A $1.5 Billion Settlement That Caught My Attention
As an attorney, I handle a substantial number of copyright
infringement cases at my firm. I've represented copyright holders in
negotiating settlements against businesses that built AI-powered services by
feeding in other people's copyrighted work without permission and then profited
from it. Given that background, the recent news that a U.S. court gave final
approval to Anthropic's $1.5 billion copyright settlement really caught my
attention, for a number of reasons.
What Happened: The Case at a Glance
Here's a quick rundown of the case, based on recent reporting:
•
In 2024, authors Andrea Bartz,
Charles Graeber, and Kirk Wallace Johnson, along with several publishers, sued
Anthropic, claiming the company used their books without permission to train
its AI model, Claude.
•
In June 2025, the U.S. District
Court for the Northern District of California found that training an AI model
on lawfully purchased books could qualify as fair use.
•
However, the court also found
that Anthropic's downloading of roughly 7 million pirated books to build its
internal library could not be considered fair use.
•
A separate trial on damages for
that piracy was set to follow. To avoid the risk of a damages verdict,
Anthropic proposed a settlement of about $1.5 billion (roughly 2.2 trillion
Korean won), which the court granted preliminary approval in September 2025.
•
In July 2026, Judge Araceli
Martinez-Olguin of the Northern District of California rejected objections that
the settlement amount was too low and granted final approval of the roughly
$1.5 billion settlement.
(Source: Intellectual Property NEWS, “U.S. Federal Court Approves
Anthropic's $1.5 Billion Copyright Settlement,” Issue No. 2026-31, August 4,
2026.)
Reading this, four things stood out to me — and I want to walk through each one.
Two Things That Struck Me as a Korean Lawyer
First, the sheer size of the damages. To be candid, copyright
damages in the U.S. tend to run far higher than what we typically see in Korea.
If you divide the roughly 2.2 trillion won settlement by the 7 million
infringing books, that works out to about 310,000 won per book. Compared to the
domestic cases I've handled, that's roughly three to nine times higher. Having
personally represented copyright holders in settlement negotiations against
Korean businesses that trained AI on unauthorized content, I feel that gap even
more acutely.
Second, how much the process limits what the parties themselves can
do. In this case, the court granted preliminary approval to the settlement
Anthropic itself proposed, and then, when the copyright holders objected that
the amount was too low, the court overruled them and granted final approval
anyway. Watching this kind of settlement process — quite different from what we
have in Korea — made me wonder whether it leaves too little room for the
parties' own autonomy. I understand that the American class action system is
designed to deliver fast, efficient relief to large groups of rights holders.
Still, a structure where individual objections simply don't carry much weight
feels like a real trade-off, separate from questions of procedural efficiency.
The Fair Use Line That Should Worry More Than Just
Anthropic
What I found most interesting about this ruling is the distinction
the court drew: training an AI on legally purchased books may qualify as fair
use, but downloading pirated books and storing them in a library does not. In
other words, the court treated the act of downloading and storing content
differently from the act of actually using that content to train a model. If
Korean courts applied the same standard, a lot of people here could find
themselves exposed to copyright liability. From what I've seen in practice,
downloading and keeping illegally distributed material is, unfortunately, a
fairly common habit in Korea. That's precisely why, in my own cases, I haven't
pursued individuals who merely downloaded and kept such files — I've focused on
representing copyright holders against the people who distributed the material
publicly in the first place.
That said, this ruling gives me real pause as someone who represents
copyright holders for a living. In practice, the people who actually get sued
are almost always the original uploaders; the far larger group of people who
download that material and reuse it in other ways is rarely targeted. Seen from
the uploader's side, it can understandably feel like bad luck — “why me, when
so many others did the same thing and reused the content besides?” Korea hasn't
yet seen many cases that treat downloading and reuse itself as the problem. But
this U.S. ruling suggests that such conduct may not qualify as fair use either,
meaning it could amount to copyright infringement. That's a signal worth taking
seriously — similar issues may well surface in Korean courts down the road.
There aren't many Korean law firms that handle copyright litigation
as a core practice area, and mine has built up considerable experience in this
field over the years. If you have questions about any of the issues raised
here, please feel free to reach out by email anytime.
