MorseSign inStart freeOpen Morse

What does 'no logs' mean for an AI chatbot? Six apps' claims, and what each still keeps

The Morse teamPublished 7 min read

A "no logs" claim from an AI chatbot means the company says its own servers don't keep your prompts or its replies once it has answered. As of November 2026, five of the six apps we checked that make such a claim still keep something else, such as your account, analytics, reported chats, or copies held by the model provider. Meta's Incognito chat sends search queries out instead, and none of the six has published an independent audit of the claim for its chat.

We read the privacy pages of Proton's Lumo, DuckDuckGo's Duck.ai, Brave Leo, Venice, Kagi Assistant, and Meta AI's Incognito chat on November 17, 2026, and quote each company's own words below. Kagi and DuckDuckGo deserve credit for listing each exception with its time limit, and Proton names the metadata Lumo doesn't store, down to timestamps and IP addresses.

What does "no logs" mean for an AI chatbot?

It means the company promises not to write your conversation to its own storage after it answers. The term comes from VPNs, where it means the provider keeps no record of what you do online. DuckDuckGo's VPN policy puts it as "We don't keep logs of your VPN activity" (DuckDuckGo).

The app still has to read your message to answer it. Proton says Lumo's server "decrypts the message temporarily to process it through the language model" (Proton). A no-logs promise covers what happens after that moment, and only on the company's own machines.

Which AI chatbots claim no logs, and what do they still keep?

Each row quotes the company's own page, read November 17, 2026.

AppIts wordsWhat it still keepsIndependent audit of the chat claimSource
Proton Lumo"we keep no logs of what you ask, or what Lumo replies"Your account email; the last 4 digits of your card on paid plans; history on Proton's servers under zero-access encryption; an anonymized copy of any feedback you sendNone published for LumoProton, policy
Duck.ai"does not record or store any of your chats"Chats on your device; topic categories for a random sample of messages; feedback you share; model providers may keep prompts "at most within 30 days"None published for Duck.aiDuckDuckGo, terms
Brave Leo"We don't store or retain prompts, responses, context, or personal data on our servers"Large prompts cached for minutes; feedback with the full conversation for 1 year; a query's broad category, language, and modelNone foundBrave
Venice"we will not have access to or store the content of your Prompts or Outputs"Your actions, such as "that you created a chat"; device details including your IP address; Google Analytics data; in Anonymous mode, whatever the provider keepsNone foundVenice, privacy page
Kagi Assistant"requests to AI tools are not saved or associated with a customer account"Threads for one day by default; reported threads for 30 days; debugging logs for 7 to 90 days; model providers' copies for 24 hours to 30 days, or 2 years for prompts Anthropic flagsNone foundKagi, LLM privacy
Meta AI, Incognito chat"Meta can't access it," and chats "won't be saved, stored"Search queries sent to "select search engines" by default; the help page lists nothing elseNCC Group reviewed a different feature on the same systemMeta, NCC Group

Lumo goes furthest on metadata. Proton says Lumo doesn't store "timestamps, IP addresses, or chat context" (Proton).

Does "no logs" cover the model provider?

Only when the app runs the model itself. Proton says Lumo's models run "exclusively on servers Proton controls" (Proton). Kagi, Duck.ai, and Venice send many prompts to other companies' models, and those companies have their own rules.

Kagi's LLM privacy page lists how long each provider keeps prompts sent through Kagi, from 24 hours for Gemini to 30 days for GPT and Claude, and 2 years for Claude prompts Anthropic flags (Kagi); we cover the rest of Kagi's setup in is Kagi Assistant private, and how providers justify those copies in what is AI abuse monitoring.

Venice says it plainly for its Anonymous mode. Its privacy page says "you should assume the provider is storing your content" (Venice). Its privacy policy describes "a zero data retention policy with our model providers" (Venice), and its privacy page ties that to Private mode, where "no prompt or response is stored."

Duck.ai's help page says providers delete prompts "immediately after generating a reply," while its terms, updated November 16, 2026, allow "at most within 30 days, with limited exceptions for safety and legal compliance" (DuckDuckGo, help page); we compare Duck.ai's relay with Lumo and Brave Leo in Duck.ai vs Lumo vs Brave Leo, and explain these agreements in what is zero data retention.

What data does a no-logs AI chatbot still collect?

Usage data, mostly. Duck.ai matches each prompt to a similar prompt from a public dataset of chatbot conversations, and keeps the match "long-term" (DuckDuckGo). DuckDuckGo says the embedding of your own prompt is "never written to disk." Brave uses its STAR and Nebula systems to learn "the broad category of a user query, the language the query is written in, and the AI model used" (Brave).

Feedback is the common exception. Brave keeps the "full text of the current conversation" you rate for 1 year. Kagi keeps an Assistant thread you report for 30 days, where its team can read it (Kagi).

Operational logs sit outside the chat claim too. Kagi keeps load balancer and virtual machine logs for 7 days and sampled error reports for 90 days, not linked to an account (Kagi). Venice collects the IP address, time zone, and browser type of people who use it without an account, receives "IP address, device type, device identifiers" and other device details, and uses Google Analytics (Venice). Brave, by contrast, says "We do not collect identifiers that can be linked to you (such as IP address)" (Brave). We compare what 14 apps collect beyond the chat in what data AI chatbots collect.

Has any AI chatbot's no-logs claim been audited?

Not that we found. Two of these companies have had outside audits of their VPNs. DuckDuckGo says Securitum inspected its VPN between October 2025 and January 2026 and "validated our no-logs policy" (DuckDuckGo). Proton says Proton VPN has passed "a fifth consecutive annual third-party audit" of its no-logs policy (Proton VPN). Neither company's AI chat pages point to a no-logs audit of Lumo or Duck.ai.

Meta comes closest. NCC Group's August 2025 assessment covered WhatsApp's message summaries, part of the Private Processing system that Incognito chat uses (NCC Group). NCC Group found that users must still trust Meta "to not insert any malicious behavior in artifacts that are not open source." The report doesn't cover Incognito chat itself.

Proton's Lumo policy says it discloses "the limited user data we possess on a Lumo account" only under "a binding request coming from the competent authorities" (Proton). Kagi offers a different signal. Its warrant canary listed "0 Warrants from any government organization" when we read it (Kagi).

How can you check a no-logs AI chat claim yourself?

Not from the outside, since a no-logs claim describes servers you never see, but you can read the right pages:

  1. Find the list of exceptions. Look for feedback, reported chats, debugging logs, and analytics, as in the table above.
  2. Find out who runs the model. If it's another company, read that company's retention, as Kagi lists it.
  3. Look for an audit report you can open. DuckDuckGo and Proton link theirs for their VPNs.
  4. Check where your history lives. Lumo, Duck.ai, and Venice (Venice) keep it on your device or under encryption they say they can't open; our guide to AI that doesn't save your data compares them.

Enclave-based chats change the question, because your own browser can check what the server runs. In Venice's E2EE mode, on its Pro plan, the prompt is encrypted on your device and "only decrypted inside a verified TEE" (Venice). We walk through those checks in how to verify AI enclave attestation.

How Morse handles what a no-logs claim leaves out

A no-logs promise covers one company's servers, on that company's word. The model provider's copy, your account, and your stored history fall outside it. Morse's servers don't keep messages or log message content. Your account holds your sign-in details, plan, balance, and a usage ledger with no content. Your chat history is stored in your browser, encrypted, not on Morse's servers (Morse privacy policy).

What the model's host keeps depends on the privacy mode you pick for each chat:

  • Private, so the model host has agreed to keep nothing too. It's the default for a new chat. Open-weight models like DeepSeek, Kimi, and GLM, plus xAI's Grok, run here. Morse has zero data retention agreements with its inference, search, and other providers, so they've agreed to keep nothing. Where a provider confirms zero retention on each reply, Morse checks it and refuses a reply that doesn't.
  • E2EE, so your own browser checks the server before you send anything. Your browser checks the enclave's proof that it's genuine before anything leaves your device, then encrypts the message for that enclave alone. Morse's servers relay text they can't read. Every E2EE reply has a lock you can open to see each check that passed, and a report you can download. The enclaves are run by third-party providers, not by Morse, and the guarantee rests on the enclave hardware and attestation working as designed. E2EE runs open-weight models with no web search or image generation, and comes with Pro, at $19 a month.
  • Anonymous, for GPT, Claude, and Gemini. Morse's proxy removes your account, email, and IP address and sends the message under Morse's own account. The provider sees the message without knowing it's you, and may keep it for a limited time to check for abuse, under its own policy.

Morse doesn't train AI on your chats, sell your data, or show ads. Because it doesn't store chats, Morse usually has very little to give authorities. Clearing your browser or losing the device deletes your chats, and Morse can't recover them, so keep a copy of anything you need. For more on legal requests, see can AI chats be subpoenaed.

Start free with an emailed code, and your first chat opens in Private mode.

Common questions

Is 'no logs' the same as zero data retention?

No. A no-logs claim is about the chat app's own servers, while zero data retention is an agreement between the app and the model provider that answers. Venice and Duck.ai make both promises, and their pages show the provider half has its own exceptions.

Does a no-logs AI chatbot keep my IP address?

It depends on the app. Proton says Lumo doesn't store IP addresses and Brave says Leo doesn't collect them, while Venice's privacy policy says it collects the IP address of people who use it without an account.

Is a no-logs AI chat end-to-end encrypted?

Usually not. Proton says Lumo's server decrypts each message 'temporarily' to run it through the model, so a no-logs promise covers what happens afterwards. Venice offers an E2EE mode on its Pro plan, where only a verified enclave decrypts the prompt.

What can a no-logs AI chatbot give the police?

Whatever it keeps. Proton's Lumo policy says it discloses 'the limited user data we possess' when a binding legal request requires it, and Kagi's warrant canary said '0 Warrants' when we read it.

Ask what you actually want to ask.

80+ models, characters, and images, with every chat encrypted and kept on your device. Free to start.

Start a private chat

Read next