Two people copy copyrighted material without the rights-holder's direct permission. One has no legal defense available at all, regardless of intent. The other gets a genuine day in court. That's the real asymmetry underneath a common complaint — not that corporations always win, but that the law gives them a fact-specific hearing individuals never get in the first place.

Ripping a DVD you own to watch on another device means breaking CSS, the disc's access-control encryption — and under the DMCA's anti-circumvention provision, that's the violation, entirely separate from whether the underlying copying would itself be fair use. The Second Circuit upheld this directly in Universal City Studios v. Reimerdes (2001): the act of circumvention is unlawful whether or not the resulting use would be infringing, and personal use is not a defense.[1] A person who owns the disc, has no intent to share it, and only wants to watch their own purchase on a device that can't read the original still has no legal cover if they break the encryption to do it.

Region coding is a separate case from DRM circumvention, and it's worth keeping distinct — but it's the same underlying mechanism this whole piece is actually about: a cost the rights-holder created, paid by someone else. A DVD bought legally in Tokyo, played on a real Japan-market player bought specifically to read it, involves no circumvention at all — the player reads the disc exactly as its region coding intends, no encryption defeated, unambiguously legal. But the two players themselves are functionally the same hardware — the same laser, the same disc-reading mechanism, reading the same physical format. Region-locking isn't a technical requirement of the format; it's a firmware restriction, and manufacturers implement it because the CSS licensing terms that let them decrypt DVDs at all require it. Nothing about the technology needed a second device to exist. The requirement is a business decision made by the studios that license the format — and the actual cost of that decision, a second piece of otherwise-redundant hardware, lands entirely on the person who already paid once for a disc they're legally entitled to watch.

Now the corporate side of the same general behavior — copying copyrighted material without a license to build something new. Anthropic settled Bartz v. Anthropic for $1.5 billion in August 2025, the largest publicly reported copyright recovery in US history, covering roughly 500,000 books downloaded from pirate libraries LibGen and PiLiMi and kept in the company's training corpus.[2] That is real accountability, not a company getting away with it — the number is the largest in the history of American copyright litigation.

But the reason it happened is the precise point, and it's not the one the settlement headline suggests. Judge Alsup ruled that training an AI model on the plaintiffs' books was itself fair use — a real, litigated legal conclusion, not a default assumption. What the ruling did not excuse was downloading the books from pirate sites instead of acquiring them legitimately. Anthropic lost on the sourcing, not on the training.[2] An individual doesn't get a comparable two-part inquiry for breaking DRM — there's no version of "the underlying use was fine, only the method was the problem" available to them at all.

The two other major AI-copyright fights running at the same time show a genuinely mixed, not one-sided, record. Getty Images lost the core of its case against Stability AI in the UK High Court in November 2025 — the court rejected Getty's central copyright claim over Stable Diffusion's use of scraped Getty images, though Getty won a narrower trademark point and has since been granted leave to appeal.[3] The New York Times' ongoing suit against OpenAI and Microsoft is still active — and as recently as September 2, 2026, the Justice Department formally filed in support of OpenAI and Microsoft, arguing that the success of the American AI industry is itself a national security interest.[4] That's not corporations quietly winning by default. That's the federal government taking an explicit legal position that a copyright holder's claim should yield to an industrial-policy interest — a position no individual DMCA defendant has ever had available to them.

Why does this matter? The sharper version of the complaint isn't "corporations get away with it and individuals don't" — Anthropic's $1.5 billion settlement and Getty's UK loss both cut against that reading. The sharper version is that the law gives corporations a real, fact-specific fair-use fight — one with actual variables, actual court dates, actual chances to lose big or win outright — while it gives an individual circumventing DRM for personal use no fight at all, just a bright-line violation regardless of what they intended or whether anyone was harmed. One side gets litigated nuance. The other gets strict liability. That gap, not any single company's outcome, is the actual asymmetry.