Who we serve / SaaS & AI

    Pen testing for SaaS and AI-native products.

    SOC 2 and vendor-required tests, plus dedicated coverage for AI-app threat models — prompt injection, tool abuse, RAG data-leak paths, and agentic-loop abuse. Fixed-scope, retest-included, and delivered on your audit or security-questionnaire deadline.
    Tenant latticeT-01T-02T-03T-04T-05T-06T-07T-08T-09T-10T-11T-12ModelToolsCross-tenant readIsolation · authz · prompt boundary12 tenants

    Recurring finding

    Multi-tenant platforms are the SaaS default, and a broken isolation boundary is the finding that ends deals mid-diligence. This is the archetype we pull out of B2B SaaS environments most often.

    Cross-tenant exposure

    high

    Multi-tenant platforms leaking one customer's data to another through broken isolation boundaries.

    Why it mattersA contract-breaking incident waiting on the next curious customer.

    What we test for SaaS & AI

    Multi-tenant boundaries, AI surfaces, and the questionnaire evidence your buyers ask for.

    • SOC 2 CC-scope testing with auditor-ready evidence

    • Vendor security questionnaire (SIG, CAIQ) evidence

    • AI-app threat modeling: prompt injection, tool abuse, agent loops

    • RAG and vector-store data-leak paths

    • Multi-tenant isolation and org/workspace IDOR

    • OAuth, SSO, and SCIM integration surfaces

    • Webhook signature and callback flow abuse

    • LLM key exposure and cost-abuse pathways

    • PCI DSS 11.4.7 support for customer-run external testing

    RelatedSOC 2 pen testing →·SOC 2 Type II pen testing →·PCI DSS 11.4.7 multi-tenant testing →

    How it shows up

    The archetype, anonymized.

    Multi-tenant platforms are the SaaS default, and a broken isolation boundary is the finding that ends deals mid-diligence. This is the archetype we pull out of B2B SaaS environments most often.

    SaaS & AI attack path04 hops
    1. 01 · Entry

      Tenant-scoped API token

    2. 02 · Boundary

      Object reference not bound to tenant

    3. 03 · Privilege

      Role assignment editable by member

    4. 04 · Impact

      Shared object store read across tenants

    OutcomeOne customer reading another customer's data — the finding that ends deals.

    See all recurring findings →

    From the register

    What a saas & ai report actually lists.

    SaaS & AI — findings excerpt05 rows
    Filter
    • PT-01

      Cross-tenant object access via unscoped identifiers

      Application API

      critical
    • PT-02

      Member role can self-elevate to workspace admin

      RBAC

      high
    • PT-03

      Prompt-injected tool call reaches internal endpoints

      LLM feature

      high
    • PT-04

      Object store bucket readable with any valid tenant key

      Cloud config

      medium
    • PT-05

      SSO logout does not revoke API sessions

      Authentication

      low

    SaaS & AI FAQ

    Frequently asked.

    Yes. Every AI-app engagement includes prompt-injection, tool-abuse, agentic-loop, and RAG data-leak testing as first-class objectives — not a bolt-on. We threat-model your specific system prompt, tool inventory, and data-retrieval flow, then hunt against it.

    Yes. A SOC 2-scoped external and web-app engagement runs 5 to 7 business days for Focused engagements from kickoff to signed report. Tell us the fieldwork date during scoping and we sequence to hit it without cutting validation corners.

    We spin up multiple test tenants under NDA and test the boundaries between them directly — org-to-org, workspace-to-workspace, and role-to-role. Cross-tenant leakage is a specific objective, not something we hope to stumble into.

    Next step

    Scope a SaaS or AI pen test.

    Thirty-minute scoping call. Fixed-scope quote within one business day. Report ready for your SOC 2 audit or your next security questionnaire.

    Request a pen test scope call →