The Big Story: AI startup competitiveness at stake in copyright litigation
Startups could face higher costs and fewer opportunities to build with AI if courts require developers to license copyrighted material used to train AI models. This week, Engine released a new report examining the potential costs of those requirements and filed an amicus brief in the ongoing In re: OpenAI, Inc. Copyright Infringement Litigation. Although the parties in the case are the largest AI companies and largest rightsholders, the outcome could have far-reaching implications for startups’ competitiveness in the AI ecosystem.
The court is weighing whether using copyrighted works to train AI models constitutes infringement or is protected under the fair use doctrine. The brief, joined by Developers Alliance and the startup Lazy Moose Co., argues AI training is a fair use, explaining that AI training is highly transformative and how the resulting models enable startups to create innovative new services. AI models at or near the frontier are trained on billions of works comprising all of the Internet and more. If the court finds that training is not a fair use, the result would incentive anyone who has ever created anything to demand copyright royalties. The brief cites Engine’s new report showing just how large those costs could be—licensing the works needed to train a model at or near today's frontier could reach more than $10 trillion. That enormous potential amount, in practice, would likely instead lead to changes how models are developed—resulting in lower quality models trained on less data.
Of course, most startups aren't developing large AI models from scratch. Instead, they're building unique products on top of existing foundation models from the largest labs, meaning any increase in the cost of developing those models could affect their price, quality, and availability for startups. Higher cost or lower quality of the models startups rely on to build innovative new services will undermine their competitiveness and shrink the overall U.S. AI ecosystem. That’s why startups are paying close attention to these cases and why courts should find that AI training is a fair use.
Policy Roundup:
New York enacts new AI chatbot restrictions. Governor Kathy Hochul (D) signed legislation on Wednesday that bans certain AI chatbot features for users under 18 and requires chatbot services to verify the ages of all of their users. The law aims to carve out customer service and business uses, but the requirements may still impact startups offering services with chat-based user interfaces.
Advisory committee explores avenues for capital access expansion. The Securities and Exchange Commission Small Business Capital Formation Advisory Committee met this week to discuss lowering barriers to initial public offerings, ways to make regulation crowdfunding a more attractive option for raising capital, and an ongoing effort to expand the ways investors can qualify as accredited. Making these existing funding pathways more accessible and useful for startups is poised to bolster innovation, particularly by entrepreneurs located outside typical startup hubs and those from non-traditional backgrounds who may lack access to investor networks.
Oregon data center restrictions delay broadband projects. The Oregon Department of Transportation has paused at least 20 permits for broadband network expansion projects as officials consider whether they are connected to data centers. The delays make concrete how restrictions on infrastructure development can negatively affect connectivity essential for startup innovation.
Lawmakers seek greater transparency around AI safety. Three House Democrats wrote to leading AI companies pressing them to report significant AI safety incidents to Congress within 48 hours. The lawmakers are in line to chair key committees in the upcoming Congress if their party retakes the majority in next week’s election, previewing the role and perspective they expect to take in AI policymaking.
Telecom regulator advances efforts to boost connectivity via satellite. At the Federal Communications Commission’s October open meeting Thursday, commissioners considered multiple items aimed at bolstering “direct to device” satellite connectivity. The technology allows devices to connect to satellites without the need for additional antennas, and can be leveraged to connect to startups’ services by users in rural or seaborne locations.
On the Horizon:
TUE 11/3: Voters will head to the polls for the 2026 U.S. midterm elections to determine control of Congress, with potential implications for key startup policy debates around AI, immigration, trade, and more.
Startup Roundup:
#StartupsEverywhere: Somerville, Massachusetts. After witnessing the 2022 baby formula crisis, founder and CEO Annie Lu joined her brother in starting their Physical AI company, Laminar. They help process manufacturers ensure sustainable production of their products through real-time chemical data from their sensors and AI models they’ve created. We sat down with Annie to discuss her company, the importance of protective patents, and Laminar’s emphasis on providing sustainable processes to customers.