We will need some clear definitions rather quickly…
https://thenextweb.com/news/trump-ai-military-memo-autonomous-weapons
Trump signs memo putting ‘most advanced AI’ into military hands and banning vendors from pulling the plug
President Trump signed a national security presidential memorandum on Friday ordering the US military and intelligence agencies to accelerate their adoption of cutting-edge AI. The directive, NSPM-11, establishes a framework for “rapid onboarding of the most advanced AI models from multiple vendors.” It also bars any company from disabling, degrading, or modifying an AI system that warfighters depend on without prior government approval.
That vendor restriction is the most striking provision. It means an AI company cannot pull a deployed model from military use unilaterally, even if the company has safety concerns about how it is being used. The clause lands directly in the context of the Pentagon’s ongoing feud with Anthropic, which was blacklisted as a supply chain risk after refusing to allow its Claude models to be used for autonomous weapons or mass surveillance.
Let the arguments begin!
https://scholarship.law.vanderbilt.edu/jetlaw/vol28/iss3/5/
When Robots Read Westlaw: Linking Inputs and Outputs in Generative AI Fair Use Analysis
Generative artificial intelligence has revealed a novel tension in copyright law: a two-stage act (input and output) of copying that traditional fair-use doctrine never squarely anticipated. In the input stage, vast swaths of copyrighted expression are reproduced and ingested to train large language models. In the output stage, those models generate works that may substitute for or dilute the market of the originals. Courts now face both stages. Their rulings diverge when they consider training and outputs together. In Bartz v. Anthropic PBC and Kadrey v. Meta Platforms, Inc., the courts characterized training uses as “highly transformative,” yet reached opposite conclusions on market harm because the evidentiary records differed. By contrast, in Thomson Reuters v. ROSS Intelligence, the fair-use defense failed outright: ROSS’s AI-generated outputs were found to compete directly with Thomson Reuters’s paid research tools. These cases confirm that transformation alone is not dispositive: the fourth statutory factor—whether outputs substitute for or erode the market for the original—can ultimately determine the outcome. This Note therefore advances an “input-output linkage” framework. It would enable courts to evaluate the permissibility of unlicensed training (input) by looking to the commercial purpose, functional use, and market effects of the resulting outputs. This framework helps explain why wholesale copying for training may qualify as fair use when outputs serve new, non-substitutive functions (as Kadrey intimates), yet fails when outputs compete with the original work or its derivatives (as in Thomson Reuters). Aligning fair use doctrine with copyright’s utilitarian goals in this way preserves incentives to create while also permitting socially valuable AI innovation. Still, adjudication alone cannot fully resolve the structural licensing challenges posed by generative AI. This Note argues that to supply ex-ante clarity and enable scalable markets for training data, targeted legislative or regulatory measures are needed. Such interventions would ensure that copyright law protects both creative labor and accommodates the need for training data.
