Windows Errors? Fix Them Before They Spread
Repair common Windows errors and clear accumulated junk for a smoother, more stable PC - no reinstall needed.Free scan · no reinstallCrashes, No Sound, or Screen Glitches?
Random freezes, missing sound and display glitches usually trace back to one bad driver. Find and replace yours safely.Free scan · under a minuteTo limit whether an AI coding assistant uses your work to improve models, identify your exact product, plan, and selected model, then change that product’s training or model-improvement setting. That setting does not necessarily stop your prompts or code context from being sent to a provider, retained, logged, or processed for safety and service operations. Review each of those data flows separately—and confirm your organization’s rules if you use a work account.
Start by identifying which assistant and account you use
Privacy rules can differ between an individual subscription and a business workspace, and between an IDE extension, a command-line tool, and a hosted service. The model matters too: an assistant may offer models hosted by different providers, and a personal API key can put processing under the provider’s terms rather than the assistant vendor’s usual arrangements.
- Write down the product surface (for example, IDE extension or CLI), account tier, workspace, and selected model.
- Check whether you selected a model using your own provider API key.
- For a managed account, ask your administrator which privacy settings, models, and agent capabilities are allowed.
Do not apply a consumer plan’s terms to a commercial workspace or API account. Anthropic distinguishes consumer Claude plans—including Claude Code used with those accounts—from commercial use; GitHub likewise distinguishes individual Copilot plans from Business and Enterprise. See Anthropic’s consumer training guidance and GitHub’s model-hosting guidance.
Understand which data flow each setting controls
“Not used for training” is not the same as “not processed.” An assistant may send prompt text and relevant code to a model provider so it can generate a response, even when those inputs are excluded from model training. Retention, operational telemetry, optional logging, feedback, and safety review can have their own rules.
The Tool Desk
Outbyte PC Repair FREEClear out junk files and repair common Windows errorsFree Scan →Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →#1 Best Overall
- Training or model improvement: whether interactions may be used to improve models.
- Inference: the prompt and code context sent to a model provider to produce an answer.
- Retention and logging: whether prompts, responses, or related records are stored, and for how long.
- Telemetry and feedback: service-use events or feedback such as a rating; these may be handled separately from prompt content.
- Safety and policy review: separate processing that may apply even if ordinary interactions are not used for model improvement.
For example, Google’s Gemini Code Assist Standard and Enterprise documentation defines Customer Data to include prompts, responses, conversation history, snippets of open and adjacent files, and cursor location. It says prompts and responses are not stored in Google Cloud by default, while customers can configure Cloud Logging to store inputs and responses. Google describes Service Data, including some telemetry, separately; listed examples include request or response events without request contents, response reactions, accepted-suggestion character counts, and UI interactions. Its statement that customer data is not used to train models without permission is specific to those editions, not a blanket statement about every Gemini product. See Google Cloud’s Gemini Code Assist security and privacy documentation.
Review the controls for your specific assistant
Cursor: enable Privacy Mode and check provider exceptions
- Open Settings → General → Privacy Mode. Cursor documents Cmd+Ctrl+Shift+J on Mac and Ctrl+Shift+J on Windows and Linux as shortcuts to open the relevant settings.
- Enable Privacy Mode if you want code excluded from training under Cursor’s stated policy. Cursor says prompts and code context are still sent to model providers when AI features are used.
- If you use a personal API key, check the provider’s privacy terms. Cursor says personal API key use follows the provider’s policy. Some models are outside Cursor’s zero-data-retention agreements, are off by default, and require admin approval.
- For a team account, ask whether an administrator enforces Privacy Mode and restricts models or agent permissions. Cursor documents organization-level enforcement, model restrictions, agent permissions, and audit logs.
Cursor describes Privacy Mode as preventing code from being used for training by Cursor or model providers; that promise does not mean no data leaves the editor. These are Cursor’s stated policies and controls, not an independent verification of implementation. Details are in Cursor’s privacy documentation.
GitHub Copilot: check the individual setting or business policy
GitHub says individual subscribers can manage whether Copilot interaction data—including prompts, suggestions, and code snippets—is used for model training in their account settings. Its documentation says opting out does not affect feature access. The cited documentation does not provide a stable full click-by-click path, so follow the settings link in GitHub’s current Copilot documentation rather than relying on an assumed menu location: GitHub Copilot.
For Copilot Business and Enterprise, GitHub says it does not use customer data to train AI models. Still check the hosting and processing details for the model selected: GitHub’s reference distinguishes providers and hosting arrangements, and data processing depends on the selected model. Consult GitHub’s model-hosting documentation and have an administrator confirm the applicable workspace policy.
Claude and Claude Code: distinguish model improvement, Incognito, and feedback
For Claude Free, Pro, and Max accounts—including those using Claude Code—Anthropic’s consumer guidance says chats and coding sessions may be used for model improvement when the user opts in, when conversations are flagged for safety review for safety purposes, or through another explicit opt-in. Commercial users are subject to separate terms, so do not assume this consumer guidance applies to Claude for Work or API use.
Anthropic says Incognito chats are not used to improve Claude, even if Model Improvement is enabled. That does not make the setting a universal deletion or no-processing control. A thumbs-up or thumbs-down can include the related conversation and may be retained for up to five years. Anthropic’s retention guidance also says opted-in data may be retained in de-identified form for up to five years in model-training pipelines. For policy-flagged sessions, it describes retention of inputs and outputs for up to two years and trust-and-safety classification scores for up to seven years. These are distinct retention cases, not one general retention period. Read Anthropic’s model-training guidance and retention guidance.
Rank #4
Gemini Code Assist Standard and Enterprise: check logging and IAM as well as training
Google’s documentation says it does not use customer data to train models without permission for Gemini Code Assist Standard and Enterprise. It also says the service is stateless and does not store prompts and responses in Google Cloud by default; an organization can configure Cloud Logging to store inputs and responses. Ask your administrator whether logging is enabled and who can access those logs. Google says processing generally occurs near the request origin, but regionality is not guaranteed.
For managed access, Google documents IAM access management. Ask the administrator to confirm who can use Gemini Code Assist and whether the organization’s logging and data-handling settings match its policy. The product-specific details are in Google Cloud’s security and privacy documentation.
Best Value
Use this setup checklist before sharing real code
- Find the data controls. In the product’s official documentation, search for labels such as privacy, data controls, model improvement, training, or telemetry. Settings and model lists can change, so use the current product instructions.
- Choose a training preference. Turn off model-improvement use if that is your choice, or use a private or Incognito mode where available. Check what the control covers and whether safety review, feedback, or previously processed data has separate treatment.
- Review the context sent with a request. Check whether the assistant can include open files, nearby files, conversation history, or other editor context. Avoid sending credentials, secrets, regulated data, or proprietary code until your employer’s policy and the relevant provider terms allow it.
- Check retention and telemetry independently. Determine whether prompts and responses are stored, whether logging is enabled, what telemetry is collected, and whether submitting feedback attaches the related conversation.
- Get work-account controls confirmed. Ask an administrator about allowed models, organization-level privacy settings, agent permissions, logging and audit access, and the agreement covering the account.
- Recheck when your setup changes. Repeat the review after changing plans, choosing another model, adding a provider API key, or enabling a new IDE or agent feature.
How to compare assistants without relying on a “private” label
When choosing between products or plans, compare the actual controls and terms for the account and model you will use. A single privacy label can hide important differences:
- Whether training use is enabled by default or controlled by the user or administrator.
- Which prompts, files, and other context are sent to providers for inference.
- Whether prompts and responses are retained, and whether optional logging changes that.
- What telemetry and feedback include, and how safety review is handled.
- Whether a provider-hosted model or personal API key has separate terms or retention.
- Whether administrators can enforce settings, restrict models, manage agent permissions, and inspect audit records.
Base the decision on the vendor documentation and your organization’s agreement for the exact tier and selected model—not on another provider’s privacy commitments.
Quick Recap
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.
