Services
Build a defined AI workflow, assess an existing system, or fix a known control gap. We agree on the access, scope and written outputs before work begins.
Discuss a projectAI security & engineering
Build a useful workflow with controls from the start. We work on bounded AI workflows, model and tool integration, retrieval, access controls, evaluation, and the practices needed to run them.
Typical outputs
Agreed implementation, documented controls, evaluation results and handover.
Explore implementationUnderstand the risks before expanding access or relying on the output. We review agents, MCP integrations, retrieval systems and AI-assisted coding workflows, then explain the issues, their impact and the changes needed.
Typical outputs
Scoped findings, technical context and a prioritized remediation plan.
Compare assessment optionsClose the gap between a finding and a working fix. We make scoped changes to code, permissions, data boundaries, secrets handling, evaluation and operations, starting from our assessment or an existing report.
Typical outputs
Agreed changes, verification results, remaining issues and handover notes.
Explore remediationWe describe checks on our own implementation or remediation as verification or retesting. Independent assessment requires a separate scope and a conflict review.
Selected technical work
Some questions need a closer look at a device, an architecture, or a technical claim. We start by establishing what can be assessed with the materials and access available.
A defined device or firmware question, with feasibility and methods agreed first. Observations and limits are documented; no blanket assurance of authenticity or absence of compromise.
Explore device assessmentsSelected architecture, product-security and technical-risk reviews for engineering, leadership and counsel.
Read the technical review scopeAdvisory, briefings & workshops
These established formats remain available alongside implementation and remediation.
Technical context for board oversight, risk appetite, and launch decisions.
Independent advice for boards and executive teams on AI-system access, security controls, vendor claims, and changes that affect launch plans. Available as a briefing or recurring support.
Briefing from $7,500; advisory scoped
Briefings are fixed-fee from $7,500. Ongoing board advisory is priced around meeting frequency, access, materials, and reporting.
Discuss board advisoryAdversarial testing for AI agents and LLM-based systems.
Prompt-injection testing, memory-poisoning analysis, adversarial path modeling, and toolchain-manipulation research for systems moving toward production access.
Common scopes include AI agent production-readiness review, MCP tool-permission review, AI-assisted CI/CD workflow review, RAG boundary review, and AgentBoundary certification readiness.
Scoped per engagement
Fixed fee based on the systems reviewed and the decision you need to make.
Discuss an assessmentSenior technical judgment without a full-time executive hire.
Regular access to a Principal Engineer for product-security questions, AI-system changes, customer reviews, and roadmap tradeoffs.
Monthly retainer, scoped
Monthly retainer based on access, meeting frequency, response expectations, and system complexity.
Discuss fractional advisoryTechnical findings for counsel and leadership.
Independent technical assessment, written attestation, and litigation support for questions about system behavior, security controls, and technical claims.
Scoped per engagement
Scoped per engagement. Expert-witness hourly rates are not published.
Make an inquiryPractical AI-security exercises for your team.
Work through AI-security threats and controls with your security team or leadership, using an agenda tailored to your systems.
Startup office hours
Bring one focused question about an AI system, product security, launch readiness, customer review, or technical roadmap. One session per company; formal assessments, written reports and follow-on work are scoped separately.
Best fit: early-stage teams with a specific decision to make and enough context to use a focused working session.
Before you engage
The sample package is fictional and illustrates the structure of a review. Actual scope, findings and deliverables depend on the engagement.
For a separately scoped third-party review, read the independent AI control review. The AgentBoundary program has its own criteria, limits and public registry.
For work across client or portfolio systems, see the MSP partner kit, VC and PE portfolio review, or law firm expert review.
Scope & pricing
Published assessment packages, briefings and workshops retain their stated scope and starting prices. Implementation, remediation, device work and custom advisory are quoted after a discussion of access, complexity and deliverables.
