Your coding agent’s ownership clause does not tell you whether your prompts or code may be used to improve a model, how long they may be retained, or which settings change that outcome. Those are separate questions, and the answer can vary by product, plan, account type, and organization agreement.
The comparison below reflects first-party guidance available as of October 7, 2026. Policies and plan access can change, so check the current terms and settings for the exact account you use before submitting proprietary material.
As an Amazon Associate I earn from qualifying purchases.
| Product and account covered | Ownership wording | Model-improvement use | Retention and controls |
|---|---|---|---|
| GitHub Copilot Free, Pro, and Pro+; policy announcement effective April 24, 2026 | GitHub says it does not claim ownership of inputs or outputs. | GitHub’s March 25, 2026 changelog says interaction data may be used unless the user opts out. The named categories include prompts, suggestions, code snippets, and associated context. | The cited announcement does not state a retention duration. It describes an opt-out; check current Copilot settings and terms. The announcement does not establish the same policy for Business or Enterprise. |
| Claude Free, Pro, and Max, including Claude Code consumer accounts | Ownership terms are not stated in the cited consumer guidance. | Use of chats and coding-session data for model improvement depends on the user’s setting. Incognito chats are excluded even if improvement is enabled. | When improvement is enabled, eligible data may be retained in de-identified form for up to five years in training pipelines. This consumer guidance does not establish a universal period for organization services; agreements and service may differ. |
| OpenAI individual services, including Codex; ChatGPT Business and API accounts are separate contexts | Individual terms say users retain rights in inputs and own outputs as between the user and OpenAI, subject to applicable law; outputs may not be unique. Business terms similarly state customer ownership of output. | Individual-service content may be used for training unless the user opts out. Business and API inputs and outputs are excluded by default, unless an organization explicitly opts in where available. | The cited materials do not state one general retention duration. ChatGPT training controls apply to Codex content, including Computer Use screenshots; Codex full-environment controls are managed separately. Business commitments depend on the applicable agreement. |
Does ownership mean the provider will not use my code?
No. Ownership language and data-use rules answer different questions. A provider may say that you own outputs or that it claims no ownership of inputs and outputs while separately describing how submitted material is processed, retained, or used to improve a service.
For OpenAI individual services, the Terms of Use say users retain rights in inputs and own outputs as between the user and OpenAI, subject to applicable law. The terms also caution that outputs may not be unique. OpenAI’s May 2025 Business Terms state that customers own output. GitHub’s March 25, 2026 changelog says, “GitHub does not claim ownership of your inputs or outputs.” These statements do not, by themselves, establish confidentiality, copyright status, trade-secret protection, or whether disclosure is permitted under your employer’s contracts.
Is my data used for model training?
It depends on the service, account, and setting. “Data” may include more than the code returned by an agent: prompts, generated suggestions, snippets, conversation history, context, and even screenshots can be within a policy’s scope.
GitHub Copilot individual plans
GitHub’s March 25, 2026 changelog announced that, starting April 24, 2026, interaction data from Copilot Free, Pro, and Pro+ users would be used to train and improve models unless users opted out. The announcement specifically lists inputs, outputs, code snippets, and associated context. It does not establish that the same rule applies to Copilot Business or Enterprise. GitHub’s Terms of Service describe AI feature terms as addressing data use for model development and improvement and the controls available, so consult the current terms and account settings for the rule that applies to your plan.
Claude consumer accounts and Claude Code
Anthropic’s consumer guidance covers Claude Free, Pro, and Max, including accounts using Claude Code. It says model-improvement use depends on the user’s settings and can involve the related conversation and associated content. Incognito chats are not used for improvement even when the setting is enabled. Organization accounts are governed separately by the applicable product terms and agreements.
OpenAI individual, Business, and API accounts
OpenAI says content from individual services, including Codex use, may be used to train models unless the user opts out. The Codex guidance says ChatGPT training controls apply to content processed through Codex, including screenshots captured by Computer Use. It also identifies full-environment Codex controls as a separate control area.
Rank #3
OpenAI says business-product and API inputs and outputs are not used for training by default, unless an organization explicitly opts in where available. Under OpenAI’s May 2025 Business Terms, customer content may be used as needed to provide the services, comply with law, enforce policies, and prevent abuse; the terms say it will not be used to develop or improve services unless the customer explicitly agrees. The applicable agreement matters: do not assume an individual-account setting or term governs an organization workspace.
How long do providers store coding-agent data?
Retention is not the same as training. A setting that prevents model-improvement use does not, on the evidence available here, establish how long data is kept for service delivery, security, abuse prevention, or legal obligations. Look for retention and deletion terms separately from training controls.
Anthropic’s consumer retention guidance gives a specific duration: if a user allows chats or coding sessions to be used to improve Claude, data may be retained in de-identified form for up to five years in model-training pipelines. The period applies to new or resumed chats after the setting is enabled. It is not a universal retention period for every Claude service. Anthropic describes organization retention separately; it can depend on the service and agreement, with zero-data-retention arrangements given as one example.
The Tool Desk
Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →Outbyte Driver Updater FREEScan for outdated or missing drivers - takes under a minuteDriver Scan →The cited GitHub and OpenAI materials do not establish a single general retention duration for the products and accounts compared here. Check the applicable service terms and any organization agreement rather than inferring a storage period from an ownership clause or training opt-out.
Best Value
Which settings or organization agreements change the result?
Start with the account you actually use, not the vendor name alone. Personal subscriptions, organization workspaces, and API accounts may have different default rules and administrators may control settings on managed accounts.
- For GitHub Copilot Free, Pro, or Pro+, check the current account controls for the opt-out described in GitHub’s April 24, 2026 policy announcement. For Business or Enterprise, verify the terms for that specific plan; the individual-tier announcement does not settle their treatment.
- For Claude consumer accounts, review the model-improvement setting and distinguish ordinary chats from Incognito chats. For organization use, ask which product terms, retention settings, and agreement apply.
- For OpenAI individual accounts using Codex, review ChatGPT’s data controls and Codex’s separate full-environment controls. For Business or API use, confirm the organization’s settings and whether it has explicitly opted in to model improvement where available.
OpenAI’s current ChatGPT Business help guidance says the same privacy commitments apply when a workspace uses Codex. It also notes that Codex seats ceased to be available to some new ChatGPT Business plans from June 24, 2026. That is a plan-entitlement change, not a general change to the privacy commitment.
Quick Recap
What to check before sending proprietary code
- Identify the exact product and account: note the coding agent, plan, whether you are signed in personally or through an organization, and the relevant terms’ date.
- Check what counts as interaction data: look for prompts, outputs, code snippets, conversation context, files, and screenshots—not only code the agent generates.
- Verify the training default and control: establish whether model-improvement use is on by default, whether an opt-out exists, and whether it applies to the agent surface you use.
- Read retention and deletion terms separately: find the stated duration, what triggers deletion, and whether an organization agreement changes the default.
- Ask your administrator or counsel when needed: regulated, confidential, or employer-owned code may be subject to rules that provider ownership language does not resolve.
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
