● 100% free · 60 seconds · no card

Can your AI agent hurt you today?

You wired an AI agent into your email, your data, maybe your payments. Here's the question nobody asked first: what could it do if it went wrong? Find out in 60 seconds.

6pass/fail gates

20controls in the full standard

60sto know where you stand

↓

The free check

Check every box you can honestly say yes to

If you can't point to real proof — a config, a log, a written policy — leave it blank. Guessing "yes" is how people get hurt.

1. Do outside-facing agents read only cleared data?

No agent talking to customers, web, or email can reach your sensitive raw data.

2. Is untrusted input structured before an action-taking agent sees it?

Raw emails/web text never flow straight into an agent that can do things.

3. Is no single agent both reading the world AND able to send/publish/pay?

The #1 exploited failure — one agent reads, a different gated agent acts.

4. Are secrets in a manager, never in files an agent can read?

No API keys in .env, config, or prompts the agent can echo back.

5. Do irreversible actions wait in staging for a human to approve?

Sends, payments, deletions, publishes don't fire until a person promotes them.

6. Is there one kill switch every agent checks before it acts?

One switch stops all agent activity — and you've actually tested it works.

Get the printable checklist — free

We'll email you the full "Can Your AI Agent Hurt You?" one-pager — yours to keep, share with your team, and run against any vendor's AI.