Non-technical FAQ
Motive, bias, and business model — the questions an editor asks before a technical review even starts.
Who decides what counts as a "euphemism" worth flagging?
The phrase library is built around a specific test: does this phrase abstract real, specific, physical harm to people or the environment, and is there a specific legal citation (Rome Statute, Geneva Conventions, UDHR, or the Planetary Boundaries framework for ecological harm) that names what's actually happening? "Collateral effects" flags because it's mapped to specific language in the Geneva Conventions about civilian harm. A phrase doesn't get added because someone finds it distasteful — it needs a citable legal anchor.
Isn't this just left-wing/political framing dressed up as a tool?
The FPÖ "Remigration" example in the press release is drawn from Der Standard's own description of the term, not the project's own political judgment. The same detection logic applies regardless of who's speaking — a corporate press release using "resource optimization" to describe layoffs, or a government statement using "enhanced interrogation," get the same treatment: name the specific legal or factual reality the phrase is standing in for. The project's founder describes his own position as "open-minded, not far-left" — the tool's actual behavior is checkable against that: it flags institutional euphemism wherever it appears, not language from one side of a political spectrum specifically.
Is this censorship, or does it block/change anything?
No. It doesn't remove, hide, block, or edit anything. It underlines a phrase and, on click, shows a translation and citation alongside the original text, which stays exactly as written. Nothing is altered, nothing is blocked from being sent or published.
Who's behind this — a company, a nonprofit, a team?
One person: Andreas Fleschurz, based in Wiener Neustadt, Austria, building and operating it as an individual project, not through any company. No investors, no funding round.
Why does it cost money for some features if the mission is about transparency?
The core detection — the static phrase library, on every supported platform — is free, permanently, no account required. Paid tiers cover the real infrastructure cost of AI-assisted detection (API calls aren't free to run) and features aimed at organizations specifically (shared team seats, a private company phrase repository, MDM deployment). The public dataset itself — the actual transparency artifact — is free to browse for everyone regardless of tier.
Does the free tier actually do anything meaningful, or is it a limited trial?
It's the full static-detection feature set on every supported platform, permanently, not a time-limited trial. What's gated behind paid tiers is specifically the AI-assisted detection (for phrasing the static list misses) and organizational features — not a crippled version of the core idea.
What happens to data collected through the free public dataset — can someone be identified from it?
No. Contributed detection events are anonymized before they're stored — no email, no API key, no account identifier travels with a submitted phrase. The privacy policy states this is a one-way process: once anonymized, an individual submission can't be traced back to the person who triggered it, which also means an individual contribution can't be selectively deleted on request (a user can stop future contributions any time by disabling data sharing in settings).
Is this trying to replace human judgment about language?
No — it's trying to make the human judgment easier to exercise by surfacing what a phrase is doing before a reader or writer has decided whether to think critically about it. The plain-language translation and citation are a starting point for a reader's own judgment, not a verdict.
What's the long-term plan — is this going to stay free, get acquired, shut down?
No plans to sell or shut it down. The founder has stated the free tier and the open dataset are core to the project's actual point, not a growth-stage loss-leader to be cut later.
Questions not covered here: af@unmask.tools