AI Transparency Notice

Effective date: 20 October 2026 Last updated: 20 October 2026

This notice explains how Taskwich, operated by Vyxos s.r.o., uses artificial-intelligence (AI) features and how we label them, so you always know when you are interacting with AI. It implements our transparency obligations under the EU AI Act (Regulation (EU) 2024/1689, Article 50), which apply from 2 August 2026.

It complements — and does not replace — our Privacy Policy and Terms of Service.

The only AI feature in Taskwich is Mind Dump. This notice describes that feature, including its voice input (Talk), and nothing else. If we add another AI feature, we will update this notice before it goes live.

Mind Dump

You type or dictate a free-text note; Mind Dump sends that text — together with the name and the start of the description (up to 500 characters) of each workspace project you can see — to a third-party AI model, which suggests structured tasks you can review, edit, or discard. No task is created until you confirm it. When you use Mind Dump you are interacting with an AI system, not a human (AI Act Art. 50(1)).

Mind Dump only suggests. Because you review and author every task, no legal or similarly significant decision is made about you automatically (GDPR Article 22 does not apply). Suggestions may be inaccurate or incomplete — review them before you rely on them. If suggestions look wrong or inappropriate, you can flag them with Report, which sends our support team your reason, any note you add and a copy of the suggested titles.

Voice input (Talk). On the web and in the Android app you can dictate your note instead of typing it. When you choose Talk — or press and hold ＋ in the Android app — your microphone audio is streamed over an encrypted connection directly from your browser or phone to an OpenAI speech-recognition model, which transcribes it live into your note. Talk only turns speech into text; it does not identify who is speaking. The transcript is ordinary text in your note, which you can edit before you ask for suggestions. When you dictate by holding ＋, it is sent for suggestions as soon as you finish; you still review every suggestion. We store neither the recording nor the transcript; we keep a record of each session (start, end and length) to count your plan's voice minutes.

How it is labelled. The Mind Dump panel carries a clear "AI" badge at the point of use, and suggested tasks are shown as AI suggestions for you to accept, edit, or discard.

Who provides the AI, and what data is sent

The models are provided by OpenAI, L.L.C. (United States) — task suggestions use the gpt-5.4-nano model via the OpenAI API, and Talk uses OpenAI's Realtime API for transcription — under a data-processing agreement that incorporates the EU Standard Contractual Clauses and OpenAI's EU–US Data Privacy Framework certification.

  • Data sent for suggestions: the free-text note, plus the name and the start of the description (up to 500 characters) of each workspace project you can see (so suggestions can be filed to the right project). No account email or other identifiers are sent.
  • Data sent when you use Talk: your microphone audio for the length of the session (at most 5 minutes), your device's IP address, up to three language hints and a one-way hash of your account ID, which OpenAI uses to detect abuse without learning who you are.
  • No training: the personal data you enter or dictate is not used to train third-party AI models (OpenAI does not train on API data by default).
  • Retention: your input, including Talk audio, is retained by OpenAI for at most 30 days, then deleted.
  • International transfer: this involves a transfer of data to the United States, described in our Privacy Policy.

Please do not type or say sensitive personal data (GDPR Article 9 special categories) in Mind Dump — it is not needed to use the feature.

What Mind Dump is not

Mind Dump provides productivity assistance only. It is not used for any purpose listed as high-risk in Annex III of the AI Act (such as decisions about employment, creditworthiness, eligibility for services, or law enforcement), and it makes no legal or similarly significant decision about you. Because it assists you rather than autonomously generating published content, the AI Act's content-marking obligation (Article 50(2)) does not apply to it.

Contact

Questions about our AI features or this notice: [email protected] or [email protected].