Increasingly yes, and we do not pretend it is universal. Some assessors are satisfied with the platform report, which is often more detailed than a manual one because every finding carries reproduction steps and evidence. Others want human attestation. For those, our in-house pentesters review the results and sign off, as an add-on. If your assessor has a view already, tell us early and we will scope to it.

It gives you continuous testing and continuous evidence across your estate. Talk to your assessor about your specific scope.

Escape maps findings to the OWASP Top 10, OWASP API Top 10, PCI DSS, SOC 2, DORA, CRA, HITRUST CSF, NIS2, HIPAA and GDPR and more, giving you audit-ready evidence of continuous security testing.

AI pentesting is built for the application and external network layers, with findings mapped to the controls they affect. PCI DSS also requires internal network penetration testing and a segmentation test twice a year, which is a different exercise.  Reach out to us about your specific scope and what we cover today.

Escape is a multi-model system. Escape routes work across its own models, running on Escape-controlled infrastructure, and frontier commercial APIs from OpenAI, Anthropic, and Google. No single provider powers the product end to end, because crawling, exploit design, validation, and evidence write-ups each reward different model strengths. The routing, prompts, skills, and reporter loop are Escape IP. Customer data isn't used to train any of these models. Private deployment is available, and internal or non-public systems are reached through a private location rather than by opening anything up. We do not train models on customer data. More information on data handling.

Testing is non-destructive by default, scoped to an allowlist, rate limited, and restricted to the window and source IPs you set. There is a global kill switch. Most regulated teams start on staging with production-like data, then extend once they have seen a full assessment.

That is what customers tell us, for two reasons. The obvious one is finding issues before a researcher submits them. The less obvious one is triage: upload your past bounty and pentest reports, and Escape validates what is still open, what is already fixed, and what was never real, then converts the live ones into regression tests. It is not a replacement for your programme. It reduces the duplicate and low-value end of it.

Two layers before anything reaches you. A validator agent replays the finding with no prior context to check that it reproduces, and a false positive detector reads it against how your application normally behaves. Flagged items are not silently deleted, so you can open them and see the reasoning, and anything you mark yourself trains the agent so it does not come back. Results improve as you feed in context: roles, access policies, threat models and documentation. If you need a zero false positive report for a specific engagement, our pentesters can verify findings by hand.