Privacy Policy
Last updated: 30 September 2026
This page explains what we process when you use Still Needs a Human, why we process it and who else sees it. It is written to be read rather than to be technically unassailable. If something here is unclear, ask us and we will answer plainly.
Who we are
Still Needs a Human Ltd is a UK company based at Building 3, Chiswick Park, 566 Chiswick High Road, London W4 5YA. This policy covers the Still Needs a Human platform (the What's Wrong QA checks, LQA scorecards and Eval engine bake-off), the in-editor browser extension and the TMS and editor integrations. We act as the controller for your account data, and as a processor for the translation content you send us to check.
What we process
- Account data: your name and email used to sign in and run your workspace. If you choose to sign in with Google, Google shares your account email and basic profile with us for authentication only.
- Translation content: the source and target segments, files and job metadata you upload, connect through a TMS or editor integration or send from the browser extension. We process it to produce your QA reports, LQA scorecards and MT evaluations.
- Requests from the site: the email, company and message you send through the contact or request forms, kept so a person can reply.
- Connection token and settings: for the browser extension, a connection token and your settings are stored locally in your browser (
chrome.storage.local) so it can authenticate you. The token is revocable from the panel. - Operational logs: minimal technical logs needed to run and secure the service.
How your content is used and kept separate
Your content is processed only to produce your own quality results. Each workspace is isolated: content, configurations and reports are never shared between workspaces. We do not use your content to train our models, we do not sell or share your data and we run no advertising or third-party tracking.
AI-assisted review (opt-in)
The AI review layer is a profile setting: the profiles we ship have it on, and a profile can switch it off. When it is on, the relevant segment text is sent to an AI model provider to help grade quality. AI calls run on our own accounts with each provider (platform keys), not on keys you supply.
Your workspace chooses which providers may receive its text. A workspace admin ticks them in Settings under AI providers, and the choice also holds for fallbacks and second opinions. New workspaces allow Google Gemini, OpenAI and DeepInfra; DeepSeek and the machine translation engines stay off until an admin allows them.
Text sent for review is fenced as untrusted content and used only to return findings for that run. We do not use it to train our models.
The browser extension
When you click Run What's Wrong inside a supported CAT editor, the extension reads the job you currently have open (source and target text plus match and lock metadata) from your own logged-in editor session and sends it to your workspace server to produce a report. It never modifies the job, and it activates only on the supported editor sites, never on any other website.
Service providers
We rely on a small set of subprocessors. Each one handles data only to run the service.
- Railway: hosting for the application and its data.
- Resend: transactional email, such as invitations, password resets and contact messages.
- Google Gemini: AI review and second opinions, when your workspace allows it.
- OpenAI: AI review and second opinions, when your workspace allows it.
- DeepInfra: AI review and second opinions on open models, when your workspace allows it.
- DeepSeek: AI review and second opinions, only if your workspace allows it.
- DeepL: machine translation in Eval, only if your workspace enables it.
- Google Translate: machine translation in Eval, only if your workspace enables it.
- Microsoft Translator: machine translation in Eval, only if your workspace enables it.
- Google: sign-in, only if you choose to sign in with Google.
- Crowdin: when you connect it, we read the strings you check and post findings back as comments.
- Smartling: when you connect it, we read the jobs you check and post findings back as issues.
- Phrase: when you use the browser extension in Phrase, it reads the job you have open in your own session.
- Dropbox: when you connect a folder, we read the files you drop there and write the reports back.
The fonts on this site are served from our own servers, so reading these pages sends no request to a font provider.
Storage and retention
Reports and scorecards are stored under your workspace account. Uploaded job files used to produce a report are removed automatically (within 24 hours). The locally stored extension token is removed when you disconnect. You can ask us to delete your workspace data.
Your rights
Under UK GDPR you can ask to access, correct or delete your personal data, and object to certain processing. To make a request, reach us through the contact form.
Contact
Still Needs a Human Ltd, Building 3, Chiswick Park, 566 Chiswick High Road, London W4 5YA. Questions or requests? Reach us through the contact form.