Two separate copyright actions landed against OpenAI and Microsoft within days of each
other, and together, they mark a structural turn in AI copyright litigation.
The first development: according to reports, a coalition of approximately 400 local and
regional U.S. newspapers has filed a federal copyright lawsuit against OpenAI and
Microsoft. The coalition is reportedly led by WEHCO Newspapers Inc., whose publications
include the Arkansas Democrat-Gazette and the Chattanooga Times Free Press. The complaint
reportedly alleges that automated tools were used to systematically scrape paywalled
content and strip copyright management information from articles in violation of the DMCA;
specific tool names cited in the complaint could not be independently confirmed. The filing
date is approximately June 24, 2026.
The second development is more precisely sourced, and more analytically significant. On
June 25, 2026, the New York Times applied to submit a third amended complaint in its
copyright lawsuit against OpenAI and Microsoft. According to
36Kr’s reporting, the amended complaint
accuses Microsoft of “actively inducing, assisting, and facilitating large-scale copyright
infringement”, not as a neutral infrastructure provider, but as an active participant. The specific mechanism: a customized supercomputing system containing more than 285,000
CPU cores and 10,000 GPUs, built to enable OpenAI’s AI training operations.
That’s a different legal theory than what’s been at the center of AI copyright litigation
so far. Prior suits focused on whether training on copyrighted data constituted infringement
– a question about what OpenAI did with the data. The NYT’s amended complaint asks whether
the company that built the computers enabling that training should also be liable. That’s
an infrastructure liability theory, and its success or failure will matter well beyond
this specific case.
Verdict
This follow-up to TJS’s June 26 coverage of the NYT amended complaint adds the affirmative
infrastructure claim that wasn’t the focus of that earlier brief.
Don’t expect these suits to resolve quickly. Copyright litigation of this scale and novelty
runs on multi-year timelines. What the suits do right now is shift the risk calculus for
any company in the AI supply chain, not just the model developers, but the infrastructure
providers who make large-scale training possible.
The real question is whether courts will treat purpose-built AI training infrastructure as
meaningfully different from general cloud computing. If Microsoft’s supercomputer is
characterized as a specialized infringement-enabling tool rather than neutral
infrastructure, the contributory liability analysis changes. That distinction will take
years to adjudicate, but compliance and risk teams at hyperscalers and AI infrastructure
providers should be tracking this now, not when a ruling arrives.
What to Watch
OpenAI has reportedly responded with a general statement defending the use of publicly
available training data as fair use. That defense addresses the training data question
directly; it’s less obviously responsive to the infrastructure-as-enablement theory in the
NYT’s amended complaint.
TJS synthesis: The WEHCO coalition and the NYT amended complaint are procedurally
distinct but analytically connected. Both assert that the harm to publishers went beyond
unauthorized access to content, that systems were deliberately built to extract, process,
and strip identifying information from that content at scale. The infrastructure liability
framing is the more novel legal argument, and if courts accept even part of it, the
contractual indemnification structures between AI model developers and their compute
providers will be stress-tested in ways neither party currently anticipates. That’s the
document review that should be happening in legal departments now.