Privacy Policy

1. Who we are

Service: Random Assistant (random-assistant.com).

Legal name & registered address: Random Solutions s.r.o., IČO 29929962, Nové sady 988/2, 602 00 Brno, Czech Republic.

Data protection contact: privacy@random-assistant.com. We have not appointed a data protection officer.

Competent supervisory authority: Úřad pro ochranu osobních údajů (ÚOOÚ).

Part A. Where we are the controller

2. What we process, why, and on what basis

Recipients for Part A: our hosting and delivery providers (Hetzner, Cloudflare), Google Workspace (our company email), Stripe (payments and partner payouts), PostHog and Sentry (analytics and error reports; see Section 10 for their regions), AhaSend (transactional email, Netherlands), and the accounting and tax advisers we engage. We send marketing email only with your separate consent. You may object at any time to processing based on legitimate interest (Art 21); see Section 8. If you signed up through a partner's referral link, that partner receives a pseudonymous customer number (not your name), the date you signed up, whether you are on a trial, paying or have cancelled, how many of your invoices were paid, and the commission it earns from them, to check its payout. Partners never receive your name, contact details or content.

Personal use. If you use Random Assistant for your own purposes, we are also the controller of the data in the accounts you connect for those purposes; the legal basis is the contract (Art 6(1)(b)) for your own data and our legitimate interest in providing the Service you asked for (Art 6(1)(f)) for other people in your mailbox and calendar.

Employer accounts. If an account you connect belongs to your employer, your employer remains the controller of that account and decides whether you may connect it. On a personal subscription we process the data from it as controller, as described above; where your employer uses Random Assistant under our Data Processing Agreement, we are its processor. Its administrator can ask us to disconnect the account and delete the data we hold from it.

For personal use, everything in Part B that says ‘your company’ means us: what Random learns is done for you under the Terms, not under a Data Processing Agreement; you and the other people in your mailbox and calendar exercise your rights with us at privacy@random-assistant.com, and they may object to our processing of their data at any time.

Part B. Where your company is the controller

3. Workspace data we process for your company

We process workspace data only on your company's instructions, as set out in the Data Processing Agreement. Your company's own privacy notice governs this data and tells its staff and contacts about it. If you contact us about workspace data, we forward your request to your company.

We classify data by sensitivity. Class A is encrypted at rest with AES-256-GCM, using per-tenant keys derived via HKDF-SHA256 from a deployment master key held outside the database; Class B is stored structured with row-level tenant isolation; Class C is operational metadata.

We do not use workspace data to train AI models, and the AI model providers we engage do not either.

What Random learns

To work better for you, Random learns from the sources you connect. It keeps what it learns in your workspace and uses it for you: in briefs, meeting preparation, drafts and chat, without being asked. It learns only these fact types:

• Facts about contacts and their companies
• Facts about deals and projects
• Commitments and agreed next steps
• Preferences and plans, including travel and how answers should look
• Profile: role, company, goals, key contacts, languages, tools
• How people address you: the names others use for you in emails, and the name you sign with (only where switched on)
• Writing style: tone and signature
• Working patterns: working hours

Not every source feeds every type. Per source:

Learning is on by default for each source you connect, and you are told so when you connect it. You can switch learning off for any source, at any time, under Settings > Connectors. The source then keeps working for what you connected it for, and nothing new is learned from it; what was already learned stays until you forget it. You can review your learned profile and confirm or forget each fact, and you can forget single memories. A forgotten memory is hidden at once and deleted after 7 days, the window in which you can undo it. For a request about one person, your company can forget the memories and facts about them, and we help on request. Random refuses to store special-category data, such as health, religion, political opinions, ethnic origin, trade-union membership, sex life or sexual orientation, biometric data, or criminal records, as a memory or learned fact, about you or anyone else, including where it would only be inferred. The check is automatic and can miss an unusual wording, so you can always delete a memory yourself. Nobody else in your company, including its owners and administrators, can see what Random learned about you.

Learning is done for your company under the Data Processing Agreement (Section 2 and Annex I). Your company is the controller for it and informs its staff and contacts; we publish a template notice it can use.

4. AI processing

To produce a draft, brief or summary, or to learn from a source, we send only the content that task needs (not your whole mailbox) to an AI model provider. Every account is served in the EU: requests go through OpenRouter's EU endpoint (eu.openrouter.ai), limited to EU data regions and to providers that do not collect your data or train on it. If that route is unavailable, the request fails; it never falls back to the US. For dictation, a Mistral AI speech model turns your voice into text through the same endpoint. Where live dictation is switched on, the text you see while you speak comes from Mistral AI directly, on its EU endpoint (api.eu.mistral.ai); Mistral AI keeps that audio and text for up to 30 days for abuse monitoring and never uses them for training. What Random learns is stored in your workspace, never in a model, and never used for another customer. The providers are named in the sub-processor list and in DPA Annex III.

We do not join or record your meetings. We work from the transcript or summary your own meeting tool sends us; that tool processes your meetings under your own contract with it.

5. Google, Microsoft and Slack data

Google user data. Random Assistant'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. For your mail and calendar, we request only the access needed to read them and to create drafts and tasks you approve. If your workspace connects Slack, a follow-up email draft built from your mail, with its subject and its recipient, is also shown on an approval card in Slack, which stores it in the US unless your contract with Slack provides EU data residency (see Slack data below). Gmail and Calendar access is brokered by Nylas, our sub-processor, which processes it in its EU region (Ireland). No one at Random Solutions s.r.o. reads your mailbox content except with your consent for a specific support case, to investigate security incidents, or where the law requires it. Google user data is never used to develop, improve or train generalised AI or machine-learning models.

Google Drive. If you connect Google Drive, Random uses only the drive.file permission. It can open only the files you pick in Google's file picker and the files it saved to your Drive at your request; it cannot see the rest of your Drive. It copies the files you pick into your Data Room. When you ask Random to save something to Google Drive, it creates a new file in the Drive your workspace connected; it never changes, moves or deletes files there. Google stores that file under your contract with Google, and deleting data in Random, or asking us to erase it, does not remove it from Drive. Drive does not go through Nylas: it uses our own Google client, connected through Nango, which we host on our EU infrastructure.

Microsoft 365 data. Outlook mail and Microsoft calendar access is brokered by Nylas under delegated permissions you or your tenant administrator grant; your administrator may need to approve the app. When you disconnect the account or close your workspace, we delete the data obtained through it as described in the Data Processing Agreement.

Slack data. When your workspace connects Slack, Random uses a bot token, never a user token. It reads the direct messages you send to the Random bot, the messages in which you @mention it, and the messages you choose with Track this or by reacting to them with the pushpin emoji (📌). Slack also tells Random about the reactions added in the conversations the bot is in; Random acts only on a pushpin from the person who connected Slack and discards every other reaction. In a channel, Random reads messages only if the Random bot was invited to it and the person who connected Slack to Random switched that channel on (off by default). When a channel is switched on, Random posts a notice there so its members know. In a switched-on channel, Random checks only the messages that person writes or that mention them, and keeps only the commitments it finds, each with the sentence it came from, as follow-up proposals for that person to approve; it does not keep the rest of the messages. Slack sends Random every message in the channels the bot is in; Random discards all others without reading or keeping them. It never reads direct messages between people. If anyone else writes to the bot, Random only tells them who it works for and keeps nothing of the message. It answers an @mention in your direct messages; in the channel it only posts a short note that it answered you directly. Random also writes to Slack so its work can be approved there. It posts approval cards only into the direct messages between the person who connected Slack and the Random bot; the card for a follow-up email draft, built from any mailbox your workspace connected, shows the draft's subject and text and its recipient's name and email address. It also adds a status reaction (for example "seen", "done" or "waiting for your approval") to the messages it works on: the ones you send it or @mention it in, a message you pick with Track this or the pushpin (which can be someone else's), and, in a switched-on channel, a message written by or mentioning that person from which it files a commitment. The reaction appears on the message where it was posted, so other members of that conversation can see it. This is part of the service you instruct us to provide under the Data Processing Agreement, and Slack stores these cards and reactions under your contract with Slack. Disconnecting Slack, or deleting data in Random, does not remove cards and reactions already in Slack; Slack keeps them under your contract with Slack. Slack text is handled like a chat in the app: it is kept as your chat with Random and in what that chat creates, such as follow-ups, tasks and memories (Section 3), all encrypted, and it is never used to train AI models. Slack keeps its own copy under your contract with Slack. When you disconnect Slack in Random, or Random is removed from your Slack workspace or its access is revoked, Random deletes the Slack token, the Slack IDs and names stored for the connection, the list of switched-on channels and every Slack follow-up proposal you have not approved. Your follow-ups and chats stay until you delete them or their retention period (Section 6) ends, and follow-ups keep their link to the Slack message.

6. Retention schedule (workspace data)

What the assistant uses as context to do a better job for your workspace is kept while your account is active: meeting transcripts and recaps, email triage results, drafts and your edits, daily briefs and follow-ups. That is how it can later know who needed a reply, what was agreed in a meeting, what you promised whom and how you write. It is deleted when the account ends or when you ask for erasure; for the items listed under Delete sooner below, also sooner if the workspace owner shortens the period. Data the assistant does not reuse keeps a short, fixed period. Memories and the learned profile follow their own rows below: a memory that is not used is archived, and deleted 12 months later unless you pin it (on Enterprise plans unused memories stay active).

Delete sooner. The workspace owner can shorten these periods under Settings > Privacy > Delete sooner, to 30 days, 90 days or 1 year. The period chosen then applies to meeting transcripts (counted from the meeting; the recap is kept), email triage results, drafts and your edits, daily briefs, completed follow-ups (counted from completion), and to meeting prep packets and recruiting reviews where it is shorter than their own period. Open follow-ups are kept.

Other people in your data. Meeting guests and the people who email you appear in transcripts, triage results and drafts. Their data follows the same rule: kept while the account is active, unless the owner shortens it. What Random learned about them follows the memory and learned-profile rows, and your company can forget the memories and facts about one person. Your company is the controller for it, so they exercise their rights through your company, and we help it answer them (Section 7).

Copies held elsewhere. The accounts you connect (mailbox, calendar, meeting tool, CRM) keep their own data under your own contract with them; deleting data here does not delete it there. AI model providers receive only what a task needs, under terms that exclude training; each provider's retention terms, where confirmed, are in DPA Annex III. Our backups hold copies for about 30 days.

Some records outlast an erasure request. Proof of your agreement: we keep your workspace ID, company name, the Terms and DPA versions you accepted and the dates, for 3 years after the contract ends, to prove the contract (GDPR Art. 17(3)(e)). Then it is deleted. Usage records: deleted. Only anonymous monthly totals per AI model remain, merged with other deleted accounts, with no link to you. Billing records: kept as tax law requires. Audit log entries, including the record of the deletion itself: kept until the audit log deletes them after 12 months.

7. Your rights (GDPR Articles 15 to 22)

For workspace data (Part B), please contact your company first; we help it answer you. For Part A data, contact us at privacy@random-assistant.com.

• Access (Art 15): request a copy of your data via the dashboard export or by emailing privacy@random-assistant.com.
• Rectification (Art 16): correct inaccurate data in the dashboard settings or by emailing us.
• Erasure (Art 17): request deletion in Settings > Privacy & data or at /legal/delete. We email a single-use confirmation link. After you confirm, we review the request and, within 3 working days, delete your workspace data, revoke your connected accounts and remove the account record. If you have a subscription (active, in trial or with a payment overdue), cancel it in Plan & billing first; there are no further charges after that. Backups are deleted within about 30 days. What outlasts the request is listed in Section 6.
• Restriction (Art 18): contact us to pause processing pending dispute resolution.
• Portability (Art 20): request a structured, machine-readable export via the dashboard data export or by emailing us.
• Automated decision-making (Art 22): Random Assistant does not make solely automated decisions producing legal or similarly significant effects on you. Drafts, summaries, and prep packets are advisory; you remain in control.
• Complaint: you may complain to the data-protection authority where you live or work, or to our lead authority, Úřad pro ochranu osobních údajů (ÚOOÚ) (Art 77).

Giving us account and billing data is needed to conclude and perform the contract; without it we cannot provide the Service.

8. Your right to object (Art 21)

You may object at any time, on grounds relating to your particular situation, to processing based on our legitimate interest (Part A: website analytics, error reports, security logs, server logs, product usage). Email privacy@random-assistant.com. Your browser's Do Not Track or Global Privacy Control signal also switches off our page-view counting.

9. California

We do not sell personal information and do not share it for cross-context behavioural advertising. Where the CCPA applies to a customer, we act as its service provider under Section 15 of the Data Processing Agreement. Requests: privacy@random-assistant.com.

10. Sub-processors and international transfers

The recipients of personal data are our sub-processors, each named with purpose, location and transfer basis in the sub-processor list; the AI model providers behind OpenRouter are in DPA Annex III. By category: hosting and network (Hetzner, Cloudflare), email and calendar connection (Nylas), AI processing (OpenRouter, Mistral AI), email delivery (AhaSend), payments (Stripe), error monitoring and analytics (Sentry, PostHog) and web search (Staan). We notify customers 30 days before adding or replacing one, as the Data Processing Agreement sets out.

Personal data leaves the EU/EEA where a recipient is a US company or processes some data outside the EU, as the list states for each: Google (our company email, Google Workspace), OpenRouter and the US-owned model host behind it (Google; DPA Annex III), Cloudflare, Stripe, PostHog and Sentry, Mistral AI for its own sub-processors outside the EEA, and Nylas for backups and for access by its support engineers in the United States and Canada (mailbox content itself is processed in Ireland; Nylas is also certified under the EU-US Data Privacy Framework). For these transfers we rely on an adequacy decision (including certification under the EU-US Data Privacy Framework, where the recipient is certified) or on the Standard Contractual Clauses, with encryption at rest and in transit and data minimisation as supplementary measures. For a copy of the safeguards, write to privacy@random-assistant.com. Web search (Staan, France) may fall back to Staan's own search providers in the United States: only where your workspace's Web research setting is on (it is off by default); when Staan's EU index lacks results, Staan may pass the query on, on the basis of your opt-in when you turn Web research on, after the notice shown on the switch (Art. 49(1)(a) GDPR).

11. Security measures

We protect personal data with technical and organisational measures, including encryption at rest with per-workspace keys, encryption in transit, separation of each workspace's data, access control, audit logging and backups. The full list is Annex II of the Data Processing Agreement.

12. Breach notification

For Part A data, if a personal-data breach is likely to result in a risk to your rights, we notify the competent supervisory authority within 72 hours of becoming aware (Art 33), and affected people without undue delay where the risk is high (Art 34). For workspace data (Part B), we notify your company within 48 hours, or 24 hours for confirmed exfiltration, as the Data Processing Agreement sets out. We maintain an internal incident-response procedure that governs how we detect, contain, and report incidents.

13. Cookies and analytics

We set strictly necessary cookies only: the session cookies app_access and app_refresh (httpOnly, Secure) and a CSRF cookie, app_csrf. We use a product-analytics sub-processor, PostHog (EU cloud, Germany), to count page views on this website and measure feature usage in the app. In your browser it runs in cookieless, in-memory mode, so it writes no analytics cookie and keeps no analytics identifier in your browser storage. Error reports from the app go to Sentry (EU region) without request bodies, cookies or user details. We do not use advertising or cross-site tracking cookies, so we show no cookie banner. The full list is at /legal/cookies.

14. Children

Random Assistant is offered to businesses and to adult consumers. It is not directed at children, and we do not knowingly collect personal data from children.

15. Changes to this notice

We post changes on this page and update the effective date above. Customers are notified by email of changes that materially expand the purposes of processing.

16. Contact

Data protection contact: privacy@random-assistant.com. We have not appointed a data protection officer.

Registered address: Nové sady 988/2, 602 00 Brno, Czech Republic.