IP Report: The White House Just Called AI Model Weights Intellectual Property
President Trump’s June 2, 2026 Executive Order Promoting Advanced Artificial Intelligence Innovation and Security established federal policy treating AI model weights as intellectual property protected from foreign adversary theft, a designation with real consequences for how companies structure AI asset protection. The order directs NSA, CISA, and the Departments of Treasury and Defense to design a voluntary framework by August 1, 2026 allowing developers of covered frontier AI models to engage the federal government before release, with access capped at 30 days and subject to IP nondisclosure agreements. The order expressly prohibits mandatory licensing or preclearance requirements, positioning the United States against the regulatory path the European Union has taken with the EU AI Act. The IP protection framing matters because it signals that the administration views AI model weights as analogous to trade secrets or proprietary technology, which will likely influence how the Department of Justice and federal agencies approach investigations of AI-related IP theft by foreign state actors.
In a separate development this week, a California federal court ruled that writers including John Carreyrou must pursue their copyright infringement claims against Anthropic, Apple, Google, Perplexity AI, Nvidia, and xAI in separate individual lawsuits rather than as a coordinated action. The procedural ruling does not reach the merits but eliminates the litigation economy a consolidated case would have provided to plaintiffs. Publishers including Elsevier, Hachette, Macmillan, Cengage, and McGraw Hill are separately pressing a putative class action against Meta Platforms Inc. alleging that Meta torrented copyrighted books from pirate sites and stripped copyright management information from the acquired works, a theory that bypasses fair use analysis entirely by relying on DMCA Section 1202. The unresolved question across all of this AI copyright litigation is whether any appellate court will issue a controlling fair use ruling before the sheer volume of settlements reshapes the licensing market for AI training data.