Applications + APIs
Product penetration testing
Authentication, authorization, business logic and data access across the product and the services behind it.
- Web and API testing
- Identity and role boundaries
- Business-logic abuse
- Cross-customer access
Penetration testing and security review
Test the systems that carry real business risk.
Choose a focused penetration test, a broader security audit, or an AI and agent assessment. Every engagement includes hands-on testing, prioritized findings, detailed evidence, and reporting your engineering and assurance teams can use.
On-premises · Hybrid · AWS · Microsoft Azure · Google Cloud
Assessment + advisory services
Coverage follows the attack path.
A single engagement may cross several of these areas. We shape the work around the product, its deployment and the decision your team needs to make.
External exposure
Attack-surface assessment
Public applications, infrastructure and third-party paths reviewed from an unauthenticated attacker’s position.
- Asset discovery
- Exposed services
- Public application paths
- Validated impact
Native software
Mobile and desktop application testing
Client-side controls tested with the APIs and services they call, not as an isolated binary review.
- iOS and Android
- macOS and Windows
- Local storage and IPC
- Deep links and updates
AI safety + security
AI and agent security assessment
We test whether AI features stay inside their data, action and safety boundaries under adversarial use, untrusted retrieved content and model or prompt changes.
- Prompt injection and retrieval poisoning
- Sensitive data and tenant isolation
- Misuse and harmful-output controls
- Agent permissions and human oversight
Architecture + deployment
Security audit and risk assessment
We review the product and the environment beneath it—from on-premises and hybrid deployments to all three major public clouds.
- On-premises and hybrid
- Amazon Web Services
- Microsoft Azure
- Google Cloud
After the assessment
Remediation and security advisory
We help the responsible team turn a finding into the right engineering change, then test the completed fix against the original path.
- Root-cause review
- Control and architecture guidance
- Engineering working sessions
- Fix validation
AI safety use cases
Test the AI inside the product—not as an isolated model.
We examine how models, retrieval, memory, tools, people and ordinary application controls behave together. The objective is a demonstrated failure path and a control the product team can improve.
Data boundaries
Can the system reveal data it should not know?
Retrieval authorization, tenant isolation, memory, training-data exposure and sensitive output handling.
Cross-tenant disclosure · secret leakage · unsafe retrievalAction boundaries
Can an agent do more than the user is allowed to do?
Tool permissions, service identity, approval steps, transaction limits and human escalation controls.
Unauthorized actions · privilege expansion · approval bypassBehavioral safeguards
Do misuse and safety controls hold under adversarial use?
Prompt injection, jailbreak resistance, harmful-output controls, abuse paths and policy enforcement across languages and channels.
Safeguard bypass · harmful content · product misuseSystem integrity
What happens when the context, model or supply chain changes?
Retrieved-content poisoning, model and prompt changes, plugin trust, evaluation gaps and unsafe fallback behavior.
Poisoned context · untrusted components · silent regressions
Common systems
RAG applications · Copilots · Customer-service agents · Transactional agents · Content generation · Evaluation pipelines
How an engagement works
A clear scope, a controlled test and evidence your team can act on.
We agree what may be tested, how far proof should go and who to contact before any active work starts.
Define the decision
Start with the launch, customer request or risk question the assessment must support.
Set access and limits
Agree targets, accounts, exclusions, test windows and stopping conditions in writing.
Test and communicate
Reproduce credible paths and report urgent issues while the assessment is still under way.
Report and support
Deliver prioritized findings, work through engineering questions and record the status of agreed fixes.
Detailed reporting
One factual record for engineering, risk and assurance.
Decision summary
Agreed scope and exclusions, methods, material findings, demonstrated business impact, priority, limitations and unresolved risk.
Technical findings
Affected assets, prerequisites, exact reproduction steps, retained evidence, control failure and specific remediation guidance.
Assurance record
Dates, tested boundaries, finding status and approved evidence that can support customer reviews, procurement diligence and internal governance.
Remediation and verification
Recommended controls, ownership, finding status and the result of agreed fix validation.
New assessment
Scope your next engagement.
Send a short description of the system, the reason for testing and your target date. We will reply with the questions needed to shape the scope.
Scope an engagement