The White House just picked a side in the fight that will determine whether AI companies owe billions or owe nothing.
The Summary
- The Trump administration filed an amicus brief supporting OpenAI against the New York Times' copyright lawsuit, arguing that blocking AI training on copyrighted material threatens national competitiveness
- The government's position could reshape AI copyright law globally, setting precedent for how foundation models are built and who pays for the data
- If the administration's view wins, every major AI lab gets a green light to train on published content without licensing deals
The Signal
The New York Times sued OpenAI in late 2023, claiming the company illegally trained its models on millions of articles without permission or payment. Now the federal government is telling the court: making that illegal would hand China the AI race.
The administration's brief argues that training AI models on copyrighted content falls under fair use, the same doctrine that lets search engines index websites and researchers quote books. The logic: the models aren't republishing the work, they're learning patterns from it. Block that, the government says, and you kill the entire foundation model industry in the US.
"The government's stance could redefine AI copyright norms, impacting global AI competitiveness and innovation policies."
This isn't abstract legal theory. It's about billions in licensing fees and who controls the data pipeline for AGI. If the Times wins, OpenAI, Anthropic, Google, and every other lab building multimodal models faces a choice:
- Pay publishers for training data retroactively and going forward
- Train only on licensed or public domain content, crippling model quality
- Move training operations to jurisdictions with looser copyright enforcement
The timing matters. China's DeepSeek models already train on vast internet corpora with zero copyright friction. The administration is framing this as national interest, not corporate welfare. If US companies get bogged down in licensing negotiations while Chinese labs hoover up everything, the capability gap widens fast.
Key implications:
- Sets precedent for Web4 agents that generate content vs. Web2 platforms that host it
- Determines whether foundation models are infrastructure (like roads) or products (like cars)
- Shapes whether tokenized creative work gets protection or gets absorbed into training sets
The court hasn't ruled yet, but the Justice Department weighing in shifts the frame from "did OpenAI steal?" to "can America afford to tie one hand behind its back?" Publishers see existential threat. AI labs see existential necessity. The judge gets to decide which existential risk matters more.
The Implication
If you're building agents, watch how this lands. A ruling that requires licensing every training source makes foundation models more expensive to build and gives incumbents with deep pockets and existing content deals a moat. A ruling that blesses fair use opens the floodgates for smaller labs and open source models to compete on compute and architecture, not legal budgets.
For anyone creating content, this is the moment to think hard about watermarking, attribution tech, and whether you want your work feeding models or fighting them. The legal framework getting built right now will govern the next decade of AI development. The administration just signaled it cares more about winning the race than protecting old media business models.