OpenAI, Microsoft win partial dismissal in developer AI training suit

Litigation concerning generative AI training data typically follows a pattern where plaintiffs assert multiple overlapping legal theories, including direct copyright infringement, breach of contract, violation of the Digital Millennium Copyright Act (DMCA), and unfair competition. Courts frequently evaluate these claims at the motion-to-dismiss stage to determine if plaintiffs have sufficiently pled facts to support each distinct cause of action. Partial dismissals are common in complex technology litigation, serving as a judicial filter that removes legally insufficient theories while allowing fact-intensive inquiries like fair use to proceed to discovery. This procedural mechanism distinguishes between claims that fail as a matter of law and those requiring evidentiary development. Historically, software copyright cases involving AI have also intersected with open-source licensing disputes, adding contractual dimensions to pure copyright questions. The recurrence of such filings across different plaintiff classes—developers, artists, publishers—suggests a coordinated stress-testing of legal boundaries rather than isolated grievances.

On September 16, 2026, a federal court issued a partial dismissal ruling in favor of OpenAI and Microsoft regarding a lawsuit brought by software developers alleging copyright infringement through AI training. According to Reuters Legal, the defendants successfully fended off specific claims within the litigation, resulting in a narrowing of the legal theories available to the plaintiffs [2][3]. The court’s order dismissed certain causes of action for failing to state a valid claim while preserving core copyright allegations related to the ingestion of proprietary source code for model training. This procedural victory establishes that not all theories of software copyright liability apply equally to generative AI development, though it does not constitute a final adjudication on the merits of fair use or direct infringement.

The plaintiffs, representing software developers, had alleged that OpenAI and Microsoft unlawfully profited from the unlicensed ingestion of proprietary source code to build their AI products. They contended that this conduct violated copyright protections and applicable software licenses. In response, OpenAI has consistently denied copyright infringement, arguing that AI training on publicly available code constitutes fair use under U.S. copyright law. Microsoft similarly contended that its development of GitHub Copilot complied with relevant open-source licenses and copyright statutes. The partial dismissal suggests the court found deficiencies in how certain non-copyright or ancillary claims were pled, distinguishing them from the central question of whether training itself is infringing.

Simultaneously, the legal landscape surrounding AI training data remains highly active and contentious. On the same day as the partial dismissal, attorney Matt Platkin announced the filing of a separate lawsuit against OpenAI and Microsoft on behalf of more than 150 publications [1]. This new complaint alleges systematic theft of protected news articles and other works to build AI businesses, highlighting that while software developer claims face procedural hurdles, media publishers continue to press similar theories of unlawful conduct. Additionally, market observers noted reports of fresh lawsuits targeting these companies over AI copyright infringement, indicating a sustained volume of litigation despite individual procedural wins for defendants [4].

The sequence of events on September 16 underscores a bifurcated legal reality for AI firms. While they secured a tactical win in narrowing the software developer suit, the immediate filing of parallel litigation by news publishers demonstrates that the broader dispute over training data provenance is far from settled. The partial dismissal clarifies which legal vehicles are viable for code-based claims but leaves the fundamental tension between intellectual property rights and AI development unresolved. Stakeholders now await further proceedings on the surviving copyright claims, which will likely test the fair use doctrine more directly than the dismissed theories did.

Software developers and news publishers risk having their primary legal theories invalidated, forcing costly retooling of litigation strategies or settlement at lower valuations. OpenAI and Microsoft benefit from reduced exposure to ancillary claims but still face existential fair use battles on core copyright issues. Over 150 publishers’ newly filed suit indicates financial and reputational risks persist despite procedural wins. The outcome will define licensing economics for training data and set compliance baselines for the entire generative AI sector.