Methodology and artifacts
Review work is anchored in system boundary, authority model, abuse cases, supporting artifacts, and remediation path.
- Approach page
- Sample review package
- Enterprise buyer packet
Proof and trust
Credibility without overclaiming.
A central trust page for methodology, sample artifacts, security practices, founder background, support programs, publications, and certification limits.
What you receive
Concrete outputs for a specific decision.
- Methodology summary
- Sample artifacts
- Security and data handling practices
- Founder background
- Program and support disclosures
- Publication and research links
- Certification scope and limitation language
Work product
Trust materials index
This page links the claims a buyer may rely on to the supporting material available on the site or through a scoped diligence request.
Founder background
Principal-led review by Kevin O'Connor, with senior national-security engineering and advanced threat research leadership experience.
- About
- Media and speaking
Security practices
Public security contact, PGP path, vulnerability disclosure route, NDA-first engagement handling, and instructions not to send sensitive materials through public forms.
- Security
- PGP
- Security diligence FAQ
Published analysis
Open research and field notes cover AI-agent authority, MCP/tool risk, RAG boundaries, prompt injection, and production-readiness review.
- Insights
- OWASP Top 10 for AI
- Field Notes
01
What is supportable
Methodology
Assessment work starts from the decision, system boundary, authority model, abuse cases, artifacts, and remediation path.
Background
Principal-led review grounded in national-security engineering, advanced threat research, AI-security assessment, and high-assurance technical communication.
Programs and support
TKOResearch may disclose approved program participation or support only where supportable and relevant.
02
What is not claimed
- No customer logos without explicit approval.
- No certification claims beyond a defined scope and date.
- No guarantee that an AI system is safe in every configuration.
- No broad assurance where only a narrow review was performed.
Next step