The US government is thinking about changing the rules around how AI systems can use copyrighted works like books, movies, and songs to learn. They want to hear what people think. A lot of the biggest AI companies gave their opinions.
Most of the companies said they don’t think AI should have to pay fees to use copyrighted works to train models. Here’s a quick look at what some of them told the government:
Meta (Facebook) said that copyright owners wouldn’t make much money anyway if fees were required. They would have to try contacting millions of people who own copyrights. However, each work is only a small part of all the data used to train an AI.
Google compared using copyrighted works to train AI to read a book. Just like reading doesn’t require permission, analyzing data to build AI systems shouldn’t either. They said the goal is to learn facts and ideas, not copy the actual works.
Microsoft warned the rules could hurt small AI startups. It’s too hard and expensive for most companies to get permission from millions of copyright owners. This could limit AI progress to just a few big tech giants.
Apple mostly talked about who can own the copyright if an AI writes computer code. They think the human programmers working with AI tools should be able to copyright the code the AI helps create.
Other companies like Anthropic, Adobe, Hugging Face, and Stability AI said current laws are fine or that training AI is fair use similar to other exceptions in copyright law like research. One venture capital firm said changing the rules now could ruin past investments of “billions and billions” in developing AI.
The comments show these companies want to keep using copyrighted works without payment so they can continue advancing AI. But will the government agree? Stay tuned as the rules are decided!
