AI Knowledge Supply Chain
Where AI's knowledge comes from, who owns it, and what happens when AI starts producing the content it learned from — copyright, training data, and synthetic-content authenticity.
Current understanding
The corpus treats AI's input and output as one supply-chain problem rather than two. Who Teaches the Machines What to Know? sets the upstream question directly: who decides what AI knows? The argument is that training-data choices are policy decisions disguised as engineering decisions, and almost no legal voice is at the table. Inside the Machine: Trust, Truth, and the Future of Knowledge in an AI World extends the question to truth itself — Bilva Chandra on AI safety inside frontier labs and inside the institutions trying to govern them. Downstream, The AI Content Tsunami with Guy Shahar takes the same supply chain to its terminal point: with Sora 2, Veo 3, and Marble, the cost of producing high-fidelity video, audio, and text approaches zero. The legal consequences — evidence, identity, deepfakes — are arriving faster than doctrine. The copyright fight is the policy proxy. Who Pays for the Truth? The UK's Copyright Battle with Big Tech tracks the UK government's choice between protecting creative industries and watering down copyright to attract US tech investment. The framing in the corpus: this is not a niche IP dispute, it's a referendum on whose data rights apply to AI training. The EU is mostly watching for now.
Tensions
- If AI is trained on copyrighted work without licensing, the value capture is upstream of the legal services industry — but the litigation lands in legal services. Who profits, who pays?
- Synthetic content (deepfakes, generated audio) is already in court evidence. Doctrine is years behind. Practising lawyers will be the ones inventing the standards.
- The UK and US are moving in different directions on training-data copyright. The EU has not yet shown its hand. Mino's customer base is in the jurisdiction that hasn't decided.
- Knowledge provenance for in-context retrieval (e.g. RAG) is a separate problem from training-data provenance — and is barely discussed in the corpus.
Mino relevance
EU sovereignty is a real Mino differentiator. Dutch-resident, EU-data-trained, traceable provenance — these are specifications, not slogans, when Mino sells against US tools whose training data is contested. Strategic implication: lean harder into the supply-chain argument as a Mino positioning pillar. Content opportunity: a Mino piece on "what your AI was trained on, and why your liability turns on the answer" would land cleanly with the in-house and law-firm audiences who follow this corpus.
Sources
4- Inside the Machine: Trust, Truth, and the Future of Knowledge in an AI WorldLaw What's Next · Apr 13, 2026
- Who Teaches the Machines What to Know?Law What's Next · Apr 12, 2026
- Who Pays for the Truth? The UK's Copyright Battle with Big TechLaw What's Next · Mar 13, 2026
- The AI Content Tsunami with Guy ShaharLaw What's Next · Nov 19, 2025