1# Zed AI Improvement
  2
  3## Agent Panel
  4
  5### Opt-In
  6
  7When using the Agent Panel, whether through Zed's hosted AI service or via connecting a non-Zed AI service via API key, Zed does not persistently store user content or use user content to evaluate and/or improve our AI features, unless it is explicitly shared with Zed. Each share is opt-in, and sharing once will not cause future content or data to be shared again.
  8
  9> Note that rating responses will send your data related to that response to Zed's servers.
 10> **_If you don't want data persisted on Zed's servers, don't rate_**. We will not collect data for improving our Agentic offering without you explicitly rating responses.
 11
 12When using upstream services through Zed AI, we require assurances from our service providers that your user content won't be used for training models. For example, usage of Anthropic Claude 3.5 via Zed AI in the Assistant is governed by the [Anthropic Commercial Terms](https://www.anthropic.com/legal/commercial-terms) which includes the following:
 13
 14> "Anthropic may not train models on Customer Content from paid Services."
 15
 16We also have a [zero-data retention agreement](https://privacy.anthropic.com/en/articles/8956058-i-have-a-zero-data-retention-agreement-with-anthropic-what-products-does-it-apply-to) with Anthropic.
 17
 18When you directly connect Zed with a non Zed AI service (e.g., via API key) Zed does not have control over how your data is used by that service provider.
 19You should reference your agreement with each service provider to understand what terms and conditions apply.
 20
 21### Data we collect
 22
 23For prompts you have explicitly shared with us, Zed may store copies of those prompts and other data about the specific use of the Agent Panel.
 24
 25This data includes:
 26
 27- The prompt given to the Agent
 28- Any commentary you include
 29- Product telemetry about the agentic thread
 30- Metadata about your Zed installation
 31
 32### Data Handling
 33
 34Collected data is stored in Snowflake, a private database where we track other metrics. We periodically review this data to improve our overall agentic approach and refine the product via our system prompt, tool use, etc. We ensure any included data is anonymized and contains no sensitive information (access tokens, user IDs, email addresses, etc).
 35
 36## Edit Predictions
 37
 38By default, when using Zed Edit Predictions, Zed does not persistently store user content or use user content for training of its models.
 39
 40### Opt-in
 41
 42Users who are working on open source licensed projects may optionally opt-in to providing model improvement feedback. This opt-in occurs on a per-project basis. If you work on multiple open source projects and wish to provide model improvement feedback you will have to opt-in for each individual project.
 43
 44When working on other projects where you haven't opted-in, Zed will not persistently store user content or use user content for training of its models.
 45
 46You can see exactly how Zed detects open source licenses in: [license_detection.rs](https://github.com/zed-industries/zed/blob/main/crates/zeta/src/license_detection.rs).
 47
 48### Exclusions
 49
 50Zed will intentionally exclude certain files from Predictive Edits entirely, even when you have opted-in to model improvement feedback.
 51
 52You can inspect this exclusion list by opening `zed: open default settings` from the command palette:
 53
 54```json
 55{
 56  "edit_predictions": {
 57    // A list of globs representing files that edit predictions should be disabled for.
 58    // There's a sensible default list of globs already included.
 59    // Any addition to this list will be merged with the default list.
 60    "disabled_globs": [
 61      "**/.env*",
 62      "**/*.pem",
 63      "**/*.key",
 64      "**/*.cert",
 65      "**/*.crt",
 66      "**/secrets.yml"
 67    ]
 68  }
 69}
 70```
 71
 72Users may explicitly exclude additional paths and/or file extensions by adding them to [`edit_predictions.disabled_globs`](https://zed.dev/docs/configuring-zed#edit-predictions) in their Zed settings.json:
 73
 74```json
 75{
 76  "edit_predictions": {
 77    "disabled_globs": ["secret_dir/*", "**/*.log"]
 78  }
 79}
 80```
 81
 82### Data we collect
 83
 84For open source projects where you have opted-in, Zed may store copies of requests and responses to the Zed AI Prediction service.
 85
 86This data includes:
 87
 88- the edit prediction
 89- a portion of the buffer content around the cursor
 90- a few recent edits
 91- the current buffer outline
 92- diagnostics (errors, warnings, etc) from language servers
 93
 94### Data Handling
 95
 96Collected data is stored in Snowflake, a private database where we track other metrics. We periodically review this data to select training samples for inclusion in our model training dataset. We ensure any included data is anonymized and contains no sensitive information (access tokens, user IDs, email addresses, etc). This training dataset is publicly available at [huggingface.co/datasets/zed-industries/zeta](https://huggingface.co/datasets/zed-industries/zeta).
 97
 98### Model Output
 99
100We then use this training dataset to fine-tune [Qwen2.5-Coder-7B](https://huggingface.co/Qwen/Qwen2.5-Coder-7B) and make the resulting model available at [huggingface.co/zed-industries/zeta](https://huggingface.co/zed-industries/zeta).
101
102## Applicable terms
103
104Please see the [Zed Terms of Service](https://zed.dev/terms-of-service) for more.