Casey Goes to the White House + The Copyright Battle Over Artificial Intelligence + HatGPT

Episode Summary

Executive Summary: This episode centers on two major themes: the Biden administration’s new AI executive order and the legal fight over whether AI companies can train models on copyrighted works. The hosts debate open vs. closed AI, government oversight, and the coming regulation of powerful models, then interview copyright scholar Rebecca Tushnet about fair use, training data, and creators’ rights. A lighter Hatch GPT segment closes with AI absurdities and self-driving car news.

Main Topics: Biden’s AI executive order (Priority: 5/5): The hosts review the administration’s sweeping AI executive order, focusing on disclosure requirements for large model training, safety testing, and government monitoring of frontier systems. Open source vs. closed AI debate (Priority: 5/5): They discuss the tech industry split over whether open development is safer and more democratic or whether closed systems with fewer actors are more responsible for high-risk models. White House AI policy perspective (Priority: 4/5): Kevin’s White House visit highlights how Biden aides frame AI as both promising and dangerous, with emphasis on balanced regulation and early intervention. Copyright and AI training data (Priority: 5/5): Rebecca Tushnet explains why many AI training and output questions fit within existing copyright doctrine and fair use, while acknowledging unresolved litigation and policy tensions. Litigation against Stability AI and others (Priority: 4/5): The episode covers a court ruling dismissing many artist claims against Midjourney, DeviantArt, and Stability AI, while allowing a direct infringement claim to proceed. Hatch GPT news roundup (Priority: 2/5): The segment satirizes AI failures and tech news, including an AI Seinfeld loop, a barge-based compute cluster, Microsoft’s problematic poll, and Cruise pausing driverless operations.

Key Arguments: The executive order is significant because it tells the AI industry that Washington is watching early, rather than waiting years to react as it did with social media. The order’s frontier-model reporting threshold is meant to capture the biggest systems, not ban model development outright. Open-source advocates argue regulation could entrench incumbents and create regulatory capture, making it harder for smaller competitors to challenge the largest AI firms. The administration says AI has both democratizing and proliferating effects, so policy must balance opportunity and risk. Existing copyright doctrine, especially fair use, can likely address many AI training and output disputes without wholly new law. Training on large amounts of copyrighted material may be legally defensible if the output is transformative and non-infringing, though that remains contested in court. Individual creators may struggle to extract meaningful compensation because licensing power is concentrated among large rights holders and publishers. Regulation of AI may still lag until a serious harm or public crisis forces stronger legislative action, as has happened historically with other technologies.

Data Points: Executive order length: More than 100 pages - The hosts describe the AI executive order as comprehensive and very long. Training threshold: 10^26 floating point operations - One trigger for reporting requirements is training a model with computing power above this threshold. Threshold in plain language: 100 septillion flops - The hosts translate the executive-order trigger into a more readable number. White House event timing: Monday before Halloween - Kevin visited the White House during Halloween festivities. Legal filing outcome: Most claims dismissed - In Anderson et al. v. Stability AI et al., the judge dismissed claims against Midjourney and DeviantArt and allowed one direct infringement claim against Stability AI to proceed. Safety-side example: 3-second voice clip - During Biden remarks, a quote referenced how a short voice sample could be used deceptively. Driverless car pause: All driverless operations in the United States - Cruise paused U.S. driverless taxi operations after a California regulatory suspension.

Pivotal Quotes: "If I were still in venture capital, I would say the technology is democratizing. If I were still in the defense department, I would say it's proliferating and they're both true." — Arthi Prabhakar: White House OSTP director describing the administration’s balanced view of open-source AI. "This is not going to be another social media where you have a decade to sort of build and chase growth and spread your products all over the world before we start holding hearings." — Kevin Roose: Summarizing the administration’s intent to regulate AI earlier than past technologies. "I think that the law has the tools to handle this, that they're pretty conventional questions." — Rebecca Tushnet: Her core view that current copyright law can address many AI-related disputes without entirely new doctrine.

Implications: Listeners should expect faster, more proactive AI regulation, especially for frontier models and safety testing. The copyright fight is likely to continue, but current law may already shape outcomes. AI governance will be fought in courts, agencies, and globally, with Europe possibly setting the standard.