AI Governance Institute
← News

Wikimedia's NYT v. OpenAI Brief Draws a Copyright Line AI Buyers Must Track

What happened

The Wikimedia Foundation filed an amicus brief in the ongoing New York Times v. OpenAI and Microsoft copyright lawsuit. The foundation argued that pre-training large AI models on copyrighted works is transformative and should qualify as fair use under US copyright law. Wikimedia drew a firm line at output. When an AI system reproduces copyrighted text verbatim in response to a user query, that reproduction should require a license from the rightsholder. The foundation also warned that an overbroad ruling against pre-training could damage open-web knowledge ecosystems and constrain open-source AI development. The brief adds a prominent institutional voice to a significant case. It has already surfaced internal communications suggesting OpenAI and Microsoft were aware of potential legal exposure from their scraping practices.

Why it matters

  • ·The two-part framework Wikimedia proposes, permissible training versus infringing verbatim output, is not yet settled law, but it signals how courts may eventually draw the line. Enterprises that deploy AI tools generating large volumes of text could face copyright liability even when using models from vendors with broad training data practices.
  • ·AI procurement due diligence programs currently lack reliable tools to verify whether a vendor's model produces verbatim reproductions of copyrighted material at scale. If the Wikimedia framework is adopted, that gap becomes a direct legal exposure requiring contract protections and output monitoring.
  • ·A ruling that treats verbatim AI output as requiring a license would reshape vendor indemnification negotiations across the industry. Compliance teams that have not reviewed AI vendor contracts for copyright liability allocation face elevated risk. This includes indemnification scope and output-use restrictions, particularly if the case resolves against the output-generation model.

Stay ahead of stories like this

Get every US AI governance development like this one, plus the rest of the week's developments. Every Thursday.

Free every Thursday. Unsubscribe anytime.

Governance controls affected

What to do now

  • ☐Ask your AI vendors whether their contracts include indemnification for copyright claims arising from model outputs, not just training data, and flag any that do not.
  • ☐Review your current AI output monitoring processes to determine whether they can detect verbatim reproduction of third-party copyrighted text at meaningful scale.
  • ☐Confirm whether your AI procurement due diligence process asks vendors how they handle requests that return near-identical copies of source material, and update vendor questionnaires if it does not.
  • ☐Work with legal counsel to assess whether any existing AI deployments involve use cases, such as summarization, content generation, or search, where verbatim reproduction is a plausible output, and document that assessment now before a ruling changes the risk profile.
  • ☐Monitor the docket in New York Times v. OpenAI and Microsoft for trial scheduling and preliminary rulings, and set a trigger to revisit vendor contracts and output policies if the court adopts a verbatim-reproduction standard.

What to watch next

The New York Times v. OpenAI and Microsoft case is likely to produce significant rulings on fair use and AI output liability within the next twelve to eighteen months. Compliance teams should watch for a court ruling on the training-versus-output distinction Wikimedia has now formally placed before the court. Anthropic's $1.5B copyright settlement has already shown that training data provenance carries real financial exposure. A court-drawn line on verbatim outputs could add a second category of liability. Congress has not yet moved on AI and copyright. However, a court's dismissal of Google AI Overviews suits pointing publishers to Congress suggests that legislative action remains possible if courts decline to extend existing copyright doctrine to AI outputs.

Related Coverage

Research2026-11-18

Expired Domain Hijack Altered Kokoro AI Model Weights in 87 Seconds

Security researchers took control of the Hugging Face publisher account for Kokoro, a text-to-speech model with nearly 13 million monthly downloads. They did this by registering its publisher's expired domain and resetting the account password. They then demonstrated that a model's neural weights could be altered in 87 seconds using a consumer graphics card, producing different audio output without changing any application code. The researchers did not upload the modified weights, but found that Hugging Face would not have flagged the modified release as suspicious.

Enforcement2026-11-17

New Commerce Rule and Blackwell Chip Findings Put Open-Source AI Supply Chains at Risk

A new US Department of Commerce rule announced November 16, 2026, targets remote access to Nvidia chips via third-country data centers. This workaround has powered China's open-source AI models. Separately, unverified research by former OpenAI researcher Will DePue claims Chinese AI companies had substantially greater access to restricted Nvidia Blackwell chips than previously estimated. Together, these developments create new export control compliance questions for enterprises that source open-weight models trained in China.

Research2026-11-16

Alpha School AI Misses Math Target for Half of K-8 Students

Internal data from Alpha School show that only 42% of students met the network's core promise of learning twice as fast in both math and reading. Former employees report the AI generated faulty lesson plans acknowledged internally to sometimes do 'more harm than good,' and that the company scraped third-party course content without permission. Alpha's platform now reaches more than 2,600 public school students in Texas and South Carolina.