Does Sourcegraph Cody train on your data?
Sourcegraph · AI coding assistant · official site ↗
Doesn’t train on your dataZero-retention availableNo training to opt out of
Does Sourcegraph Cody train its models on your data?
No — Sourcegraph Cody does not use your content to train its AI models by default.
Sourcegraph does not train its own or its providers' LLMs on your code. Cody sends code snippets to third-party model providers (Anthropic, OpenAI) only to generate responses, under zero-retention agreements; Cody is now the Sourcegraph Enterprise code-intelligence assistant.
Can you opt out?
No training opt-out is needed because Sourcegraph does not train on customer code by default and its third-party LLM providers do not train on your codebase. Enterprises can further restrict by bringing their own LLM keys (Azure OpenAI, Amazon Bedrock) to keep inference inside their own cloud.
Zero retention / DPA
By default Cody accesses LLMs via Anthropic and OpenAI APIs with zero retention on both inputs and outputs; the same applies through the Sourcegraph Model Provider (Cody Gateway). For autocomplete via Fireworks.ai, no customer chat or autocomplete data is stored. Cody Enterprise is SOC 2 Type II, GDPR and CCPA compliant with a zero-data-retention and no-training posture. DPA ↗
What the listicles get wrong
Cody Free and Pro were retired on 23 July 2025; Cody now ships as the Sourcegraph Enterprise assistant (Sourcegraph also points users to Amp for agentic coding). Retention specifics for Anthropic/OpenAI are governed by the referenced Cody notice/terms.
Verdict by plan tier
Enterprise (Sourcegraph Enterprise)No trainingSourcegraph will not train on your company's data and its third-party LLM providers do not train on your specific codebase; snippets are used solely to generate responses under zero-retention agreements.
Get notified when this changes
We track Sourcegraph Cody's data-training and retention policy. Leave your email and we'll send one note if it changes.
Frequently asked questions
Does Sourcegraph Cody train its AI models on my data?
No — Sourcegraph Cody does not use your content to train its AI models by default. Sourcegraph does not train its own or its providers' LLMs on your code. Cody sends code snippets to third-party model providers (Anthropic, OpenAI) only to generate responses, under zero-retention agreements; Cody is now the Sourcegraph Enterprise code-intelligence assistant.
Can I opt out of Sourcegraph Cody training on my data?
There is no training opt-out to set for Sourcegraph Cody: No training opt-out is needed because Sourcegraph does not train on customer code by default and its third-party LLM providers do not train on your codebase. Enterprises can further restrict by bringing their own LLM keys (Azure OpenAI, Amazon Bedrock) to keep inference inside their own cloud.
Does Sourcegraph Cody offer zero data retention (ZDR) or a DPA?
By default Cody accesses LLMs via Anthropic and OpenAI APIs with zero retention on both inputs and outputs; the same applies through the Sourcegraph Model Provider (Cody Gateway). For autocomplete via Fireworks.ai, no customer chat or autocomplete data is stored. Cody Enterprise is SOC 2 Type II, GDPR and CCPA compliant with a zero-data-retention and no-training posture.
Is Sourcegraph Cody safe to use with confidential or proprietary data?
It depends on your plan tier. Enterprise (Sourcegraph Enterprise): Sourcegraph will not train on your company's data and its third-party LLM providers do not train on your specific codebase; snippets are used solely to generate responses under zero-retention agreements. Always confirm current terms with Sourcegraph before sending confidential data — this is cited public information, not legal advice.
Sources
https://sourcegraph.com/docs/cody/faq
https://sourcegraph.com/blog/cody-is-enterprise-ready
https://about.sourcegraph.com/terms/cody-notice
This page is cited public information, not legal or compliance advice. Whether Sourcegraph Cody trains on your data, and any zero-retention or DPA option, can depend on your specific plan, region and contract. Always confirm current terms with Sourcegraph before sending confidential or proprietary data.