Security

Last updated: 22 October 2026

This page is for IT and security teams reviewing Arasso for their organisation. It covers where meeting data goes, what is stored and for how long, where it is processed, and what the desktop app needs on a managed computer. Our Privacy Policy is the binding document; this page describes the same practices in technical detail.

At a glance

• No audio is stored. Meeting audio, microphone audio and the Interpreter’s synthesised voice are processed in memory while a session runs and are never written to our servers.
• No AI training on your content. Arasso does not train AI models on customer content, and neither do the AI providers we use.
• Business accounts store no conversation content on Arasso servers.
• No access to Microsoft 365 or Teams. The app requests no permissions on work accounts and never joins a meeting.
• No administrator rights are needed to install or run the app, and every Windows program file is digitally signed.
• Encrypted everywhere. Every connection uses TLS.

How data flows

1. The Arasso desktop app captures audio on the user’s computer and streams it over an encrypted WebSocket (TLS) to the Arasso relay at api.arasso.im.
2. The relay runs in Tokyo, Japan, with a second server in Los Angeles, USA. Each connection is served by one of them.
3. Captions: the relay streams the audio to our speech provider, which transcribes and translates it in the provider’s region nearest the relay: Japan for the Tokyo relay, the USA for the Los Angeles relay. The relay sends the captions back to the app, which shows them on screen.
4. Interpreter: the user’s microphone audio follows the same path to the speech provider for transcription. A translation provider (USA) translates the transcribed text, with a second translation provider (USA) as a fallback. The speech provider then synthesises the translated voice, and the app plays it into the meeting through a virtual microphone on the user’s computer. The other participants’ speech is captioned as in step 3.

Business accounts

Arasso stores no meeting content for Business accounts. Business accounts are for organisations, and the commitments below apply in both Captions and Interpreter mode.

• No meeting content on our servers. No audio, transcripts or translations are written to Arasso servers. Text exists in memory while a session runs, and on the user’s own computer as local history.
• Storage fails closed. At the start of every session the relay reads the account type, and it records caption text only for an account it has confirmed is not a Business account. If that check fails, nothing is stored.
• No AI training. Neither Arasso nor its AI providers train models on meeting content.

What is stored for Business accounts

Providers by mode

Processing region

Each session’s speech is processed in the speech provider’s region nearest the relay that serves it. For a Business account we can fix that region, Japan or the USA, for all of its users, whichever relay they connect to; if the fixed region is unavailable, the session does not start rather than run elsewhere. Translation in Interpreter mode runs in the USA.

Setting up Business accounts

We set up Business accounts for an organisation. Users’ email addresses can be registered before they first sign in, so the commitments above apply from their first session. Ask us at any time to confirm the status of an account.

Access by Arasso staff

There is no meeting content from Business accounts for Arasso staff to access. Usage records and diagnostic events are used only for billing, support and keeping the service working, and access to production systems is restricted.

Further requirements

If your organisation needs single sign-on, central administration or another control, tell us at [email protected]. We build Business features around what customers need.

What is stored

For accounts other than Business accounts, Interpreter mode stores only the transcript of what the user said; the translation of the user’s speech and the synthesised voice are not stored. Business accounts store no caption text at all (see Business accounts).

Local history. The app saves caption text as transcript history on the user’s computer, in the app’s data folder under the user’s profile. Users can delete a single session or all history from inside the app at any time. Deleting the app’s data folder also removes it.

Retention and deletion

• Caption text is deleted 90 days after the session ends.
• Diagnostic events are deleted 90 days after we receive them.
• To delete an account, email [email protected]. We delete or de-identify the account’s records within 30 days, except where we must keep them for legal, tax or accounting reasons. Deleting an account also deletes its usage records, caption text, sign-in records, problem reports and the diagnostic events sent by the app. Server-side diagnostic events that reference the account are removed by the 90-day job.

AI processing and training

Arasso does not train AI models on customer content, and neither do its AI providers.

• Speech provider (speech recognition, caption translation, Interpreter voice): does not use customer content to train or improve its models under its terms of service, and does not keep the audio or transcripts of real-time requests. It holds ISO/IEC 27001:2022 certification and a SOC 2 Type 2 report, and we have a signed data processing agreement with it.
• Translation providers (Interpreter only): do not use data sent through their paid APIs to train their models. They keep the text sent to them and the translation returned for abuse monitoring only, then delete it: up to 30 days for the primary provider, and up to 55 days for the fallback provider when it is used. Captions do not use these providers.

Where data is processed

Arasso is administered from Australia. The full list of subprocessors is available on request.

Encryption

• Every connection is encrypted with TLS: between the app and the relay, between the relay and each provider, and between the relay and the database.
• The database is encrypted at rest by the provider.
• The app keeps its sign-in token in Windows Credential Manager or the macOS Keychain. Arasso stores no passwords: users sign in with a one-time email code or with Google.

Operational security

• Administration. One person, the founder, administers Arasso’s production systems. Every administrative account (code hosting, servers, database, speech and translation providers, payments, domain and website, email) requires multi-factor authentication, and no account is shared.
• Keys. Provider API keys are never in the source code or in the desktop app. They are held as encrypted secrets on the server platform and on the administrator’s computer.
• Administrator’s computer. The computer used to administer Arasso has full-disk encryption.
• Incidents. If a personal data breach affects a Business customer’s data, we notify that customer without undue delay and within 72 hours of becoming aware of it, with what we know at that point, and follow up as we learn more.
• New providers. Customers with a signed data processing agreement are told at least 30 days before a new provider starts processing their personal data, and can object. If a provider has to be replaced at once because of an outage, a security problem or the provider stopping its service, we tell them as soon as possible.
• Limited exposure. Arasso servers store no transcripts, translations or audio from Business accounts, so a breach of those servers could not expose the content of a Business customer’s past meetings, beyond any text a user chose to paste into a problem report.

The desktop app on a managed computer

Account access. Arasso does not connect to Microsoft 365, Teams, Outlook or the Microsoft Graph API, and requests no OAuth permissions on work accounts. It does not join meetings as a participant and installs nothing into meeting apps. Users sign in to their own Arasso account with an email code or Google sign-in, which shares their basic Google profile (email address, name and profile picture).

Audio capture. Captions listen to the computer’s system audio output: everything the computer plays, not only the meeting app. Arasso’s own sound is excluded on macOS and on Windows 10 version 2004 or later. Users can switch the Captions input to the microphone instead. On macOS, capturing system audio requires the user’s approval in the system permission prompt.

Microphone. The microphone is used by Interpreter, and by Captions only when the user selects it as the input. When Interpreter is set up with “Send my own voice to the meeting” turned on, the app passes the microphone to the virtual microphone. By default it opens the microphone only while a meeting app is reading the virtual microphone; a setting lets the user keep it open. That audio stays on the computer and is never sent to Arasso.

Start at login. Starting Arasso at login is optional and per user. It is offered, pre-selected, on the last step of Interpreter setup; the installer never turns it on. It can be turned off in Windows Settings, Apps, Startup, in macOS System Settings, General, Login Items, or by running Interpreter setup again with the option cleared.

Installation and updates. The app installs per user, into the user’s profile, and needs no administrator rights to install or run. The Windows installer and every program file it installs are digitally signed (Authenticode, SSL.com). The Mac app is signed with an Apple Developer ID and notarised by Apple. The app checks for updates and verifies each update’s signature before installing it. If the Microsoft Edge WebView2 Runtime is missing, the Windows installer downloads it from Microsoft.

VB-CABLE (Windows, Interpreter only). For the Interpreter’s voice to reach a meeting app on Windows, the user installs VB-CABLE, a free virtual audio driver published by VB-Audio. Arasso does not bundle it. Installing it requires administrator rights once. Captions do not need it. On Mac, Arasso installs its own Arasso Microphone driver instead, which asks for an administrator password once.

Network access. For proxy or firewall allowlists, the app itself connects to:

Sign-in with Google and checkout open in the user’s web browser, which then contacts Google (accounts.google.com) or Stripe (checkout.stripe.com, billing.stripe.com).

Documentation and contact

We have a data processing agreement ready to sign and can share the full list of subprocessors. If your organisation has its own template, send it to us and we will review it. Security and privacy questions, and requests for documentation: [email protected]