Security and trust
We test your security. Here is how we protect it.
Customers trust a pentest firm with their staging apps, their test accounts and a list of their weaknesses. This is how we handle that trust.
Data handling
How we handle your data.
Secrets encrypted at rest
Test users' passwords, two-factor secrets, saved sessions and your certificates are encrypted at rest. Passwords are never shown in the app.
Deleted after 30 days
Engagement data, including the agent's workspace, requests and evidence, is deleted 30 days after you accept the final report.
Deliverables kept 6 years
Signed reports and letters are kept for 6 years, so anyone holding a copy can still verify it when an auditor asks.
Never used to train models
Customer data is never used to train models.
Encrypted detailed report
The detailed report holds working exploits, so it is released encrypted, with a password your admins see in Ambit and that is never sent by email.
Least access
The agent's container receives only its own engagement's scope and test users. It never holds Ambit's keys.
Agent safety
The controls around the agent.
- Verified scope
- Testing touches only domains your admin proved control of, by DNS record or file. Proof is checked again daily; if it disappears, testing pauses.
- Written authorization
- Nothing starts until someone with authority signs the authorization to test. Once signed, the scope is locked.
- One way out
- The agent runs in an isolated container on a network whose only exit is Ambit's proxy. Anything that ignores the proxy has no route.
- Enforced rate and hours
- The proxy paces every request to the rate you chose, and the agent only tests during the hours you allowed.
- Identified traffic
- Every request carries an Ambit-Pentest header with your engagement reference and an AmbitPentest/1.0 user agent token.
- A log of every request
- The proxy writes one line per request, outside the agent's reach: the record of everything we sent, and when.
- No evasion
- No captcha solving, no getting around bot protection or a WAF, no residential proxies, no account creation. The agent asks you instead.
- No load testing
- Limits are checked with a handful of requests. Ambit never floods your app.
- Human review
- A certified pentester leads every engagement, reviews the findings before they are released to you, and signs the report.
- Stop at any time
- Your team can pause or stop a run from the engagement page.
Subprocessors
Who else touches customer data.
The providers we use to run Ambit.
| Provider | Purpose |
|---|---|
| Amazon Web Services | Hosting, storage and the isolated test runners |
| Anthropic | The AI model the agent uses |
| WorkOS | Sign-in to the Ambit app |
| Kernel | Live browser sessions, for signing in your test users |
On request
Documents for your security review.
- Data processing agreement For engagements that touch personal data, or for EU and UK customers.
- Master services agreement For customers who want a reusable agreement for every test.
- AI testing addendum How AI is used in your test, for security reviews that ask for it separately.
- Authorization, rules of engagement and statement of work Filled from your test's setup and signed in the app. Templates on request.
- Sample report and attestation letter Public, on the sample reports page.
Report a vulnerability
Found an issue in Ambit?
Email [email protected] with what you found and how to reproduce it. Read the policy first: it says what is in scope and how we work with you.
Get started
Your next pentest, in days.
Tell us what to test. A certified pentester leads the engagement, the agent does the legwork, and you get a signed report your auditor can use.
Fixed price. Results in days. Retest included.