Use cases
The risk isn't the AI tool. It's what gets pasted into it.
People use AI to debug, summarize, draft, and analyze real work — so real secrets and personal data end up in the prompt box. Here's what Data Loss Prevention for AI tools looks like on the sites your team already uses.
- ChatGPT
- Claude
- Gemini
- Microsoft Copilot
Free for individuals • per-device pricing for organizations
Engineering · debugging a deploy
$ deploy --prod # keeps failing
AWS_SECRET_ACCESS_KEY=AKIA••••••••EXAMPLE
DATABASE_URL=postgres://10.2.0.14/app
Caught before it sent
1 credential found · AWS access key. Redact it and the rest of the config still goes through.
Why people paste sensitive information.
It isn't carelessness. The same handful of everyday tasks are exactly the ones that pull real data into a prompt.
Debugging
Pasting an error, a log, or a config to find a fix — the fastest way to unblock.
Analysis
Dropping in a spreadsheet or export to summarize, rank, or pull out the numbers that matter.
Writing
Handing over a contract, memo, or ticket to draft, rewrite, or tighten the language.
Real moments
You'll recognize these. So will your team.
None of them are careless — they're normal ways people work with AI. That's exactly why the check has to live in the browser, where the data is written.
Debugging a failed deploy
“Why does this keep crashing in prod?” — pasted with AWS_SECRET_ACCESS_KEY=AKIA…EXAMPLE still in the config.
InputGuard flags the live key at paste and offers to redact it before the prompt sends.
Cleaning up a headcount sheet
“Rank these reports by tenure” — one row reads Dana Lee · SSN 502-44-1990 · $148,000.
The SSN is caught and redacted, while the rest of the sheet still goes through for the analysis.
Summarizing a customer export
“Which accounts are overdue?” — the CSV holds jordan@acme.com and card 4111 1111 1111 1111.
Card numbers and emails are found in the attachment before it's ever uploaded to scan.
Redrafting a contract clause
“Tighten this indemnity clause” — between Acme Corp and Initech LLC under Project Atlas.
Business-sensitive party and project names are flagged so the team decides before sharing.
Drafting a reply from a ticket
“Write a response to this customer” — the thread still has (415) 555-0148 and account #80231.
Personal details are caught in the moment, so they don't get pasted into the draft.
Different teams, the same blind spot.
- Consultants & agencies
- Legal & compliance
- Finance
- HR & people
- Engineering
- Support & sales
For organizations
Make AI safe to allow across the team.
Blocking AI outright just pushes people to personal accounts. InputGuard gives teams browser-native AI DLP, so you can allow the tools people need while managing how they're used.
Managed from the admin portal
- Decide which AI sites are protected, allowed, or blocked
- Roll policy out across devices instead of trusting each person to configure it
- Review metadata-only audit activity without reading prompts
- Give a safe default to new AI tools before anyone uses them
Start with your own browser. Scale to your team.
Join early access to protect the AI tools you already use, then roll it out when you're ready.
Free for individuals • per-device pricing for organizations