What happened
The New York Times and The Daily News have accused OpenAI of withholding evidence in their ongoing copyright lawsuit. According to a report from TechCrunch, the outlets claim an OpenAI engineer admitted in a deposition that the company had already conducted internal searches for copyrighted works in its training data — a capability OpenAI had previously argued was technically burdensome to perform. The publishers are now asking the court to sanction OpenAI for allegedly disrupting the discovery process.
How the room's reading it
OpenAI's official response frames the allegations as an attempt by the Times to invade user privacy as its legal case weakens. Legal watchers on X see a messy, but not unusual, discovery dispute in a high-stakes corporate lawsuit. The more potent reaction is coming from developers and infra teams. For them, this isn't about legal tactics. It's about existential risk. The consensus among practitioners is that if a firm with OpenAI's resources faces these accusations, smaller teams building on scraped data are exposed to enormous liability, making data provenance a critical issue.
Sailfish's take
This dispute makes an abstract risk feel very concrete. We've seen teams treat training data as a problem for tomorrow, but this filing suggests tomorrow is here. Any product built on a model with an opaque data lineage is inheriting this legal uncertainty. It's not just about what the model can do — it's about what it was trained on. We believe the era of asking for forgiveness on data scraping is closing fast. If you're shipping a generative product, this is the week to start a serious audit of your data sources and dependencies. The bill is coming due.