By default 20xwork does not use customer data to train, fine-tune or evaluate any model, on any plan. There is no setting a user or admin can flip that changes this; it changes only by written agreement between 20xwork and your organisation.
The model providers we use. The language models behind the product run on Microsoft Azure (Azure OpenAI and Azure AI Foundry) under Microsoft's enterprise terms, which prohibit the use of customer prompts and outputs for training. The vendors behind those models do not receive your organisation's identity and have no training rights over your data.
Your own assistant is governed by your plan. When a 20xwork card appears inside an employee's Claude, Claude Code, Copilot or other assistant, that session is governed by your organisation's agreement with that vendor, not by 20xwork. Business plans from these vendors do not train on your data by default; personal consumer plans may. We recommend that seats using 20xwork run on a business plan.
The one door in: the Development Partner Programme. Organisations may choose to help improve 20xwork's own models by joining the programme. It works like this:
- Opt-in only, by written agreement signed by your organisation, never by an individual user.
- What may be used: de-identified work signals, recommendations and outcomes, for example "the next step named an owner and a date" and whether the card was kept, adapted or dismissed.
- What is never used: record contents, email or message text, documents, names, email addresses or any identifier of a person, account or record.
- De-identified data is retained for the programme for up to 5 years and is deleted at your request on leaving.
- You can leave at any time by email to adharsh@20xwork.ai. Data already used to train a released model cannot be extracted from it, which is why the programme uses only de-identified signals.
- Members get a say in the roadmap and early access to what the programme produces.
Whether your organisation is in the programme is shown to admins in Organisation settings under Data and privacy.