New AI act forces companies to disclose use of copyrighted content

New legislation introduced in the US House of Representatives demands that companies disclose copyrighted content used to train artificial intelligence models.

Adam Schiff, a United States representative, has introduced new legislation that requires all companies dealing in artificial intelligence (AI) to disclose what copyrighted material they used to train their generative AI.

If passed, this bill, known as the Generative AI Copyright Disclosure Act, will force AI companies to be transparent about what content their AI is trained on or face a fine of at least $5000.

This new act will apply to new and existing models and demands the formal submission of all copyrighted works “used in building or altering the training data set for that system,” a press release on the Adam Schiff website reads.

The information must be filed to the Copyright Office 30 days before the model becomes publicly available.

Generative AI models are trained on a multitude of content, including anything from articles to social media posts.

Naturally, as these models scrape the internet for a diverse range of data, which is required to train generative AI, they will most likely encounter copyrighted content that could be used without the creator's knowledge.

There have been many disputes over using copyrighted content to train AI models.

One notable case involved the New York Times and its legal battle against tech giants OpenAI and Microsoft.

The New York Times sued OpenAI and Microsoft, accusing them of using millions of newspaper articles without permission to train artificial intelligence technologies.