The Information: Meta has reportedly limited engineer use of Claude Code and Codex because rival model outputs could con…
Rohan Paul Twitter · Rohan Paul (@rohanpaul_ai) · 2026-06-29
Meta has reportedly restricted engineers from using Claude Code and OpenAI Codex over concerns that rival model outputs could contaminate Meta's AI training data and violate contractual terms with Anthropic and OpenAI barring use of outputs to train competing models.
Extraction
Topics: ai-training-datamodel-distillationai-ipterms-of-servicecompetitive-ai
Claims
- Meta limited engineer use of Claude Code and Codex to prevent rival AI outputs from entering Meta's model training pipelines.
- Both OpenAI's and Anthropic's terms of service explicitly bar using their model outputs to develop competing models.
- AI distillation risk arises when a downstream model trains on another model's outputs, potentially constituting unauthorized capability extraction.
- Safe mitigation strategies include ingredient tracking, clean-room rules, training-data provenance logs, and access controls separating coding-agent outputs from training datasets.
- A strong legal case in this area would typically require evidence such as mass scraping, fake accounts, or internal records showing intentional cloning.
Key quotes
Both OpenAI's and Anthropic's terms bar using output to develop competing models.
Distillation risk starts when a new model of Meta learns from another model's outputs (from OpenAI or Anthropic), so even accidental reuse of Claude or Codex answers could look like Meta extracted capability from competitors rather than built it alone.
The safest strategy could be ingredient tracking: use rival tools for ordinary productivity only when outputs are barred from model-training pipelines, evaluation sets, benchmark generation, post-training data, reward-model data, and internal datasets that later feed model development.