AI assistant privacy ratings
AI chatbots and assistants, hosted and local, rated on how they keep prompts and whether they train models on them.
18 AI assistants are rated against 13 public criteria, and 18 have enough evidence for a letter grade. Of the 12 with a known jurisdiction, 10 are based in a Five Eyes country, and 11 in the wider Fourteen Eyes.
Open-source AI assistants only · Criteria for this category · Suggest an addition · Markdown
| Name | Grade | Score | Data | Jurisdiction | Open source | No trackers or telemetry | No ads or data sales | Independent audit | Transparency report | Tells users about requests | No training on your data | Runs locally | Limited chat retention | No account needed | TLS configuration | Security headers | Modern web standards |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
Open-source desktop app for running large language models offline on a local computer, with optional connections to remote AI APIs. |
Grade B | 89 | 100% | – | Yes | Yes | Yes | No | Not applicable | Not applicable | Yes | Yes | Yes | Yes | Not applicable | Not applicable | Not applicable |
Open-source C/C++ library and command-line tools for running large language model inference locally on CPUs and GPUs, including a built-in web server and chat interface. |
Grade B | 89 | 100% | – | Yes | Yes | Yes | No | Not applicable | Not applicable | Yes | Yes | Yes | Yes | Not applicable | Not applicable | Not applicable |
Open-source tool for downloading and running large language models on a local computer through a command line, API and desktop app. Optional cloud-hosted models are also offered. |
Grade B | 81 | 100% | – | Yes | Partial | Yes | No | Not applicable | Not applicable | Yes | Yes | Yes | Yes | Not applicable | Not applicable | Not applicable |
AI assistant from Proton that runs open-weight models on Proton's servers, with no logs of chats and zero-access encrypted chat history. |
Grade C | 73 | 100% | Switzerland Outside Eyes | Partial | Partial | Yes | Partial | Yes | Yes | Yes | No | Yes | Partial | Yes | Yes | Not tested yet |
Open-source desktop app from Nomic for running large language models locally, including chatting with local documents. |
Grade C | 72 | 100% | – | Yes | No | Yes | No | Not applicable | Not applicable | Yes | Yes | Yes | Yes | Not applicable | Not applicable | Not applicable |
Desktop app for finding, downloading and running large language models locally, with a chat interface and a local API server. |
Grade C | 72 | 100% | – | No | Yes | Yes | No | Not applicable | Not applicable | Yes | Yes | Yes | Yes | Not applicable | Not applicable | Not applicable |
Self-hosted web interface for large language models that works with Ollama and OpenAI-compatible APIs, with multi-user support. |
Grade C | 72 | 100% | – | Yes | No | Yes | No | Not applicable | Not applicable | Yes | Yes | Yes | Yes | Not applicable | Not applicable | Not applicable |
AI assistant built into the Brave browser for chat, page summaries and writing help. Works without an account and can also use local or self-chosen models. |
Grade C | 61 | 100% | United States Five Eyes | Partial | Partial | Partial | No | Not applicable | Not applicable | Yes | Partial | Yes | Yes | Not applicable | Not applicable | Not applicable |
AI chat from DuckDuckGo that relays prompts to third-party models without identifying the user, with no account needed and recent chats saved only on the device. |
Grade C | 60 | 100% | United States Five Eyes | Partial | Yes | Partial | No | Partial | Yes | Yes | No | Yes | Yes | Partial | No | Not tested yet |
AI assistant from Kagi that combines several third-party language models with Kagi Search results, available with a Kagi account. |
Grade D | 54 | 100% | United States Five Eyes | No | Yes | Yes | No | Partial | No | Yes | No | Yes | No | Partial | Yes | Not tested yet |
AI assistant from Anthropic for writing, analysis, coding and research, available on the web and as desktop and mobile apps, with free and paid plans. |
Grade D | 44 | 100% | United States Five Eyes | No | No | Yes | No | Yes | Yes | Partial | No | Yes | No | Yes | No | Not tested yet |
AI chatbot from OpenAI for writing, questions, coding, image generation and voice conversations. Available on the web and as desktop and mobile apps, with free and paid plans. |
Grade D | 42 | 100% | United States Five Eyes | No | No | No | Partial | Yes | Partial | Partial | No | Yes | Partial | Yes | Partial | Not tested yet |
AI chat and image generation service that proxies prompts to open and third-party models, keeping conversation history only in the browser. |
Grade D | 42 | 100% | United States Five Eyes | No | No | Yes | No | No | No | Yes | No | Yes | Partial | Yes | Partial | Not tested yet |
AI assistant from Google for questions, writing, coding and image generation, connected to Google services such as Gmail and Drive. Available on the web and in Android and iOS apps. |
Grade F | 35 | 100% | United States Five Eyes | No | No | Yes | No | Yes | Yes | Partial | No | No | Partial | Partial | Not tested yet | Not tested yet |
AI assistant from French company Mistral AI, formerly called Le Chat, for chat, research, documents and coding tasks. Available on the web and in mobile apps. |
Grade F | 33 | 100% | France Nine Eyes | No | No | Yes | Partial | No | No | Partial | No | Partial | No | Yes | Partial | Not tested yet |
Consumer AI assistant from Microsoft for chat, search, writing and image generation, built into Windows and Edge and available on the web and in mobile apps. |
Grade F | 29 | 100% | United States Five Eyes | No | No | No | No | Yes | Yes | Partial | No | Partial | Partial | Partial | No | Not tested yet |
AI assistant from Meta built into Facebook, Instagram, WhatsApp and Messenger, and available on the web and as a standalone app. |
Grade F | 27 | 100% | United States Five Eyes | No | No | No | No | Yes | Yes | No | No | Partial | Partial | Yes | No | Not tested yet |
AI answer engine that searches the web and summarizes results with cited sources, available on the web and as desktop and mobile apps. |
Grade F | 27 | 100% | United States Five Eyes | No | No | Partial | No | No | No | Partial | No | Partial | Partial | Yes | Partial | Not tested yet |
Questions
What is the most private option among AI assistants?
No pick has been made yet. The table above is sorted by score, based on public evidence.
How are AI assistants rated?
Each entry answers 13 questions: open source, no trackers or telemetry, no ads or data sales, independent audit, transparency report, tells users about requests, TLS configuration, security headers, modern web standards, no training on your data, runs locally, limited chat retention and no account needed. Answers need links to evidence. See the full criteria.
Does jurisdiction matter?
Jurisdiction decides which laws can compel a provider to hand over data. Each entry shows its country and whether it is in the Five, Nine or Fourteen Eyes. The data a provider can hand over depends mostly on what it stores and who holds the keys. Read about jurisdictions and the CLOUD Act.