Privacy Policy
Effective 3 November 2026
Nemo is a personal AI assistant made by ASI Labs Pte. Ltd. ("ASI Labs", "we"), a company registered in Singapore. This policy explains what information Nemo collects, what it does with it, who else sees it, how long it is kept, and the choices you have. It applies to the Nemo service at heynemo.ai, the Nemo desktop and mobile apps, and Nemo on Telegram, LINE, Slack, WhatsApp and email. Expert Q&A at heynemo.ai/expert has its own notice.
The short version: Nemo keeps what you tell it and what you connect so that it can do the work you ask for. It does not sell your information, it does not show you advertising, and it does not use your data to train AI models.
1. What we collect
- Your account. When you sign in with Google we receive your email address, name and profile picture. When you sign in with Telegram we receive your Telegram id, name, username and a link to your profile photo; we keep the id, name and username, and the language your Telegram client reports. When you use Nemo on LINE, in the chat or in Nemo's pages inside LINE, we receive your LINE user id, display name and the language your LINE app is set to; we keep the id and the display name (an account made from LINE takes that name), and we use the language only to word our greeting and buttons. When you use Nemo on the web, we keep the main language your browser reports (one code, such as en-US), not its full language list. If you register a passkey we store its public key (the private key never leaves your device); if you turn on notifications or link the desktop app, we store the subscription or device record that makes them work. We record the country your sign-up came from, resolved by our web server from your IP address with a local database; the address itself is not stored with your account. If you joined through an invite link, we record who invited you. If you started Nemo from one of our ad links, or opened one in LINE, we record which link.
- What you give Nemo. Your messages and chats on every surface, the files and images you upload, your notes and Wiki sources, your preferences, and the routines and alerts you set up. The facts Nemo remembers about you, because you told it to or because they came up in conversation, each kept with where it was learned and a short quote of the message that taught it. The people you correspond with, learned from the mail and messages Nemo handles for you: name, address, how often you are in touch, and a short inferred description. When Gmail is connected, a profile of your writing style, learned from mail you sent, so drafts sound like you. From your own messages to Nemo, a short note of how you write (for example formality, length and the languages you mix), so replies match you. You can see, edit, reset or turn it off in About you, and it is deleted with your account.
- Connected apps. When you connect an app (Gmail, Google Calendar, Sheets, Docs, Drive, LinkedIn, X, Slack, Notion, an MCP server of your own and others), Nemo reads and writes that app's data only to do what you ask or what you turned on. Google services are covered in detail in section 3.
- Usage. Which features you use, when, from which surface and for how long; the AI usage and cost of your requests; and, on the public pages, the page, the referring site, the country and a visitor id computed from your IP address and browser with a secret that changes every day, so that it cannot be traced back to you. The web server in front of Nemo logs the IP address, time and path of each request, as web servers do. We use no third-party analytics or advertising trackers.
- Payments, when offered. Paid plans and credit purchases are not currently available. If we offer them, payments are handled by Stripe. We keep Stripe's identifiers for your customer record and subscription and the amounts paid. We never see or store your card number.
2. How we use it
- To run Nemo: answer you, carry out the actions you ask for, remember what you told it, and run the routines and alerts you set.
- To offer proactive help you turned on, such as the inbox digest, reply drafts and follow-up reminders.
- To understand interests, on Telegram. From the private messages you send Nemo on Telegram, an AI model notes durable, non-sensitive leisure interests (a team you support, a hobby, the kind of music you mention), each with a short quote from your message as its evidence. The people at ASI Labs who run Nemo (section 8) use these signals to understand who uses Nemo and to shape it, and to choose which of our Telegram updates you receive, for example news about a team you support. Every such update has a "Not interested" button, which stops us using the interests it was sent for and deletes the quotes behind them, and /stop ends all of our messages. The signals are never sold or shared with advertisers, and the words you wrote are never repeated back to you.
- To estimate which country you are in, for the people at ASI Labs who run Nemo (section 8): from a country an administrator set, the country your sign-up came from, your phone number's country code if you use Nemo on WhatsApp, the country our ad was aimed at, your time zone, or the language your Telegram app reports. They see it as an estimate, with where it came from, to understand who uses Nemo and how our ads perform. It never changes what Nemo does for you.
- To keep the service working and safe: measuring reliability, finding and fixing errors, preventing abuse, and enforcing our terms.
- To process and account for paid plans or credits when those purchases are available.
- To tell you about Nemo. Product emails and Telegram messages from us can be stopped at any time (the unsubscribe link, or /stop on Telegram). A message from us may mention something you made in Nemo, such as a recent picture, video or post draft. If you stop using Nemo, we may send up to three messages over the following weeks to ask whether something went wrong; they stop as soon as you reply or ask us to.
- To meet legal obligations.
We do not sell personal information, and we do not use it for third-party advertising.
3. Google user data
If you connect a Google service through Nemo's own Google client, Nemo asks Google for the narrowest permission that covers the feature:
| Service | Permission | What Nemo does with it |
|---|---|---|
| Gmail | Read, write and send email (gmail.modify) | Read and search your mail when you ask, summarise it, draft replies that you review and send from the composer, save the replies it pre-drafts for you to your Drafts folder, mark a conversation read when you open or reply to it, and move a message to Trash when you press Delete on it in Nemo's Email view. In chat, Nemo refuses to delete or trash mail, even when asked. Nemo never permanently deletes mail; the only thing it removes is the draft copy of a message you have just sent from Nemo. |
| Google Calendar | Events (calendar.events) | Tell you what is on your schedule; create, move and cancel events, and add Meet links, when you ask. |
| Google Sheets | Spreadsheets | Read and update a sheet you name or link. |
| Google Docs | Documents, and the files Nemo creates | Create and edit documents, and share or delete the ones Nemo made. |
| Google Drive | Files Nemo created, and file names | Find a sheet or document by its name when you refer to it. |
A Google service connected through Pipedream (section 5) uses the permissions Pipedream's own Google app asks for, which can be broader; Nemo uses them for the same features and nothing else.
We use this data only to provide the features you asked for, or that a feature you turned on prepares for you, in a way that is visible to you in the app. We do not use it for advertising, we do not sell it, and people at ASI Labs do not open it except with your permission to resolve a support problem, to investigate abuse or a security incident, or when the law requires it (section 8).
What Nemo keeps from Gmail: for messages it has seen in the last 31 days, the sender, recipients, subject, date, a short snippet, the labels, Gmail's message id and the unsubscribe link from the headers; whether the last message in a thread was yours; your Gmail address; the contacts and the writing-style profile learned from your mail (section 1); and the cards, digests and draft replies it prepared for you. Message bodies are fetched when a feature needs them and are not stored as such; the reply Nemo writes for you, and any summary it gives you, stay in your chat history like anything else Nemo says. Nemo keeps no copy of your calendar, spreadsheets or documents beyond the links it made and the answers it gave you.
Nemo's use and transfer to any other app of information received from Google APIs will adhere to the Google API Services User Data Policy, including the Limited Use requirements.
Google user data is sent to our AI providers (section 4) only to produce the reply, summary, draft, picture or video you asked for. We do not use Google user data, or any of your other data, to develop, improve or train generalized artificial-intelligence or machine-learning models. Google user data reaches an AI provider inside chat requests, which we route exclusively to zero-data-retention endpoints: the provider does not train on what it receives or keep it once it has answered, within the two limits section 4 describes. The one exception is a picture or video you ask Nemo to make from it. A picture's description goes to Google's picture model, on the same terms. Nemo won't make a video for you from a conversation while something it read there from your Google account is still in what it can see, and a video made only from words or pictures you give for it carries nothing Nemo read; videos go to the video model under its provider's own terms (section 4).
Deleting it. Disconnect the service in Wiki → Apps. Nemo revokes its access and deletes everything it kept from your mailbox: the headers and snippets, the thread records, the scan state, the cards and drafts built from them, and the writing-style profile. The filing rules you created yourself, and the contacts learned from your mail, are kept until you delete them or your account. You can also revoke Nemo's access at any time from your Google Account (myaccount.google.com → Linked apps; older layouts call it Security → Your connections to third-party apps).
How the connection works. Google connections are made either through Pipedream, a connection service that holds the access token and forwards Nemo's requests to Google, or directly through Nemo's own Google client, in which case Nemo holds the token itself, encrypted (section 7). Which one applies is shown when you connect; we are moving every Google connection to Nemo's own client.
4. AI providers
Nemo's replies and drafts, and its reading of the images, PDFs and videos you give it, are produced by AI models that we access through OpenRouter and the model providers behind it (at the time of writing DeepSeek, Google, OpenAI and ByteDance, among others). To produce a reply, Nemo sends the relevant parts of your conversation, your memory, the data from the apps you connected and the files you uploaded. For these chat requests we instruct OpenRouter, on every request, to use only endpoints it lists as zero data retention: the provider does not train on what it receives or keep it once it has answered. Two limits apply. Google, which runs the Gemini models, may keep a prompt its safety systems flag, for up to 90 days, to check for abuse, and does not train on it. OpenRouter screens every request for child sexual abuse material and may keep what it flags for review, reporting and legal preservation.
The models that process Google user data (section 3), who runs them, and what each does:
| Model | Run by | What it does |
|---|---|---|
| DeepSeek V4.1 Flash (deepseek/deepseek-v4.1-flash) | DeepInfra; if DeepInfra fails, another zero-data-retention provider that OpenRouter picks | Writes replies, summaries and drafts |
| Gemini 3.5 Flash Lite (google/gemini-3.5-flash-lite) | Takes over when the first model fails, and reads PDFs | |
| Gemini 3.1 Flash Image (google/gemini-3.1-flash-image) | Reads images |
We do not host these models ourselves. Nemo reaches them through OpenRouter's API, and every such request tells OpenRouter to use only zero-data-retention endpoints. We share your Google user data with them only to produce what you asked for, never for training. When we change these models, we update this list.
Images and videos Nemo generates for you, and voice notes it transcribes, also go through OpenRouter to the model you choose, under that model provider's own data policy as OpenRouter publishes it; photo background removal uses our own service or, as a fallback, Replicate. Pictures are made by Gemini 3.1 Flash Image or Gemini 3.1 Flash Lite Image (google/gemini-3.1-flash-image, google/gemini-3.1-flash-lite-image), run by Google, which we route to Google's Vertex service on the same terms as the Gemini models above. Videos are made by Seedance 2.0 Fast or Seedance 2.0 Mini (bytedance/seedance-2.0-fast, bytedance/seedance-2.0-mini), run by ByteDance; no video model offers zero data retention. If you ask Nemo to make a picture from something in your Google account, the description Nemo writes for it goes to Google's picture model. Nemo won't make a video for you from a conversation while something it read there from your email, calendar, contacts or other connected apps is still in what it can see: on the web, the conversation's last 16 messages; on Telegram, LINE, Slack and email, the last 20 messages of the last 48 hours. If it can't check, it doesn't make one. A video made only from words or pictures you give for it, such as with /video, Animate, a video idea sent on Telegram or Video Studio, isn't held, because nothing Nemo read goes into it. Videos go to ByteDance under its own terms.
We do not use Google user data, or any of your other data, to develop, improve or train generalized artificial-intelligence or machine-learning models. Google user data reaches an AI provider inside chat requests, which we route exclusively to zero-data-retention endpoints: the provider does not train on what it receives or keep it once it has answered, within the two limits section 4 describes. The one exception is a picture or video you ask Nemo to make from it. A picture's description goes to Google's picture model, on the same terms. Nemo won't make a video for you from a conversation while something it read there from your Google account is still in what it can see, and a video made only from words or pictures you give for it carries nothing Nemo read; videos go to the video model under its provider's own terms (section 4).
AI output can be wrong. Nemo shows its sources where it has them, and anything it sends on your behalf (email, posts, messages to others) is drafted for you to approve first, unless you set up a routine that sends on a schedule.
6. Where your information is
ASI Labs is a Singapore company. Nemo's servers and database are in the United States (US East), and the providers in section 5 process data in the United States and in the regions they operate in. By using Nemo you agree to your information being processed there.
7. Security
- Connections to Nemo, and Nemo's connections to the services in section 5, use TLS.
- Credentials and tokens for your connected accounts are encrypted at rest with AES-256-GCM and decrypted only on our servers; they are never included in what is sent to an AI model.
- Requests through Nemo's web app and APIs are checked against your signed-in session; we are completing a server-side secret on every database function.
- Code that Nemo runs for you executes in an isolated, single-use sandbox.
- Nemo never sends anything on your behalf without your approval, unless you created a routine that does so on a schedule.
No system is perfectly secure. If you find a security problem, write to support@heynemo.ai.
8. Access by people at ASI Labs
A small number of named administrators can open Nemo's administration console. It shows how people use Nemo (which features, how often, what went wrong) and the country estimate described in section 2, without showing what they said, with three exceptions: an administrator can open a person's Telegram conversation with Nemo, or the text of their chats with Nemo on the web, to resolve a support problem or an abuse report, and can send a message into the Telegram conversation on the team's behalf; an administrator can open a person's Telegram conversation to review what Nemo sends them on the team's behalf, such as a follow-up or the welcome message, or a reply Nemo proposes to send them, shown with the Telegram messages of theirs it was based on: their three latest and up to nine older ones about a topic; and administrators see the interest signals described in section 2, each with the short quote from your Telegram message that supports it. In a Telegram conversation, administrators also see previews of the pictures the person made with Nemo and the first frame of each of their videos, marked as private. Apart from those previews, we do not open Gmail content, documents or files this way. A message sent this way is recorded with the administrator who sent it, and so is each web chat opened this way.
9. How long we keep it
| What | How long |
|---|---|
| Chats, memory, notes, Wiki sources, uploads, routines, contacts | Until you delete them or your account. |
| Group-chat transcripts (Telegram, LINE, Slack groups) | The last 48 hours of each group's conversation; older messages are removed as the group keeps talking. |
| Interest signals from your Telegram messages (section 2) | The message itself is kept only until the next daily scan, at most a day. A signal and its quotes stay while the interest keeps coming up, are archived once it has not come up for 270 days, and are deleted 90 days after that. A signal we dismissed is deleted after 90 days, unless we were asked never to note that topic again, in which case the label alone is kept, without quotes, to honour that. |
| What Nemo kept from Gmail (section 3) | Messages seen in the last 31 days; older records are removed as Nemo keeps scanning, and everything is removed when you disconnect Gmail. |
| A memory you ask Nemo to forget | Hidden at once; the record is kept for 7 days so it is not re-learned, then deleted. |
| Results of routines | The output for 30 days, the run record for 180 days. |
| AI usage and cost records | Until you delete your account; daily totals per account are kept for accounting. |
| Visits to the public pages | 90 days, under the daily-rotating visitor id. |
| A LINE ad link you opened before joining Nemo (section 1) | 30 days, filed under a code made from your LINE user id, never the id itself. If you join in that time, the link stays with your account as where you came from. |
| Web-server logs (IP address, time, path) | About two weeks. |
| Application logs | Until the next deployment, typically a few days. |
| Your email address, after you unsubscribe | Kept on a do-not-send list so we do not write again. |
When you delete your account, your data is removed from the live service within minutes, in batches, and the access Nemo had to your connected accounts is revoked; a short-lived record of the deletion is kept until the purge has finished and is then removed. We do not run scheduled backups; a copy taken by hand for maintenance is deleted within 30 days. Any payment records stay with Stripe under its own policy. What an AI provider has already processed cannot be recalled from it.
10. Your choices and rights
- See and edit what Nemo remembers about you in Wiki → About you, or tell Nemo to forget something.
- Delete chats, notes and sources in the app.
- Disconnect any app in Wiki → Apps.
- Delete your account in Settings → Account. This is permanent.
- Stop product emails with the unsubscribe link, and Telegram messages from us with /stop.
- Ask us to remove an interest we noted from your Telegram messages, or never to note a particular one again, by writing to support@heynemo.ai.
- Ask us for a copy of your personal information, to correct it, or to delete it, by writing to support@heynemo.ai. We answer within 30 days.
Under Singapore's Personal Data Protection Act you may also complain to the Personal Data Protection Commission. Our data protection officer is Keith, who can be reached at support@heynemo.ai.
11. Children
Nemo is not directed to children under 13, and we do not knowingly collect information from them. If you believe a child has given us personal information, write to us and we will delete it.
12. Changes to this policy
When we change this policy we post the new version here with a new effective date. If a change is material, we tell you in the app or by email before it takes effect.
13. Contact
ASI Labs Pte. Ltd. — 298 Tiong Bahru Road, #05-01, Central Plaza, Singapore 168730 — support@heynemo.ai