--- title: "AI and Virtual Employees Terms — Tidy" description: "How Tidy's AI works, where it runs, what credits cover, what virtual employees may do for you, and the rules for both." url: https://www.tidy.engineering/legal/ai-terms last_updated: 2026-10-13T10:33:58.000Z --- [Terms & Privacy](https://www.tidy.engineering/legal) Tidy Engineering / Legal # AI and Virtual Employees Terms Effective October 13, 2026 ## What this covers These terms cover every AI feature in Tidy: assistant chat and drafting, extraction and triage, search, briefings, transcription, image generation, automations, and virtual employees. They are part of the [Terms of Service](https://www.tidy.engineering/legal/terms). The AI rules in the [Acceptable Use and Content Policy](https://www.tidy.engineering/legal/content-and-use) apply too. ## Where AI runs **On your Android phone.** Offline search uses a local index, not an AI model. On-device briefings use Android’s AICore, and capture processing can use the downloaded Gemma 4 E4B model. What runs on the phone stays on the phone. Some requests run on Tidy’s servers instead, depending on your device, settings, and connection. **On Tidy’s servers and AI providers.** Server AI features send what a request needs to the provider handling it, which can include a conversation and the tool results it uses. Today those providers are Google (Gemini), Groq, and Fireworks AI. Dictation is transcribed on your phone; meetings, recordings, and podcasts on Tidy’s own servers. Live voice calls are transcribed by Groq, and Groq takes over meeting and recording transcription when our servers are unavailable. Simple images are made on Tidy-managed hardware; rich images use Google Gemini. The [Subprocessors page](https://www.tidy.engineering/legal/subprocessors) keeps the current list. **At an endpoint you choose.** Supported features can use your own model endpoint, on your hardware or with a provider you pick. A virtual employee you create uses the model, endpoint, and credential you set for it, and its prompts and tool results go to that provider. Your agreement with that provider governs what it does; check its privacy and training terms first. Which model serves a feature can change as models improve. ## Your content and AI output You own your content, and you own the output Tidy generates for you: summaries, drafts, briefings, and the rest. You give Tidy only the permission it needs to process your content and run the features you use. Output may not be unique; two members can get similar answers to the same question. Don’t present AI output as human work where that would deceive someone. ## No training on your content Tidy does not train AI models on your content, and our AI providers are engaged on terms that use content only to answer the request, not for training. A provider you choose yourself is governed by your own agreement with it. Tidy’s use of information received from Google APIs adheres to the Google API Services User Data Policy, including the Limited Use requirements. Google data is never used to train general models, ours or anyone else’s. ## Feedback on AI output When you rate or correct something the AI made (a summary, a suggested task, contact, or event), we store your rating or correction, any note you add, and which item it was about, inside your workspace. We use it to improve Tidy’s results for you. It is deleted with your account, it is never sold, and it is not used to train any third-party model. ## Credits and fair use Assistant credits cover the heavier AI work. Each plan includes credits per seat every month: Core 500, Plus 1,500, Max 4,000. Credits reset each month, and unused credits don’t roll over. If you run out, AI features pause or scale back until the reset, or you can move up a plan or buy a one-time top-up in Settings. Top-up credits also expire at the next reset. Two other limits keep the service fair: email processing per inbox per day (150 on Core, 300 on Plus, 1,000 on Max) and storage per seat (10, 100, or 250 GB). The [pricing page](https://www.tidy.engineering/pricing) lists what each plan includes. The credit cost of an action can change as models and their costs change. The [pricing page](https://www.tidy.engineering/pricing) is the current statement of what each plan includes. ## Check the output AI output can be wrong, incomplete, or out of date. Review anything that matters before you act on it or send it. For legal, medical, or financial work, have a qualified person check it first. ## Virtual employees A virtual employee is an agent you set up that acts inside the access you give it. It can use tools: send email and texts, create and update calendar events, browse the web through a controlled browser, call other services, and run the automations you define. Tidy does not give it your payment details, but it can still take actions with real consequences through connected services or a browser, so grant access with care. You decide how much it does on its own. Some actions ask you first; others run within the approval setting you chose. Approvals, budgets, and a check on each tool call are the guardrails Tidy runs, and we can suspend an employee that misbehaves. You are responsible for the access you grant and the actions you approve. Virtual employees are on Plus and above. ## Changes and availability AI features change quickly. The model behind a feature can change, and features can improve, move, or be withdrawn, especially during the private beta. On-device features depend on what your phone supports. ## When something goes wrong If an AI feature does the wrong thing, such as a message sent wrong or an automation that misfired, email [support@tidy.engineering](mailto:support@tidy.engineering). We’ll help fix it, and undo it where that’s possible.