Does OpenAI Codex (ChatGPT personal plans (Plus / Pro, also Free / Go)) train on your code?
OpenAI · ChatGPT personal plans (Plus / Pro, also Free / Go) · checked 2026-10-19
Trains by default on personal plans. Turn off 'Improve the model' AND 'Include environments'.
Trains on code
Yes, unless you opt out
Default
OpenAI may train on content from services for individuals unless you opt out. 'Improve the model for everyone' covers Codex tasks. A separate 'Include environments' setting controls Codex environment context.
How to opt out
ChatGPT > Settings > Data controls > 'Improve the model for everyone' > Off (or privacy portal: 'Do not train on my content'). Then Codex settings > turn off 'Include environments'. Thumbs up/down can still send the whole conversation for training.
Retention
Not stated for Codex tasks on the pages checked. Temporary chats in ChatGPT: up to 30 days.
Telemetry
Codex CLI sends anonymous usage/health metrics by default (no PII per OpenAI). OTel log export is off by default.
Disable telemetry
In ~/.codex/config.toml: [analytics] enabled = false
Zero data retention
no · ZDR is an API/business feature.
Offline / local model
yes · Codex CLI: run with --oss and a local provider (oss_provider = "ollama" or "lmstudio").
Self-hostable
partial · CLI runs locally; cloud tasks run on OpenAI.
Score
48/100 · training 12/40 · retention 6/25 · offline 20/20 · telemetry 10/15
Classes: training Yes, unless you opt out · retention Not published · telemetry On, can disable · offline yes
Not published (not found in the vendor's official sources)
prompt_and_code_retention for Codex tasks