Meta restricts use of Claude Code and Codex to keep rival AI out of its training data
Quick Answer
Meta is restricting its engineers' use of Claude Code and Codex to prevent rival AI outputs from contaminating its training data, amid concerns over unauthorized capability transfer.
Quick Take
Meta is restricting its engineers' use of Claude Code and Codex to prevent rival AI outputs from contaminating its training data, amid concerns over unauthorized capability transfer. The company is investing billions in its own coding assistant, MetaCode, while enforcing strict policies against using external AI outputs for internal tasks.
Key Points
- Meta has temporarily halted work with Claude Code and Codex.
- Concerns over distillation could escalate relationships with partner companies.
- Meta plans to spend billions on internal AI development this year.
- Company policy prohibits using AI outputs for test tasks or code analysis.
- Distillation issues are causing friction across the AI industry.
📖 Reader Mode
~1 min readMeta is limiting how its engineers use Anthropic's Claude Code and OpenAI's Codex to prevent outputs from these AI tools from ending up in its own training data. According to internal documents obtained by The Information, Meta has even temporarily halted certain work with these models.
The company is worried about distillation - the unauthorized transfer of capabilities from rival AI models. An internal memo warned of serious escalations with partner companies if their model outputs were to leak into Meta's training data.
Meta is currently building out its own coding assistant, MetaCode, and wants to cut its reliance on outside tools, partly because of rising costs. According to an internal memo, the company is on track to spend billions of dollars on internal AI use this year alone. Company policy bars engineers from using AI outputs to create test tasks or for code analysis. Human review is still required.
Distillation is causing friction across the industry. Anthropic recently accused Alibaba of the largest known distillation attack to date, and Elon Musk had to admit in April that xAI had partially distilled OpenAI's models. Terms of service from OpenAI, Anthropic, and Google all explicitly ban using model outputs to build competing systems. Meta said it has clear rules for the responsible use of AI tools.
— Originally published at the-decoder.com
Want this in your inbox every morning?
Daily brief at your local 8am — bilingual EN/中文, free.
More from The Decoder
See more →
An AI model programmed nonstop for 19 days on a single MirrorCode task that cost $2,600 to run
Epoch AI's MirrorCode benchmark reveals Claude Opus 4.7 as the leader with a 56% solve rate, reconstructing a 16,000-line toolkit in 14 hours. Despite this, all models tested struggle with the most complex tasks, highlighting limitations in current AI capabilities. The single task consumed $2,600 over 19 days, raising questions about cost-effectiveness in AI development.
