Your systems. Your rules.
We test only what you authorize, isolate every engagement inside single-tenant August Sandboxes, and never train AI models on your data.
- #0114:02:11.482ZActor A request capturedprev_sha256ec6a5b2f…2d21entry_sha2567faf948b…f374
- #0214:02:11.739ZCross-account 200 loggedprev_sha2567faf948b…f374entry_sha25644f90778…0739
- #0314:19:54.006ZReviewer sign-offprev_sha25644f90778…0739entry_sha2568eece0c8…f80c
- Retest16:41:27.310Z403 Verifiedprev_sha2568eece0c8…f80centry_sha256b859d95d…49b2
Before and during testing
- 01Zero active testing before a written Rules of Engagement is signed
- 02Test accounts you control, never real customers
- 03Dedicated egress allowlist IPs issued per engagement in the signed RoE
- 04Agreed time windows and immediate stop on request from either side
- 05No denial of service and no social engineering unless you sign for it
- 06Engineer-verified findings with a signed Letter of Attestation
Your data
- Zero model training on your source code, traffic, prompts or findings
- Zero data retention across CyberCART API prompts and completions
- Scope-locked Linux VMs that reach only verified domains
- We keep only the evidence required for your report and retest
- 30-day artifact deletion on request, or earlier per your contract
Our promise ›
We behave like the attacker, never like a risk. You decide the scope. We prove what breaks, help you close it and leave nothing behind.
Architecture & Sandbox Isolation ›
Hard technical boundaries on every run.
Every assessment runs inside isolated single-tenant infrastructure enforced at the kernel, network and contract layers.
- 01Sandbox isolation
Single-tenant August Sandboxes with kernel and DNS scope lock
Every engagement runs inside its own single-tenant August Sandbox VM. Outbound networking is locked at the kernel firewall and DNS resolver strictly to domains you authorize, dropping all other egress by default.
- 02Egress allowlisting
Dedicated egress allowlist IPs per engagement
We do not share static public scanning pools across customers. Dedicated source egress IPs are provisioned per engagement and recorded in your signed Rules of Engagement before testing starts so your WAF and SOC can verify every request.
- 03AI data boundary
Zero model training on client code or traffic
Your source code, HTTP wire captures, prompts and findings are never used to train or fine-tune any AI model. Inference runs only on contracted endpoints with short or zero retention.
- 04Artifact retention
30-day artifact deletion on request
We retain only the proof needed for your report and post-patch retest. Sandbox volumes, raw captures and engagement artifacts are deleted within 30 days on request, or sooner when specified in your contract.
- 05Authorization gate
Zero active testing before written RoE execution
No active probing, scanning or exploitation begins until both parties execute a written Rules of Engagement defining in-scope staging hosts, test accounts, testing windows and emergency stop contacts.
Questions about security?
Ask for our security and data handling details before you sign.