Claude Code is Anthropic's coding assistant, reviewed for code context, model-improvement settings, sharing, retention, and developer controls.
75
Good
75
Good
Claude Code inherits Anthropic's controls, but repository access and integration scope matter more than the homepage suggests.
Claude Code is strongest where users and teams can limit access through account, workspace, and plan-level controls rather than broad public-sharing defaults.
The main risk is not ad-tech alone but how much repository, terminal, and integration context the coding assistant can see while working.
75
Good
Anthropic documents model-improvement controls and enterprise or work privacy distinctions.
Users can opt out of having inputs and outputs used to train models through account settings.
Anthropic separately documents how third-party training data and model-development uses are handled, instead of hiding them in one paragraph.
75
Good
65
Mixed
70
Mixed
80
Good
Anthropic Privacy Page
Open: anthropic.com1. AI Use
Anthropic documents model-improvement controls and enterprise or work privacy distinctions.
2. AI Use
Users can opt out of having inputs and outputs used to train models through account settings.
3. AI Use
Anthropic separately documents how third-party training data and model-development uses are handled, instead of hiding them in one paragraph.
4. Controls
Business or employer-provided contexts can move more control to the commercial customer than to the individual user.
5. AI Use
Anthropic may use inputs and outputs to train models unless the user opts out.
6. Code Prompts
Code, prompts, and terminal context can still be processed to provide coding assistance.
Founded
Unknown
Founder
Unknown
Parent Company
Claude Code
Lifecycle
Active
Category
AI Developer Tools
CEO
Unknown
Security Team
In house
Date Added
05-18-2026
Once you delete a chat, you cannot recover it. Deleting a chat removes it both from your visible chat history and the system after the retention window.
Business or employer-provided contexts can move more control to the commercial customer than to the individual user.
Anthropic may use inputs and outputs to train models unless the user opts out.
Code, prompts, and terminal context can still be processed to provide coding assistance.
75
Good
Anthropic documents service-provider, privacy-rights, and security handling.
The policy distinguishes commercial-customer processor contexts from consumer-controller contexts.
Users and organizations can choose whether to connect many third-party websites and services.
International-transfer safeguards and privacy-rights disclosures are documented directly in the policy.
Data can still flow through service providers, business partners, affiliates, connected services, and legal recipients.
Third-party training data sources and connected coding or agentic sessions widen the practical sharing surface.
65
Mixed
Anthropic still collects usage information, device data, and cookies or similar technologies.
Marketing and business-partner contexts still add some tracking and promotional data use around the service.
Anthropic says it does not sell personal data as defined by applicable laws.
Users can access privacy settings and rights tools.
Claude Code is not a public social feed or ad-delivery product.
70
Mixed
Retention still depends on product context, plan, safety, legal duties, and account state.
Connected sessions and copied code or logs can remain outside a single workspace lifecycle.
Anthropic documents deletion, access, and portability rights.
Privacy settings give users some ability to manage future data use.
Commercial-customer contexts can place more lifecycle control with the organization than with consumer accounts.
The policy separates some non-user training data questions into a dedicated policy instead of hiding them entirely.
80
Good
Developers can limit repository access and use privacy controls.
Users can opt out of model training through account settings.
Anthropic documents access, portability, correction, and deletion rights.
Organizations can govern access in work or enterprise contexts more centrally than unmanaged personal use.
Privacy still depends on not exposing secrets and on workspace or plan settings.
Connected third-party services and agentic sessions can widen the practical surface beyond one simple toggle.
80
Good