Warnd keeps a live inventory of every AI tool your team uses — and auto-discovers the ones they forgot to tell you about. Risk scored. Approval-ready. Built for the people who have to answer for everything.
Most teams surface 10+ unsanctioned AI tools in their first hour. And that's just the OAuth grants — sign-up emails usually double the count.
If Warnd doesn't surface tools you didn't know about in your first hour, email me directly. hello@warnd.app
Add a tool
Paste a URL — we'll fill in the rest.
Tool URL
ChatGPT
OpenAI · auto-filled from privacy policy
Tool name
Paste a URL. Warnd reads the privacy policy and scores the risk in under a second.
Recognised out of the box — plus hundreds more
485%
year-over-year growth in corporate sensitive data flowing into generative AI tools.
Cyberhaven, Q2 2024
$5.27M
average cost of a breach involving shadow data — $370K above the global average.
IBM Cost of a Data Breach, 2024
€35M
maximum fine under the EU AI Act — or 7% of global turnover, whichever is higher.
EU AI Act, in force August 2024
The status quo
Engineers paste code into ChatGPT. Marketing pipes customer data into Jasper. Finance runs contracts through Notion AI. Nobody told IT. Nobody told Legal. And when an auditor — or a vendor — comes asking, you have nothing to show them.
Spreadsheets go stale in a week. Slack threads get buried. The shared doc you started in Q1 has six entries. Warnd is the source of record that stays current — automatically.
Customer data goes into an unvetted vendor
A support rep pastes a transcript into a tool IT has never seen. The next breach notice is a coin flip.
Audit hits you with no inventory ready
You spend two weeks polling teams for a list. You still miss tools. The findings letter writes itself.
A vendor breaches and you're downstream
A third-party AI service leaks. You can't answer "did anyone here use it?" without a week of digging.
Approval takes 18 days, so people stop asking
When the official path is slow, the unofficial path wins. Shadow IT compounds — quietly, fast.
The Warnd loop
Five moves. No agents to install, no SDK, no consultants. Warnd works the second you sign in.
Google Workspace scan
143 users scanned · read-only
ChatGPT
38 users
Claude
17 users
Perplexity
9 users
Grammarly
24 users
Gamma
6 users
Most teams have authorized 30+ AI tools via Google sign-in without anyone keeping track. Warnd finds them all — read-only, no agents, no installs.
What we read
Active users · Third-party OAuth grants · App display name + scopes
What we never read
Email content · Drive or Calendar files · Passwords or content of any kind
ChatGPT
OpenAI
Notion AI
Notion
GitHub Copilot
GitHub
Midjourney
Midjourney
Jasper
Jasper AI
Cursor
Anysphere
Stop asking 'what are we using?' every time something happens. Warnd keeps a live, searchable inventory of every AI tool across every team — risk level, approval status, owner, and the URL that started it all.
ChatGPT
chat.openai.com · Free plan · privacy policy analysed
Data retention
How long is your data stored?
Third-party sharing
Data shared with external parties?
GDPR compliance
Regulatory alignment & certifications
Trains on your data
User inputs used to train models?
Security certifications
Formal security standards held
Paste a URL and Warnd reads the privacy policy in under a second. It scores the tool across five compliance dimensions and gives you a plain-language summary your CFO and your engineers will both understand.
Review queue
2 pending · 12 reviewed this week
Jasper
Jasper AI · jasper.ai
Midjourney
Midjourney · midjourney.com
Slow approval is why people stop asking. Warnd turns review into a 30-second decision: every submission lands in a queue with the risk score, the URL, and a clear approve/deny path. Every action is logged.
Inventory tells you what your team uses. Policy + playbook tell them how to use it safely. Every tool in Warnd gets its own policy (who can use it, for what, what's forbidden, when to escalate) and its own playbook (starter prompts, do's and don'ts, the example flows that actually work). Wiki-edited by the team, audit-logged on every change.
Who it's for
When the auditor, the board, or the breach notice shows up, one person owns the question "what are we using?". Warnd is for them.
The IT / security lead
"I found out about the AI notetaker from an invoice."
The compliance owner
"The audit asked for our AI register. We didn't have one."
The founder / ops lead
"Enterprise deals now come with an AI questionnaire."
Vs the status quo
Two real things compete with Warnd for your time and budget. Here's how each one actually breaks down on the dimensions that matter for AI governance.
Option A
Option B
Option C
The deadline
Brussels moved the high-risk deadline once — the 2026 date became December 2, 2027. The obligations didn't shrink, and every one of them starts from the same place: knowing what you use. An inventory with an EU AI Act tier on every tool is step zero. Start now and you're the calm one in the room.
December 2027
Get inventory-readySecurity, up front
Read-only by design
The Google connector uses read-only admin scopes. It can look — it can never touch, send, or change anything.
Nothing to install
No agent, no browser extension, no SDK. Nothing of ours runs on your machines, ever.
Isolated by tenant
Your workspace is walled off with Postgres row-level security. OAuth tokens are encrypted at rest with AES-256.
Leave with everything
Export your inventory and audit history as CSV, JSON, or PDF at any time. Disconnect the connector in one click.
Pricing
Start free. No card required. Upgrade when you need approval workflows, auto-discovery, and a longer audit history.
For solo evaluators or tiny teams — manual inventory of up to 5 tools
Forever · no card needed
1 workspace · 3 members · 5 tools
Start freeFor 5–50 person teams that need real AI governance
per month · billed monthly
14-day free trial · no card required
1 workspace · 50 members · 100 tools
Start with TeamFor organisations with stricter compliance, audit, or multi-entity needs
per month · billed monthly
Need more capacity or a custom contract? Talk to us about Enterprise →
Common questions
How does the risk scoring actually work?
When you add a tool, Warnd fetches and reads its privacy policy. It checks data retention, third-party sharing, training-on-data clauses, GDPR alignment, and security certifications — and produces a scored summary in under a second. You can re-scan any tool at any time.
Can my whole team submit tools?
Yes. Invite teammates and let them add tools they already use. Submissions land in the owner's review queue. Open workspaces let members publish directly; required-approval workspaces gate everything until an owner signs off.
Is the Google Workspace connector safe?
Yes — and we built it to be auditable. We request read-only Admin Directory and token-list scopes only. We never read email, Drive, Calendar, or any user content. Refresh tokens are encrypted at rest with AES-256. You can disconnect any time and your discovered apps stay in your workspace for reference.
How is this different from a spreadsheet?
Spreadsheets don't score risk, don't enforce approvals, don't maintain an audit trail, and go stale within a week. Warnd does all of that, and stays current automatically.
What happens when a privacy policy changes?
Re-scan with one click. Warnd re-reads the latest policy and updates the score. You're always working from current information — not a snapshot from the day someone first added it.
Where does my data go?
Tool metadata and risk scores live in your workspace, isolated by tenant in Postgres with row-level security. Privacy-policy fetches go through our service; nothing about your team is sent to third parties for advertising or training.
Do I need to install anything?
No. Warnd is a web app. There's no agent, no browser extension, no SDK. Sign in, paste a URL, you're running.
Every week without a real inventory is another week of unreviewed tools, unchecked risk, and audit exposure. Warnd fixes it in five minutes — for free.
Free forever · No credit card · Setup in 5 minutes