The safety gate before ChatGPT, Claude, and Gemini

Use AI without
leaking what matters.

GuardInput scans prompts and files before they reach ChatGPT, Claude, Gemini, or Copilot, detecting sensitive data, masking it, and keeping a clear AI Receipt.

GuardInput Scanner, Active
Masking
Input to scan
HIGH RISK · 87

Hi PERSONSarah Chen,

Q3 review for PERSONMarcus Hale (EMP-4472).

Salary: AMOUNT$94,500

CC: EMAILj.cole@acme-corp.com

EMAILhr@acme-corp.com

Token: APIKEYsk-proj-7fK9mNpQ2vRxL…

Portal: URLhr.acme-corp.com/Q3

AI-safe version
MASKED COPY

Hi PERSONSarah Chen,

Q3 review for PERSONMarcus Hale (EMP-ORG4472).

Salary: AMOUNT$94,500

CC: EMAILj.cole@acme-corp.com

EMAILhr@acme-corp.com

Token: APIKEYsk-proj-7fK9mNpQ2vRxL…

Portal: URLhr.acme-corp.com/Q3

Detected3 people2 emails1 amount1 API key1 URL

Local-first scanning

Sensitive detection runs client-side before any data leaves your browser.

Explainable detection

Every flag tells you exactly what was found, which category, and why.

Readable receipts

A structured log of every AI interaction, ready for audit or compliance review.

Before / After

Before it goes to AI,
see what it contains.

GuardInput reads your prompt or document and highlights every piece of sensitive data, before you hit submit.

Before, raw prompt
High risk

Hi there,

I need help reviewing the contract for PERSONClaire Dubois.

She works at ORGLemaire & Partners and her retainer is AMOUNT€12,400.

Email: EMAILclaire.dubois@lemaire.fr

Direct line: PHONE+33 6 12 34 56 78

API token: APIKEYsk-abc123xyzQ9…

After, AI-safe version
Ready to review

Hi there,

I need help reviewing the contract for [PERSON_1].

She works at [ORG_1] and her retainer is [AMOUNT_1].

Email: [EMAIL_1]

Direct line: [PHONE_1]

API token: [API_KEY_1]

Detection categories

EmailsPerson namesPhone numbersFinancial amountsURLsAPI keys & tokensDocument metadataSensitive context
Files

Drop a document.
Get back a safe copy.

Not just prompts, GuardInput reads whole files, strips hidden metadata, masks every sensitive field, and hands you a clean copy to attach. The original never leaves your machine.

GuardInput, Document Scanner
Waiting for file

Drop a file or browse

PDF · DOCX · TXT · CSV · up to 25 MB

Findings will appear here

Risk reduced

91

Reads real formats

PDF, DOCX, TXT and CSV, parsed locally in the app.

Sees hidden data

Author tags, GPS, track-changes and embedded metadata.

Exports a clean copy

Masked, logged on a receipt, ready to share with AI.

AI Receipt

Every sensitive AI interaction
should leave a trace.

GuardInput

AI Protection Receipt

Receipt#7FK9-MNPQ
DateJun 23, 2026 · 10:30 UTC
Sent toClaude
DocumentPDF · 3 pages

Protected before sending

3× person namesreplaced
2× email addressesmasked
1× financial amountmasked
File metadataremoved

Risk reduced

87 24

VERIFIED SAFE
Keep this receipt for your records

Human-readable summary

Document prepared for Claude

PDF · 3 pages · Submitted 10:30 UTC

Initial risk: High (87/100)

Detected: person names, emails, financial amount

Actions taken

Names replaced · Emails masked · Metadata removed

Prepared risk score

24 / 100

LOWER RISK
0100
DesktopComing soon

Protection that lives
in your menu bar.

A lightweight desktop agent watches every paste into ChatGPT, Claude or Copilot, and any app, catching sensitive data before it ever leaves your machine. No copy-paste into a separate tool. It just sits there and guards.

Desktop · macOS · Windows
10:30
AI
Claude, New chat

Message

Summarize this for PERSONSarah Chen — invoice AMOUNT$94,500, reply to EMAILsarah@acme-corp.com
Send

Works everywhere

Any AI app or browser tab, intercepts at the OS level.

Stays local

Detection runs on-device. Nothing is uploaded to scan.

One-click mask

Approve from a popup, send safely, keep the receipt.

Use cases

For every team that touchessensitive data.

Legal

Summarize a client contract safely

Extract clauses from NDAs, service agreements, and litigation docs without exposing party names, financial terms, or confidential details to the model.

Person namesContract amountsOrg names
HR

Clean CVs and performance reviews

Run AI analysis on employee records and assessments while removing identifying data before it reaches any public model.

Employee namesSalary dataMedical info
Finance

Remove sensitive reporting details

Prepare financial summaries, audit notes, and board reports for AI assistance, without exposing figures, bank details, or account data.

IBAN numbersAmountsClient codes
Support

Prepare customer tickets for AI

Anonymize complaints, chat logs, and email threads before routing them through AI triage or summarization tools.

Customer PIIOrder IDsAddresses
Engineering

Catch API keys before prompting

GuardInput detects secrets, tokens, credentials, and internal endpoints before they are copy-pasted into Claude, ChatGPT, or Copilot.

API keysAuth tokensInternal URLs

More coming

Recruitment, accounting, research, media, and more. Tell us what you need.

Submit a use case
Compliance

Make AI policies usable
at the moment of copy-paste.

GuardInput helps teams reduce unnecessary personal and confidential data sent to public AI tools. It supports better habits around GDPR, AI Act readiness, NIS2 risk awareness, and internal AI usage policies, without claiming automatic compliance.

GuardInput is a risk-reduction tool, not a legal certification. It helps you work more carefully, compliance decisions remain yours.

GDPR

Minimize personal data sent to third-party AI processors before submission.

EU AI Act

Build internal readiness habits before obligations formally apply.

NIS2

Reduce risk surface in AI-assisted workflows for covered entities.

AI Usage Policy

Enforce your organization's AI rules at the prompt level, not after the fact.

Early access

Bring the input gate
to your AI workflow.

We are building GuardInput for teams who take data handling seriously. Join early and help shape what it becomes.