DeepMind's Sholto: 'extremely unlikely' Codex trained on user transcripts
daniel_mac8 · x · 2026-09-08
Amid speculation that user data influenced model behavior, Google DeepMind researcher Sholto Douglas pushed back: based on everything visible from outside, it is extremely unlikely OpenAI pulled user transcripts or knowingly trained on them in a way that affected this. He stressed people shouldn't run away with the narrative that 'your user data isn't safe in Codex.'
Related event: OpenAI Disputes Navier-Stokes Credit and Codex Data Privacy Allegations(38 posts)→
More from Models
- Blender head-to-head: same prompt, 12 seconds, and one frontier model is in a different class — ZeroStateReflex · 2026-09-09
- Netlify adds GPT-6 Astra, Gemini 3.8 Flash, Claude Fable 5.1 and smarter Agent Runner scoping — thisiskp_ · 2026-09-09
- GLM-5.3-Flash tops agentic tool-call leaderboard at 78%, priced at just $0.50/M output tokens — shensi · 2026-09-09
- Artificial Analysis Updated Benchmarks Twice in 4 Days for Astra — py-net · 2026-09-09
- Gary Marcus: it was the harness, not the model — hidden risk of opaque scaffolding — GaryMarcus · 2026-09-09
- Inception ships Mercury 2.5: most capable diffusion LLM at 1,107 tokens/sec, 260K context — StefanoErmon · 2026-09-09