Privacy Campaign

What does our server read?
Nothing.

While most writing apps scan your texts to feed AI models, we built a platform where the server doesn't see — and never will see — what you write.

No credit card required14-day free PRO trial
The real problem

Your texts are feeding machines.

Terms of service nobody reads

Most writing platforms include clauses that allow the use of your content for 'product improvement' — which almost always means training AI models with what you wrote.

Unpublished manuscripts in public datasets

Researchers have already found fragments of unpublished books, in-progress screenplays, and confidential articles inside datasets used to train LLMs.

Your favorite editor may be a silent spy

Features like smart autocomplete, writing suggestions, and contextual correction frequently send your text to external servers — without encryption, without explicit consent.

Journalists losing source protection

Documents stored in cloud without E2EE can be accessed by the provider's employees, hackers, or governments via court orders. Your sources depend on your tool.

What if there was an editor where the server simply can't read anything? Where every word is encrypted in your browser before touching the network? Where not even Writtt itself has access to what you write? That editor exists.

Your ideas are not our training product. Every word you type is shielded with AES-256-GCM in your own browser. What reaches our servers is pure cryptographic noise — no meaning, no value, no use for any AI. We don't read. We don't train. We don't touch. The rule is simple: not even we can read what you write.

Zero-knowledge architecture means the encryption key is derived from your password via Argon2id (64MB, 3 iterations) and never leaves your device. The server stores only cryptographic blobs prefixed with enc: — data that is literally noise without the key.

Who it's for

Built by people who understand the value of an unpublished text.

Writers

Your next novel, short story, or essay shouldn't appear in an AI training dataset before it's published. Protect every draft.

Journalists

Confidential sources, ongoing investigations, sensitive drafts. Every byte needs protection that goes beyond a marketing promise.

Screenwriters

Pitches, treatments, and in-progress screenplays are valuable intellectual property. Don't risk a leak by using the wrong tool.

Researchers

Data under ethical compliance (GDPR, IRB), research notes, theses under construction. E2E encryption meets regulations by design.

How it works

Simple. For you — and impenetrable for everything else.

01
// seu manuscrito
It was a dark night...

You write normally

The editor works like any premium editor — markdown, rich formatting, focus mode. The difference is that everything happens locally, in your browser.

Text exists only in your browser
02
// AES-256-GCM
enc:7f2a9c...b4e31f
client-side · browser only

Automatic encryption

Before any transmission, your text is shielded with AES-256-GCM on your own device. The key is derived from your password and never leaves the browser.

Encrypted locally
03
a3 7f 91 c2 e8 4b 0d f5
b1 6c d4 88 2e 79 5a b3
f0 1d 63 c7 9e a2 4b 8f
// this is all the server sees

The server receives noise

What reaches our servers is just a cryptographic blob — no meaning, no structure, no utility. We don't train AI, we don't read data, we don't touch what's yours.

Zero AI training
Readable text never touches the network · Server sees only noise · AI never trains with your data

Do you know what happens to what you type?

CriteriaWRITTT PROTraditional platforms
Is your text used to train AI?Impossible — E2EEFrequently yes (vague terms)
Can the provider read your documents?No — zero-knowledgeYes — server-side access
Protection against court orders?Yes — we don't have the keyNo — data is accessible
Real end-to-end encryption?AES-256-GCM in the browserTLS in transit only
Can employees access content?Impossible — zero-knowledgePossible — server-side encryption
Export without vendor lock-in?PDF, Word, MarkdownProprietary formats
FAQ

Questions from people who write for a living.

Do you really not train AI with my texts?+
It's not a matter of policy — it's a matter of architecture. Your texts are encrypted in your browser before any transmission. What reaches our servers is cryptographic noise. We can't read it, therefore we can't train. It's mathematically impossible.
What about the editor's AI features? Are my texts sent to OpenAI/Anthropic?+
Communication is direct: your browser connects to the AI provider's API (OpenAI, Anthropic, or Google) without going through Writtt. We never see the content sent nor the response returned. The API key is stored encrypted in your profile.
What if Writtt is sold or changes ownership?+
Zero-knowledge architecture guarantees that, even in an acquisition scenario, the data is cryptographically inaccessible to any new owner. Without your password, the stored blobs are meaningless noise.
Can I export everything and leave if I want?+
Yes. Full export in Markdown, vector PDF, and native Word. No vendor lock-in, no tricks to keep you. Your data is yours.
What's the difference between 'we don't use your data' and 'we can't access your data'?+
Everything. 'We don't use' is a policy promise that can change tomorrow. 'We can't' is a verifiable cryptographic guarantee. AES-256-GCM + zero-knowledge architecture = data exists as noise on our servers. It's not a choice, it's an intentional technical limitation.

Your manuscript deserves a vault.

Start writing in absolute privacy. No credit card, no obscure terms, no fear of feeding algorithms with what's yours.

Start now — 14 days free
No credit card required14 days of full PRO
Your texts will never be used to train AI. Never.