AI News Roundup – AI at CES 2025, New Biden AI chip export rules receive backlash, New York ponders state-level AI regulation, and more

To help you stay on top of the latest news, our AI practice group has compiled a roundup of the developments we are following.

• • AI-based technologies were in the spotlight at this year’s Consumer Electronics Show (CES) in Las Vegas, Nevada, according to Fast Company. AI chipmaker Nvidia, now the world’s second most valuable company by market capitalization, unveiled Project Digits, a compact “AI workstation” computer running off of Nvidia’s GB10 system-on-a-chip and aimed at AI developers. Nvidia also announced its RTX 50-series of GPUs, which the company claims is 2.5 times faster than previous models in generative AI tasks. Samsung, LG and Google unveiled generative AI features for their TV units and software, with the former two announcing integration with Microsoft’s Copilot AI system. Automaker Honda, in partnership with Sony, announced an AI-powered voice assistant for their new lineup of electric vehicles. As TechRadar reports from the show, AI was ubiquitous, and this demonstrates that “AI’s inherent power, versatility, and unprecedented exponential growth make it almost unlike any technology we’ve encountered before.” If CES is anything to go by regarding trends in high-tech industries, AI technologies are certainly here to stay.

• • The Biden administration is rushing to implement further restrictions on the export of semiconductors used for AI applications, according to Bloomberg. The U.S. plans to expand trade restrictions that were previously limited to most of the world to nearly every country, creating three “tiers.” The most permissive tier would be occupied by a small number of U.S. allied countries, including Canada, most of Western Europe, Australia, Japan, and South Korea, who would have little restriction on their access to U.S. AI chips, while the most restrictive tier, including Russia and China, would be blocked from importing the AI chips. In the intermediate tier are several major economies, including India, Brazil, and most of Eastern Europe, such as Poland, that would be subject to limits on the total computing power of chips exported to the country. Companies based in countries in the latter group would be able to bypass national limits by adhering to a set of strict U.S. security requirements and becoming a “validated end user.” AI chipmaker Nvidia harshly criticized the rules, saying in a statement that “[t]his last-minute Biden administration policy would be a legacy that will be criticized by US industry and the global community” and that “[i]t makes no sense for the Biden White House to control everyday data-center computers and technology that is already in gaming PCs worldwide, disguised as an anti-China move.” It remains to be seen if the incoming Trump administration will maintain these restrictions after it takes power on January 20.

• • A lawmaker in New York is proposing new state-level AI regulations, according to the MIT Technology Review. Alex Bores, a Democratic member of New York’s State Assembly, is currently drafting the “RAISE Act,” or “Responsible AI Safety and Education.” Similar to SB 1047, a California bill that would have placed stringent reporting and safety requirements on the most advanced AI models that was ultimately vetoed by Governor Gavin Newsom last year, Bores’ bill would require AI developers to maintain safety plans for their models as well as protect whistleblowers who report AI models that may cause “critical harms.” The bill does differ from SB 1047 in several key ways: Bores’ bill does not require a “kill switch” for models that would halt all operations, and does not create a new governmental body tasked with regulation of AI systems. Regardless of the final status of the RAISE Act, it remains clear that, in the absence of federal regulation of AI systems, states will continue to “step up” and address AI-focused issues in the U.S.

• • New legal filings allege that Meta knowingly trained its AI models on a database including millions of pirated eBooks, according to Reuters. According to newly-public filings by a group of authors suing Meta on the grounds of copyright infringement, internal communications found in the discovery process show that Meta used the dataset LibGen, which is said to contain pirated copies of the authors’ books, to train its AI models. The use of LibGen was allegedly approved by Meta CEO Mark Zuckerberg despite concerns within the company that LibGen was “a dataset we know to be pirated.” U.S. District Judge Vince Chhabria, who oversees the lawsuit, allowed the authors to amend their complaint but expressed further skepticism regarding the merits of their arguments.

• • AI technologies could transform the future of professional sports, including soccer, according to The Athletic. Soccer, known as “football” outside of the U.S., is the world’s most popular sport, and AI technologies have already been used by professional teams to improve their play. Liverpool FC, which competes in the English Premier League, partnered with Google’s DeepMind AI subsidiary last year to refine strategies for corner kicks. Now, some propose using AI agents to simulate entire matches, modeling players based off their styles and predicting how games might turn out. Football clubs around Europe already maintain massive libraries of scouting reports and other data about their own players and their rivals’, and AI tools could be used to expedite the normal process of interpreting such data. This has led to anxiety among some scouts, who worry that their jobs may be in danger of being replaced by an AI model. Others see AI applications as simply the natural evolution of data analytics, which has had great influence in sport for decades now. Regardless, apprehension still exists with how far the technologies will go, with one figure stating that “I’m reminded of the Jurassic Park line about the scientists being so preoccupied with whether they could, they didn’t stop to think whether they should.”