Provenance, Copyright & Lawyers
less than a minute
There is an issue with provenance of the training metadata. For most LLMs use cases generating revenue today:
here are some court cases that were current in September 2025
By | Against | In | For |
|---|---|---|---|
| Getty Images | Stability AI | UK & US | ~ $1,700M |
| Authors class action → (settlement) — United States | Anthropic | USA | ~ $1,500M |
| NY Times & others | OpenAI / Microsoft | USA | $150k per work |
| Encyclopaedia Britannica | Perplexity AI | USA | not disclosed |
| Disney, NBCU, WBD | MiniMax | USA/China | not disclosed |
| Japanese news | Perplexity AI | Japan | ¥2.2 billion (≈ $15M) |
| Artists | Midjourney et al | USA / UK | not disclosed |
| Canadian news | OpenAI | Canada | not disclosed |
If you want to be ethical about your content & know where it’s coming from, check out
content Authenticity Initiative
- a trusted chain of data to trust content
Coalition for Content Provenance and Authenticity
- the trust mechanism
SMPTE
- the trust mechanism in MXF for multi-stage professional media workflow
fairlytrained.org
- they certify that models have not stolen your work or your
clients work
ISO/IEC 42001:2023
- an AI system is well managed
ISO/IEC 5259:2024
- AI data quality management