Delhi Cleared OpenAI on Copyright the Same Week a US Court Cited Section 230 to Block Texas
I expected the Delhi High Court’s ruling on ANI versus OpenAI to go the other way. News agencies have made real headway in other jurisdictions arguing that training on their reporting is itself the infringement, and I assumed an Indian court would lean toward the plaintiff on a fact pattern this sympathetic.
It didn’t, and the reasoning it used is the more interesting story than the outcome.
The Delhi High Court ruled that OpenAI’s use of ANI content to train ChatGPT wasn’t copyright infringement, because ANI didn’t show that ChatGPT actually reproduced its reports. In the same week, a US appeals court narrowly blocked Texas from enforcing its child-safety monitoring law, on the grounds that Section 230 preempts the state from imposing it. Different countries, different statutes, different subject matter entirely — and yet both courts arrived at their result by asking not “was the underlying conduct harmful” but “does the specific legal mechanism reach this specific conduct at all.”
Two Courts, One Move
Ask what both rulings actually decided, stripped of subject matter. The Delhi court didn’t rule that training on news content is fine as a general matter; it ruled that ANI’s specific evidence didn’t clear the bar of showing reproduction, which is a much narrower and more procedural finding than “AI training is lawful.” The Texas ruling is the same shape from the opposite side: the appeals court didn’t say Texas’s monitoring requirement was bad policy, it said Section 230 preempts the state from writing it at all, a jurisdictional finding that leaves the underlying child-safety question completely unaddressed. Both courts found the narrowest possible door out of a hard question, and both doors happen to favor the platform.
Why This Keeps Happening
Copyright statutes and Section 230 were both written for a legal landscape that didn’t anticipate generative AI or platform-scale content moderation at today’s volume, which means courts asked to rule on either one are stuck fitting a new fact pattern into old categories. The safest ruling a court can make in that position is procedural — evidence wasn’t sufficient, or federal law preempts the field — because it resolves the case in front of them without setting a substantive rule the court isn’t confident it can defend on appeal. That’s not judicial cowardice. It’s what happens when legislatures haven’t done the work the courts are being asked to substitute for.
The Case Against Reading Too Much Into Either Ruling
Both decisions are genuinely narrow, and treating either as a green light — “training on copyrighted news is legal,” “child-safety laws can’t survive Section 230” — overstates what either court actually held. A different plaintiff with better reproduction evidence, or a differently drafted state statute, could get an opposite result from the same courts on the same underlying questions.
Two governments, two different problems, and the same quiet admission: the law hasn’t caught up, so the judges are answering the smallest question they can get away with.