The US Copyright Office wants to hear your AI takes

The US Copyright Office is finally stepping into the wild west of generative AI, and it wants the public to weigh in before setting the rules.

The agency is opening a public comment period starting August 30th to tackle the messy collision between artificial intelligence and intellectual property. Anyone wanting to submit written comments has until October 18th, while follow-up replies are due by November 15th.

Three massive questions on the table

In its Federal Register notice, the Copyright Office laid out three primary questions it needs answered.

First, how should AI models use copyrighted material during training? Second, can AI-generated content receive copyright protection if there is no human creator involved? And third, how should copyright liability actually work when an AI system infringes on protected work?

The agency also asked for thoughts on AI potentially violating publicity rights. While the office noted this is not technically a copyright issue, mimicking an artist's voice, likeness, or visual style could easily trigger state-level unfair competition and publicity laws.

The agency noted that it has already received multiple registration applications for works containing AI-generated elements over the past several years, and public input will help guide its future rulings.

Legal fights are already piling up

This inquiry comes on the heels of major legal friction. The Copyright Office was recently sued after denying copyright to Stephen Thaler for an AI-generated image. Earlier this month, a federal judge in Washington, DC, sided with the agency, ruling that human authorship remains an absolute requirement for copyright protection.

Meanwhile, the tech industry is already facing a barrage of lawsuits over training data. Generative AI models regularly crawl public web content, and creators are fighting back.

Three visual artists have sued Stable Diffusion, Midjourney, and DeviantArt for training on their artwork without consent. Comedian Sarah Silverman and authors Christopher Golden and Richard Kadrey filed lawsuits against OpenAI and Meta, alleging their copyrighted books were used to train ChatGPT and LLaMA. On top of that, major news organizations have started blocking OpenAI's web crawlers, while Senate Majority Leader Chuck Schumer is urging lawmakers to speed up formal AI rulemaking.

Why it matters

The outcome of this proceeding could fundamentally reshape how generative AI is built and monetized.

For creators and developers, the immediate takeaway is clear: pure AI output still has zero federal copyright protection in the US. If you build a product, brand asset, or game entirely with generative tools, anyone can legally copy it without asking. If creators want ownership, human creative control must remain front and center.

For tech giants, training data liability is the real existential risk. If regulators decide that scraping public web content without explicit consent constitutes copyright infringement, the current business model behind frontier models breaks down. Companies like OpenAI and Meta would have to pay massive licensing fees or rely heavily on synthetic data, significantly raising the barriers to entry for smaller AI startups.

We think the Copyright Office is being smart by opening the floor rather than issuing blanket bans. But bureaucratic timelines move at a glacial pace compared to model development. By the time this review wraps up in late November, the technology will have already sprinted ahead.