DOJ: OpenAI AI Training Is Fair Use, Omits Equity Stake Talks

• The Justice Department filed a statement of interest asserting that training large language models on copyrighted text constitutes fair use.
• The DOJ's brief cites national security as a key reason for supporting AI development and its fair use defense.
• The administration has reportedly discussed acquiring a 5% equity stake in OpenAI, valued at approximately $42.6 billion.
• The Justice Department's three-page brief, titled "The Interest Of The United States," did not disclose the potential government ownership stake in OpenAI.
• The brief relied on blog posts from OpenAI and Anthropic to support claims of AI breakthroughs, rather than independent verification.

The absence of any mention of this substantial financial interest in the DOJ's official statement has raised questions, particularly given the department's active intervention in what is described as a 'potentially existential legal battle' for OpenAI.

The Justice Department has formally intervened in the ongoing copyright litigation against OpenAI, submitting a statement of interest that articulates the administration's official position on the use of copyrighted material for artificial intelligence training. In this three-page document, titled "The Interest Of The United States," the DOJ asserts that training a large language model (LLM) using copyrighted text constitutes fair use. This intervention comes amidst a sprawling legal battle where OpenAI faces accusations of copyright infringement, including cases initiated by stand-up comics and the New York Times.

The department's filing suggests that the legal framework should support the development of AI, arguing that rules hindering a robust AI industry in the United States could threaten national security. The brief posits that such restrictions might grant a competitive edge to foreign adversaries who operate under fewer constraints. While acknowledging that AI will play a significant role in cybersecurity, the DOJ's argument for national security is presented as a broad justification for its stance on AI training.

The Justice Department's brief outlines specific conditions for its fair use determination, noting that the material must have been acquired legally by OpenAI. It also includes a caveat that the fair use defense would not apply if the AI model were to reproduce exact copyrighted text as output. This nuanced position aims to balance the need for AI development with existing intellectual property protections, though its application in various ongoing cases remains a point of contention, especially given past instances where OpenAI has not always legally acquired training material.

A significant detail omitted from the Justice Department's brief is the reported discussions between the administration and OpenAI regarding a potential federal government equity stake in the company. Reports indicate that the federal government has been negotiating for a 5% ownership share in OpenAI. Based on the company's valuation of $852 billion, this stake would be worth approximately $42.6 billion.

The absence of any mention of this substantial financial interest in the DOJ's official statement has raised questions, particularly given the department's active intervention in what is described as a "potentially existential legal battle" for OpenAI. Critics point to this omission as a potential conflict of interest, suggesting that the government's advocacy for OpenAI's fair use defense could be influenced by its prospective financial ties to the AI giant.

This reported negotiation for a government 5% stake in OpenAI introduces a complex layer to the Justice Department's advocacy for AI training fair use. The department's brief, while asserting the importance of protecting "American ingenuity and intellectual property from exploitation and theft," notably refrains from addressing the protection of artists and creators from similar exploitation, further highlighting the perceived imbalance in its stated priorities.

The Justice Department's invocation of national security as a primary justification for its fair use stance on AI model training has drawn scrutiny. The brief argues that hindering AI development in the U.S. would empower foreign adversaries, framing the issue as critical for technological infrastructure defense. However, the application of this broad national security claim to specific copyright disputes, such as those involving stand-up comics' routines or the New York Times' restaurant reviews, has been characterized as an overreach. The argument suggests that the nation's cybersecurity does not hinge on whether an AI can generate "a passable tight five" or reproduce copyrighted reviews.

Furthermore, the DOJ's brief cites the White House cybersecurity order twice, referencing frontier model benchmarking and CISA directives, to bolster its national security claims. Yet, when substantiating the "major research breakthroughs" delivered by large language models, the department relied on blog posts from OpenAI and Anthropic, rather than independent verification. This reliance on sources with a direct commercial interest in promoting AI capabilities, despite the vast resources of the federal government, underscores a perceived lack of independent rigor in the brief's factual assertions.

While the Justice Department champions the protection of "American ingenuity and intellectual property," its brief conspicuously avoids any mention of safeguarding artists from exploitation or theft. This selective focus, coupled with the general principle that copyright law's fair use exception is intended to facilitate greater creative expression, highlights a tension. Historically, copyright disputes often center on the creative artist, but outcomes frequently benefit large corporate rights-holders who manage intellectual property interests to maximize wealth, rather than the individual creators themselves. The ongoing debate surrounding LLM copyright infringement defense continues to navigate these complex legal and ethical waters.

Lawyers advising on AI development or copyright enforcement should note the DOJ's official stance supporting fair use for AI model training, but also be aware of the potential for perceived conflicts of interest given the reported discussions of a government equity stake in OpenAI. This could influence future legal interpretations or regulatory actions concerning AI and intellectual property, particularly when national security is invoked.

Source: Original reporting via legal news analysis.