Two stories landed this week that, read together, sketch the outlines of a genuine epistemological crisis. Hyperallergic is calling Anthropic's mass book destruction the new book burning. Simultaneously, Apple has widened its trade secrets investigation into OpenAI, alleging additional former employees may have carried confidential intellectual property out the door. In one story, a corporation destroys books after training on them. In the other, human knowledge workers are accused of being the vector through which proprietary ideas migrate between AI labs. The symmetry is not accidental.
The Knowledge Pipeline Runs Both Ways
The Apple-OpenAI case positions human employees as conduits: vessels of institutional memory who carry what they know in their heads, their notes, their muscle memory of product decisions. The Anthropic book destruction story positions books as disposable training fuel, valuable only until the model has internalized them. A 2026 arXiv paper by Islam, Sarker, and colleagues on RAG-based modeling for misinformation mitigation makes a quieter version of the same argument: that LLMs need contextual, source-linked knowledge to stay accurate, not just bulk ingestion. The pipeline is the problem. What goes in shapes what comes out, and right now nobody agrees on who owns the pipe.
Cultural Memory as Contested Infrastructure
The Glasgow School of Art's inability to fund a faithful restoration of the fire-gutted Mackintosh Building completes a grim triptych. Physical archives burn. Digital archives get ingested and discarded. Human knowledge workers become liability. Brewster Kahle's argument, made in his conversation on building the global brain and public AI, that libraries are unfinished business feels more urgent than ever. The Internet Archive exists precisely because private actors cannot be trusted to preserve what they've consumed. This week proved the point twice before Tuesday.