TriOngle

AI technology and security architecture for LLM and agent systems.

Designed secure, tested by attack, governed to standard.

See the servicesWrite to us
The TriOngle constructionAn equilateral triangle. Its base is labelled Architect and drawn as a solid line; its right side is labelled Assure and drawn dashed; its left side is labelled Govern and drawn as a chain line. A circle inscribed in the triangle touches all three sides and is labelled LLM and agent systems.ArchitectAssureGovernLLM and agentsystems

Three sides of one triangle

The name carries the method. Every engagement is drawn on three sides, and every scope belongs to one of them.

Architect

Design the AI system secure from the start: trust boundaries, identity, retrieval and tool permissions decided before the code ships.

Drawn as the solid base. Scopes A1 to A3.

Assure

Attack it the way an adversary would, prove every finding with a proof of concept, and prove the fixes.

Drawn dashed, like an attack path. Scopes S1 to S4.

Govern

Meet the EU AI Act, ISO/IEC 42001 and GDPR in the same engagement, and be ready when something goes wrong.

Drawn as a chain line, the reference the others align to. Scopes G1 to G3.

What we do differently

Six commitments, each a standing rule of how TriOngle works.

  • Fixed scopes, published

    Every service on this site is a fixed scope with named deliverables, so you can see what you get before you ask what it costs. Prices are on request while we launch; each quote is for a scope agreed in writing.

  • Every finding mapped

    Each finding carries its place in the OWASP Top 10 for LLM Applications (2025),⁠[1] the OWASP Top 10 for Agentic Applications (2026),⁠[2] MITRE ATLAS,⁠[3] NIST AI 600-1,⁠[4] ISO/IEC 42001 Annex A,⁠[5] ISO/IEC 27090:2026⁠[6] and the EU AI Act article it touches. See how the mapping works.

  • Security, privacy and the AI Act in one report

    One engagement and one plain-language report for engineers, the board and your counsel. After Regulation (EU) 2026/1744, the AI Act’s rules for high-risk systems listed in Annex III apply from 2 December 2027.⁠[7]

  • Agents and MCP as fixed scopes

    Twelve agent scopes, from MCP server audits to agent identity design, each mapped to the OWASP agentic list of 9 December 2025⁠[2] and to the MCP security best practices.⁠[8]

  • A dated, versioned method

    The TriOngle Method v1.0, published on 4 November 2026, is public and vendor-neutral. Every report states the method version and the version of each framework it maps to. Tooling we build for the method is published openly.

  • Post-quantum readiness inside AI work

    NIST IR 8547 became final on 27 October 2026.⁠[9] It deprecates quantum-vulnerable public-key algorithms after 2030 and disallows them after 2035.⁠[10] We inventory the cryptography in your AI systems in the same engagement: G3 Post-quantum readiness for AI systems.

The scopes

Ten core scopes on the three sides, and twelve deeper scopes for AI agents. Each has a page with its deliverables and its framework mapping.

Architect
Assure
Govern

AI agents

All scopes and deliverablesPricing

Not yet offered

In development

Synthetic-actor defence (in development)

AI agents and other synthetic actors now operate inside company systems, often unseen.

We are building a security service for this problem, first for Swomi’s own security operations.

Status: in development. No opening date is set, and we will not describe what it does until it does it.

Tell me when it opens

The TriOngle Method v1.0

Six stages, the same on every engagement, published so you can hold us to them.

  1. Scope

    The system, its boundary, the frameworks in scope and the rules of engagement, agreed in writing.

  2. Model

    Data flows, trust boundaries, assets and actors, including every agent and tool.

  3. Test

    Design review and attack testing, as the scope sets, each test case tagged with its technique.

  4. Map

    Every finding mapped to the frameworks and the law it touches.

  5. Report

    One plain-language report, for engineers, the board and counsel.

  6. Retest and hand over

    Fixes verified, and the tests handed over to run in your CI.

Read the method

Founded by Maynard L. Dokken

TriOngle’s founder is a Canadian technology entrepreneur and inventor, an early developer of the application service provider model and of privacy-first online payment technology.⁠[11] His security record includes a 2000 filing at Milinx for threshold biometric authentication that combines voice and fingerprint verification with encryption,⁠[12] a 2009 program that gave ordinary users a free security and privacy USB key,⁠[13] and the 2023–24 filings for a Bio-Origin Validation Engine, WO 2025/000099.⁠[14]

About TriOngle

Start with a scope

Tell us the system and the scope code, or the problem if you do not know the scope yet.

inquiries@triongle.com

Security reports go to security@triongle.com under our disclosure policy. Test evidence is kept on EU servers only.

References

Numbered in the order they are first cited on this page; each entry links back to the places that cite it. Every source was opened on 4 November 2026.

  1. ↑ 2025 Top 10 Risk & Mitigations for LLMs and Gen AI Apps. OWASP Gen AI Security Project.
  2. ↑ a b OWASP Top 10 for Agentic Applications for 2026. OWASP Gen AI Security Project, 9 December 2025.
  3. ↑ mitre-atlas/atlas-data. MITRE ATLAS on GitHub. Data version 2026.05.
  4. ↑ Artificial Intelligence Risk Management Framework: Generative Artificial Intelligence Profile (NIST AI 600-1). NIST, 26 July 2024.
  5. ↑ ISO 42001 Annex A Controls List: All 38 Controls Explained. RiskProfs, 21 July 2026.
  6. ↑ New guidance to help organizations protect AI from security threats. BSI, 27 October 2026.
  7. ↑ EU Digital Omnibus on AI Enters Into Force. K&L Gates, The National Law Review, 31 July 2026.
  8. ↑ Security Best Practices. Model Context Protocol.
  9. ↑ NIST IR 8547, Transition to Post-Quantum Cryptography Standards. NIST CSRC, 27 October 2026.
  10. ↑ NIST Releases Final IR 8547 Establishing Definitive Post-Quantum Cryptography Transition Timeline and Deprecation Schedules. Quantum Computing Report, 27 October 2026.
  11. ↑ Maynard L. Dokken. maynardldokken.com.
  12. ↑ Milinx Business Group: Patents and research. maynardldokken.com.
  13. ↑ CapitalKeez: The Start & Journey of Keez & Maynard L. Dokken. capitalkeez.com.
  14. ↑ Maynard L. Dokken: Patents and innovations. maynardldokken.com.