EmbeddingGemma 2: an open, lightweight multimodal embedding model
deepmind.google·6d ago
AI Engineering
Hacker News·Aug 13·5 min read
If I own Claude's outputs why can't I train my own model on them?
TL;DR
While users own the Outputs generated by Claude, they cannot use these Outputs to train competing AI models without permission. This policy is in place to maintain safety and protect Anthropic's investments.
✦ Why It Matters
Review your AI model training plans to ensure compliance with Anthropic's policies on Claude's Outputs.
Key Takeaways
- Use Outputs for sentiment analysis and content categorization.
- Avoid training models that compete with Claude.
- Integrate Outputs into applications for enhanced features.
- Seek written permission for any competitive model training.
Full Summary
When using Claude, users own the Outputs generated from their Inputs, but there are strict limitations on using these Outputs for training AI models. Anthropic, the company behind Claude, emphasizes safety and has invested heavily in ensuring its models are harmless.
Allowing users to train new models on Claude's Outputs without oversight could lead to unsafe AI systems, as these models would lack the safety measures implemented by Anthropic. Additionally, using Claude's Outputs to create competing models undermines the company's business.
Users can utilize Outputs for non-competitive applications like sentiment analysis, content categorization, and anomaly detection, but must refrain from developing general-purpose chatbots or models for open-ended text generation.