Overview
- Reports sourced to The Information and republished by other outlets say Meta has asked engineering teams to pause certain tasks that use Anthropic’s Claude Code and OpenAI’s Codex, though Meta has not publicly confirmed the directive.
- The reported reason is concern that outputs from third‑party models could be distilled into Meta’s training sets, which could create data‑governance, IP and partner‑relation problems for the company.
- Separately, reporting says Google has limited Meta’s access to Gemini capacity, highlighting supplier constraints as Meta scales its own compute resources.
- Meta is simultaneously accelerating a multibillion‑dollar buildout of proprietary AI infrastructure and weighing financing and business moves, including leasing data‑center capacity and potential ways to fund heavy 2026 AI spending.
- The changes update a longer shift at Meta that includes open‑sourcing Llama models, splitting operations with Manus, and recent EU measures on WhatsApp access, all of which shape how Meta balances openness, partnerships, and control of training data.