without human oversight.
Legal counterpoints and the unsettled law
The Times’ position is one credible legal frame, but it is not settled. Defendants in training‑data suits typically raise fair‑use and transformative‑use defenses, arguing that data ingestion for model training can be non‑infringing depending on use and transformation. Courts have reached different conclusions in related cases, and the SDNY litigation will hinge on factual discovery about what content was used and how.
The June 25, 2026 amendment that dropped a secondary‑infringement theory against OpenAI is a reminder that pleadings and litigation strategy evolve. Discovery could materially affect the case’s trajectory. Expect judges to wrestle with technical questions about dataset composition, model behavior, and whether downstream answers produced by models recreate protected expression.
Press‑freedom pressure and subpoenas
Sulzberger also raised a separate but urgent threat to journalism: government pressure on reporting. He described Department of Justice subpoenas tied to reporting on a Qatar‑donated jet as “a gross overreach” and called the current pattern of government actions “the biggest effort to go after the press that we’ve seen from a White House probably since Woodrow Wilson.” Those are his characterizations. Verify subpoena details and legal posture in court filings and Times reporting before treating them as settled fact.
What executives and news leaders should prioritize now
If your enterprise depends on content, attention, or the public value of reporting, treat the Times’ fight as both a warning and a playbook. Below is a prioritized, time‑boxed checklist you can operationalize.
-
0-90 days, Audit and prioritize (Owner: Head of Content/Product)
Run a 90‑day crawl of your top 1, 000 pages by referral traffic. Tag pages that drive subscription signups, ad revenue, or partnership value. Identify the top 10% of pages that, if siphoned by answer engines, would materially hurt revenue.
-
90-180 days, Harden direct relationships (Owner: Head of Growth)
Boost subscription funnels, A/B test newsletter and account paywalls, and reduce single‑click referral dependence. Launch at least one product experiment to convert high‑value referrers into direct accounts.
-
180-365 days, Negotiate, pilot licensing, and legal posture (Owner: Legal/Business Development)
Explore pilots to license content to platforms with tightly scoped APIs and measurable usage terms. If licensing is impractical, join trade groups or pooled bargaining efforts to share legal risk and negotiating leverage.
-
Ongoing, Use AI with guardrails (Owner: Editor/CTO)
Adopt AI as a filter for discovery and triage, require human verification for outputs, and publish transparent usage notes where AI materially assisted reporting or audience content.
For smaller publishers that can’t litigate at scale, pooled strategies matter: trade associations, local press alliances, or pooled licensing funds can amplify negotiating power. Look to groups such as local press associations, publishing trade groups, and copyright advocacy organizations to coordinate action.
Short, honest answers leaders are asking
-
Is The New York Times actually suing OpenAI and Microsoft?
Yes. The Times filed suit in late 2023 alleging unauthorized use of its journalism; the paper amended that complaint on June 25, 2026, narrowing some claims.
-
Has search‑AI genuinely reduced traffic to publishers?
Industry analyses suggest significant effects: Columbia Journalism Review, citing Similarweb and Pew Research Center, reported an estimated ~15% decline in search referral traffic tied to AI summaries, and Pew’s July 2025 study found links in those summaries are clicked roughly 1% of the time. These are estimates across publishers, not precise measures for any single outlet.
-
Are the Times’ staffing and spending figures verified?
Sulzberger said the Times has spent “well over $20 million” on the litigation and provided subscriber and newsroom figures during the interview; treat those as the company’s statements until confirmed via investor materials or official disclosures.
-
Can AI be used ethically in newsrooms?
Yes, when used as a filter and paired with human validation. That reduces drudgery and expands reporting reach, but it isn’t a legal cure‑all for training‑data disputes or a replacement for editorial judgment.
-
What should small publishers do if they can’t afford litigation?
Prioritize direct reader revenue, join collective bargaining or licensing efforts, and collaborate with peers through trade groups to share legal and commercial strategies.
How to read the fight, and what happens next
The Times’ lawsuit forces a public reckoning over whether platforms that intermediate answers must pay for the content they ingest. Courts, regulators, and markets will all shape the outcome. Litigation will clarify some legal rules. Legislative or regulatory responses could create market structures for licensing. Platforms might pursue limited deals with publishers. The June 25, 2026 amended complaint shows the legal posture will shift as facts emerge.
For business leaders outside media, the lesson is clear. Platforms that intermediate content will face pressure to license inputs, compensate creators, or confront regulatory and market pushback. If your business depends on referral traffic, curated answers, or third‑party discovery, treat the shifting economics as a strategic risk. Audit your exposure, strengthen direct channels, and clarify ownership of your content now.
“At the end of the hardest days, what makes me feel like it’s all worth it…, reading the paper. Just to remind myself, ‘Oh, this is what I’m fighting for.’”, A.G. Sulzberger
The Times can’t unilaterally “save” journalism. What it can do, by suing, negotiating, and amplifying the debate, is force the rules of the road into public view. Expect years of litigation, licensing experiments, and shifting user behavior. If you care about trusted reporting or rely on the attention it creates, now is the time to audit exposure, strengthen direct channels, and join peers in shaping the next marketplace for content.