Secrets
Credentials that unlock your systems.
API keysPrivate keysPasswordsTokensConnection strings
Protect sensitive content before it leaves the building
Alesia inspects prompts and uploads on the device, and blocks sensitive content before it reaches an AI tool. No data ever leaves the device to be classified.
Example: in an AI chat, Alesia finds an email address and a card number in a new message. The user picks Redact and send, the values are replaced with placeholders on the device, and the AI replies using only the placeholders.
Why it matters
Most leaks aren't attacks. They're a paste.
People paste whatever gets them an answer. Usually that's fine. Sometimes it's a live key, a customer list or a contract.
Without Alesia
An engineer pastes a config file into an AI chat.
The live API key and customer PII go with it.
With Alesia
Alesia checks the paste on the device before it's sent.
The key and PII are redacted. The question still gets answered.
What it catches
Validated detectors, plus your own rules
Built-in detectors find secrets, payment data and national IDs out of the box. Add your own terms for everything else.
Payments
Validated by checksum, not just by shape.
Card numbersIBANsSWIFT codesCrypto wallets
National IDs
Check digits verified for nine countries.
DEFRITESNLBEPLUSIN
Your own terms
Exact keywords for codenames and customers, or a rule written in plain language.
"C++ code that references our billing logic"
On a match
You decide what happens next
Each policy picks one response. Start in Watch, then tighten once you trust the rule.
Block
The request is refused. Nothing is sent.
Warn
The person sees why, then chooses to continue or cancel.
Warn & redact
The person is told, and sensitive values are swapped for placeholders before sending.
Learn more about redactionWatch
Nothing interrupts anyone. The match is recorded, so you can trial a rule safely.
Private by design
Detection that never sees your data
Classification runs locally on every device. Only metadata (what was blocked, where, and why) leaves the endpoint. Alesia never asks you to route your content through a third-party cloud to protect it.
Testing
Test a model before you deploy it
Every model has test cases: content that must be caught and content that must pass. Run them after each change, pick a sensitivity, then deploy. Policies always run the deployed version.
Use cases
See it in action
Data Protection, in detail
No. Detection runs on the device. Your console receives the incident: which policy matched and what happened next, never the content itself.
Secrets such as API keys, private keys, passwords and tokens, payment data such as card numbers and IBANs, contact details, and national IDs from nine countries, each with format or checksum validation.
That depends on the policy. They're either blocked, warned with the choice to proceed or cancel, warned with the data redacted, or, in watch mode, they see nothing at all.
A determined insider can work around any on-device tool. Data Protection is built for everyday mistakes, which cause most real leaks.
Say yes to AI.
Set the rules once. Alesia enforces them on every device, and prompts stay on your devices.