Trust & security
You're pasting legal text into a website. You should know exactly what happens to it. Here it is, in plain language.
Your text is never stored
When you submit text, it travels to our API, is processed in memory, and the result travels back. We keep no copy of your text: not in a database, not in logs, not for "quality improvement", not for training anything. What we do record is a count: how many term matches your key used this billing period (on the Workspace plan), or how many billable operations (on the Developer API). That's how billing works.
The same promise covers the homepage demo: text you paste there is processed and returned, never stored. The demo needs no key and no signup; to keep it fair it allows 5 texts an hour, which we count against a one-way hash of your network address (never the address itself), and even those counter rows are deleted within a day. We also count how many demo runs happen each day: one number per day, nothing about what was pasted or who pasted it.
No AI middleman
No language model runs when you use this service. Suggestions are served from a fixed, curated glossary of 7,257 reviewed legal terms (as of the 2026-07-06 update), and your text is never sent to an AI provider. AI helped build the glossary; that work was finished and frozen into fixed entries before your text ever arrives (see how the glossary is made). If you want a full AI rewrite, we give you a prompt to carry to your own model, under your own account and policies. You stay in control; we stay out of the loop.
Auditable by design
Every highlighted term traces back to one glossary entry. Open its detail card and you can read the meaning, what to watch for, and real examples for yourself. The system is deterministic: the same input always produces the same output, every time. So when a suggestion is wrong, it is wrong in a way we can find and fix at the source. A model's one-off guess leaves nothing to inspect.
How the glossary is made
Most entries are drawn from plain-language resources published by United States government agencies and courts: glossaries, style guides, and examples of legal ideas explained well. We use those public materials; we are not affiliated with or endorsed by any government agency.
The rest were distilled from partial definitions spread across many sources. No entry rests on a single source, and no entry was written by a single AI: at least five different AI models took part in the curation, with different jobs (drafting, combining, cross-checking). A human then verified the result: every piece of data in an entry had to be corroborated by following the lead of other definitions, or confirmed in partial form in yet others. The plain-language wording follows published examples: government plain-language materials, health-literacy guides, and open example collections. This took several months, and it continues.
The scope is United States legal language: terms of art, multi-word phrases, and legal senses of everyday words. Terms with more than one sense carry their contexts on the term's detail card. Updates arrive as reviewed batches, recorded in the release notes below: no silent edits.
One honest limitation: a deterministic system can still be deterministically wrong. When an entry is wrong, tell us: hello@plainlanguage.us. It gets fixed at the source, and the fix reaches every future response.
Glossary release notes
Every glossary update lands as a reviewed batch. The batches so far:
- 2026-07-06: 7 new terms and 3 revised entries, with 42 new phrase variants the matcher now recognizes.
- 2026-06-28: about 550 new terms from the curation pipeline, and about 240 entries gained plain-language renderings they had been missing.
Current size: 7,257 terms (as of the 2026-07-06 update). This list grows with each update.
Where your data actually goes
The path of a request: your text travels over TLS to Cloudflare's edge, where our code runs. The match happens in memory against the bundled glossary, the response returns the same way, and your text is written nowhere along the way. There is no other server behind it: nowhere else for your text to go.
- Cloudflare runs the site, the API, and our billing database on its global network (their security practices: the Cloudflare Trust Hub). Your text passes through Cloudflare's infrastructure only, and our worker writes no logs: the code contains no logging calls, and Cloudflare's optional worker-logging features are switched off.
- Stripe handles payment and billing identity (your email, your subscription), under Stripe's privacy policy. Your card number never touches our systems, and your pasted text never touches Stripe.
- That's the whole list. No analytics trackers on your text, no advertising pixels, no data brokers. (The website's pages use Cloudflare Web Analytics, a cookieless page-view counter; it never sees what you type or send.)
What we do keep: a hash of your API key, your billing records, and a usage counter per customer (how many term matches or billable operations, this billing period). Numbers and billing facts, never content.
If that list of providers ever changes, the privacy policy is updated first, with a new effective date, before a new provider touches any data.
Your API key
We store only a cryptographic hash of your key. The key itself is shown to you once and never again. If it's lost, you roll a new one; if it leaks, you roll and the old one dies instantly. In the browser workspace, the key is kept in your browser's local storage, on your machine, and you can forget it there any time.
If something goes wrong
Found a security problem? Tell us at security@plainlanguage.us (the address is also published in security.txt). If a security incident affects your data, we will email affected account holders without undue delay (our target is within 72 hours of confirming it) and post a notice on the privacy policy page.
Billing you can't be surprised by
- Workspace: hit your monthly limit and requests simply pause: no overage charges, ever.
- Developer API: no monthly fee, no minimum. Every response says what it just cost, your running total is one free call away, and the monthly invoice covers exactly what you used.
- Cancellation is self-serve in the billing portal. No phone calls, no dark patterns.
Accessibility
This site is built to meet WCAG 2.1 AA: keyboard-operable throughout (including term cards and collapsible panels), visible focus indicators, a contrast-checked palette in both light and dark modes, and no meaning carried by color alone. Highlighted terms are also underlined and weighted.
How we check: keyboard-only passes over every screen and state, and contrast checks on the full palette in both themes. The honest limits: this is self-assessment. No outside audit has been done yet, and we don't claim certification. If something doesn't work with your screen reader or keyboard, tell us at hello@plainlanguage.us and we'll fix it.
Who's behind this
PlainLanguage.us is a product of extrAIntelligence LLC, an independent commercial service not affiliated with any government agency. It finds legal terms in the text you paste and suggests plain-language meanings from a reviewed glossary. Questions or concerns: hello@plainlanguage.us.
The fine print, honestly
Plainlanguage.us provides suggestions for legal terminology; it is not legal advice and is not a substitute for a lawyer. The terms of service and privacy policy say this at slightly greater length. They're short and written like this page.