On Wednesday, U.S. lawmakers delved even deeper into the impact of generative artificial intelligence on copyright protection, addressing short- and long-term questions about how tech companies collect data to train AI and generate content based on it. emphasized the question.
The hearing is the third in a series held by the Senate Judiciary Committee, following previous discussions in May and June that focused on other aspects of AI and IP, including copyright law, innovation and competition. is.
One of the main areas of discussion focused on whether to require companies to allow users to opt out of using their data to train AI models. Ben Brooks, head of public policy at Stability AI, said the company has already received more than 160 million opt-out requests from people who don’t want their images used in AI models.
But when Senator Chris Coons asked if Stability AI paid people who allowed their data to be used, Brooks avoided a direct answer, instead having a large data set. chose to point out that it is important
“It’s important to have diversity to make it viable,” Brooks said.
The debate also comes as technology companies face new legal challenges. A class action lawsuit was filed against Google this week, alleging the tech giant violated state and federal copyright and privacy laws when developing AI products. The complaint comes less than a month after OpenAI received another but similar lawsuit from the same company. Earlier this year, Stability AI was sued by Getty Images, alleging the startup illegally used millions of images to train his AI models.
The AI company executives who testified said their AI models were trained on content deemed “fair use.” But Senator Marsha Blackburn said “fair use” protection has become “a pretty useful way to steal” intellectual property. Another committee member, Senator Amy Klobuchar, suggested lawmakers should consider banning AI-generated content for some uses, suggesting how AI is used to generate misinformation about political advertising. expressed concern about being used. (It’s worth noting that the two senators disagree on virtually every other issue.)
AI experts say it’s hard to remove content that’s already been used as training data from large language models. It’s also difficult to opt out of data sharing unless you know in advance what’s being opted in, said Jeffrey Hurlston, general counsel and business and legal executive at Universal Music Group. He believes companies must obtain consent before using content to train AI, adding that some artists don’t want their music distributed for reasons beyond commercial viability.
“Creativity is the soundtrack to our lives,” Harleston said. “And without the foundation of copyright, we might never have even known about it.”
Emory Law School AI expert Matthew Sugg believes the debate is focused on commercial protections rather than on ordinary people who still want protection from deepfakes and other AI content. said he was “extremely concerned.”
One of the proposals is a new “anti-impersonation law” to protect artists from AI impersonation, put forward by Adobe vice president and general counsel Dana Rao. In his opening remarks, Rao, who oversees the company’s legal and policy team, said statutory damages should be included in the law so that artists are not obliged to prove actual damages. (Adobe and others are working to create a new standard for generative AI through organizations like the Content Authenticity Initiative, which currently has more than 1,500 authors, including Stability AI, Universal Music Group, and Getty Images. have members.)
The AI-influenced artist is represented by concept artist and visual developer Carla Ortiz, whose works include Black Panther, Guardians of the Galaxy, and Doctor Strange. In his testimony, Ortiz recalled being “horrified” to find out how his design helped train an AI model, adding, “I never worried about my future as an artist. Never,” he said. (Ortiz has sued his Stability AI in the past.)
“I realized that nearly all of my work, that of nearly every artist I know, and that of hundreds of thousands of artists, was filmed without our consent, credit, or compensation.” she said. “These works were stolen and used to train commercial technology using datasets containing billions of image and text data pairs.”
https://digiday.com/?p=510895
