US urges G-20 countries to allow AI training on creators’ work

CHAPEL HILL, North Carolina – The US urged G-20 countries on Sept 2 to develop rules governing how AI companies use copyrighted material to train their systems, as major American technology firms face mounting lawsuits over the practice.Companies including Anthropic, OpenAI, Google and Meta have been sued in the US by authors, publishers, media organisations and other creators who allege the firms misused their work to develop AI products.Tech firms train their AI models to generate images and text by feeding them vast amounts of data, often including books, music and art. They have argued that using that work to build their AI tools is legal.US Commerce Secretary Howard Lutnick told the G-20 officials at a tech-focused meeting in North Carolina that they should allow AI companies to train their models on creators’ work, while finding a way to “protect artists”.He said the countries should embrace “fair use”, a principle that allows the unauthorised use of copyright-protected works under certain circumstances.“We need frameworks that protect creators and inventors whose ingenuity drives our economies without creating the barriers that prevent the next generation of innovators,” Lutnick told the gathering of officials representing G-20 countries.He did not offer specifics on how to accomplish that.The US Justice Department separately on Sept 2 filed a brief supporting OpenAI in its dispute with The New York Times and several other newspapers over ​the company’s use of their work to train the large language ‌models behind ChatGPT, saying AI training generally makes fair use of copyrighted material.Nvidia CEO Jensen Huang appeared alongside Lutnick ahead of the commerce secretary’s speech on Sept 2.Huang urged the G-20 officials to avoid writing AI regulations focused on “theoretical harms”, and instead develop rules addressing real-world problems tied to the technology. His message echoed those of the Trump administration and other American tech executives who addressed the gathering on Sept 1.The G-20 nations will aim to avoid writing new regulations specifically for AI, according to a joint statement released on Sept 2 by the US. Any future rules should address only “novel considerations” posed by AI that are not already covered by existing laws, the statement said.Government requirements could hurt the industry’s profits if they slow the release of new models or prompt the companies to change how their products perform to address security concerns.China, which has already implemented a broad list of AI regulations, was among the signatories to the statement.Anthropic co-founder Tom Brown, Palantir’s Alex Karp and OpenAI’s Sam Altman also addressed the gathering on Sept 2. REUTERS