Authors, media, and artists sued model providers over works used in pretraining; courts began setting tests.
Since 2023, lawsuits in the U.S. and elsewhere asked whether training on protected works without a license is fair use or infringement. Cases touched text, image, and code.
Partial settlements, filtered datasets, and crawler “opt-outs” reshaped the data market. Licenses and markets for synthetic/authorized content emerged.
The legal endgame is still evolving, but it already alters costs and strategies for acquiring frontier corpora.
September 2026 reporting (Business Insider): Meta’s Model Capability Initiative (MCI) sought to collect employee keystrokes and mouse data to train computer-use AI in a TBD unit. An internal revolt followed—flyers, memes, an “Eye of Sauron” petition—then concessions and a halt after an internal data leak.
On September 15, 2026, the OpenAI Foundation announced Public Data for Health / Data for Public Health, with $40 million to UNC Lineberger Comprehensive Cancer Center to generate multimodal data that could improve personalized cancer vaccines, plus support for OpenAdmet. MIT Technology Review and UNC itself detailed the bet: higher-quality biological observations so models can advance medicine.
On 22 September 2026 Teradata expanded its Tera assistant with a Context Engine, execution Harness and reusable Agent Skills for governed data engineering and science; availability planned for Q4 2026, with claims of fewer tokens than Claude Code on SWE-bench Pro.