Independent Assessment

ISO 27090 AI Security Assessment

An independent assessment of your AI systems against the security guidance in ISO/IEC 27090:2026, done by assessors from outside your team. Most assessments take two to four weeks and end in a formal report with every finding rated by risk.

  • Every finding tied to an ISO 27090 threat and backed by evidence
  • Covers the AI systems you build and the AI services you buy
  • Led by a senior CISSP in the US or Canada

For most SMBs, a standalone ISO 27090 AI security assessment costs USD $5,000 to $15,000.

Talk to an AI security assessor

Share the AI systems in scope and your timeline, and an assessor will follow up to book a call.

What does an ISO 27090 AI security assessment cover?

A risk-based assessment of your AI systems against ISO/IEC 27090:2026, from scoping to the final report.

Frame (8)

Assess: Scoping and AI System Inventory

  • What We Do

    We confirm the AI systems in scope, including third-party models and AI services, and review your architecture, data flows, suppliers and existing risk assessment. We build an assessment plan weighted by risk and schedule interviews.

  • What You Get

  • Assessment Plan:
    Scope, criteria and method, weighted by the AI systems and data that carry the most risk.

  • AI System Inventory:
    Each system in scope with its owner, data sources, model source and suppliers.

Frame (9)

Assess: Threat Applicability Review

  • What We Do

    We review each AI system against the threats ISO/IEC 27090:2026 describes, such as data poisoning, model exfiltration, evasion, membership inference, model inversion and prompt injection, and record which ones apply.

  • What You Get

  • Threat Applicability Matrix:
    Each ISO 27090 threat marked applicable or out of scope for each system, with the reason.

Frame (10)

Assess: Document and Design Review

  • What We Do

    We review AI security policies, risk treatment, supplier agreements, model update procedures and incident response plans against the ISO 27090 guidance and your ISO 27001 or ISO 42001 commitments.

  • What You Get

  • Document Review Findings:
    Each document checked for coverage, currency and approval, and for whether it matches how your team works.

Frame (11)

Assess: Safeguard Testing

  • What We Do

    We sample evidence that safeguards work, such as training data provenance records, access and usage logs for model endpoints, input and output filtering, development environment controls and model change records.

  • What You Get

  • Safeguard Testing Workpapers:
    For each safeguard, the test performed, the evidence reviewed and the conclusion.

Frame (12)

Assess: Interviews and Process Verification

  • What We Do

    We interview engineering, data and security staff to confirm they know their responsibilities for AI systems and can show them, from data intake to model retirement.

  • What You Get

  • Interview Notes:
    What was asked, what was shown, and where practice differs from the documented procedure.

Frame (13)

Assess: Assessment Report and Findings Walkthrough

  • What We Do

    We write a formal report with findings rated by risk and recommended safeguards, walk your team through every finding, and format the results for your management review.

  • What You Get

  • AI Security Assessment Report:
    Every finding rated, tied to an ISO 27090 threat or topic, and backed by evidence.

  • Remediation Plan:
    Fixes ordered by risk, each with an owner and a target date.

What do you get from an ISO 27090 AI security assessment?

Four things you hold at the end, ready for customers and leadership.

An AI security assessment report

A formal report with scope, criteria, method and every finding tied to an ISO 27090 threat or topic.

A threat applicability matrix

Each ISO 27090 threat marked against each AI system, ready for customer security reviews.

A dated remediation plan

Fixes ordered by risk, each with an owner and a target date.

Management review input

Results formatted for your ISO 27001 or ISO 42001 management review, so leadership signs off on what was found and what changes.

When is an independent ISO 27090 assessment the right call?

For many companies, the trigger is a customer, investor or board that asks for an outside view of how AI systems are secured.

Right call when

  • Enterprise customers ask how you protect your AI features, models and training data

  • You want an outside view before adding AI systems to your ISO 27001 or ISO 42001 scope

  • The engineers who built your AI safeguards are the engineers who operate them

  • You use third-party models or AI services and want their risk assessed

Wrong call when

  • You need the lowest possible price. Our pricing is competitive, but every assessment is led by experienced assessors, goes through several rounds of QA and ends in a detailed report, so we are rarely the cheapest option

What do clients say about Truvo's ISO work?

"Truvo Cyber conducted our ISO 27001 and ISO 27701 internal audit. They were easy to work with, communicated clearly throughout the project, and delivered a thorough, professional report on time. I would recommend them to any organization working through similar compliance requirements."

Ellipse (1) (1)
Aimee Halbert

UserTesting

"Truvo is an instrumental and integrated part of our team...
They don’t just provide recommendations; they ensure we meet our stringent ISO 27001 and SWIFT compliance goals. We trust them with projects of national importance, and they deliver."

Mask Group-1
Matt Charette

CISO, Payments Canada

Ready to plan your ISO 27090 AI security assessment?

Book a strategy call. We review your AI systems and your timeline, and give you a clear picture of the scope and investment.

Book Your Free ISO Strategy Session

What do companies ask about ISO 27090 assessments?

For most SMBs, a standalone ISO 27090 AI security assessment costs USD $5,000 to $15,000. The price depends on the number of AI systems in scope, whether you build models or use third-party AI services, and the depth of safeguard testing. No prior Truvo engagement is required.

ISO/IEC 27090:2026 is an international standard from ISO/IEC JTC 1/SC 27, published in October 2026. It gives guidance on understanding, detecting and mitigating security threats specific to AI systems throughout their life cycle, from training data to models, including threats such as data poisoning and model theft.

ISO describes ISO 27090 as security guidance with no requirements for a management system, so an assessment against it gives you an independent, evidence-based view of your AI security. Certification remains available for the management systems around it, such as ISO 27001 and ISO 42001.

We assess each AI system in scope against the threats and safeguards ISO 27090 describes, from training data and model access to prompt injection, model updates and monitoring. We review documents, interview staff and sample evidence that safeguards work.

The assessment reviews your AI security program and safeguards against ISO 27090 using documents, interviews and evidence. AI penetration testing attacks a system directly to find exploitable weaknesses. Many companies start with the assessment and add penetration testing for their highest-risk AI systems.

ISO/IEC 42001 sets requirements for an AI management system, and ISO 27090 focuses on security threats to AI systems. The assessment findings can feed your ISO 42001 AI risk assessment and management review, and your ISO 27001 risk treatment plan.

Two to four weeks for most small and mid-sized organizations, from scoping to final report. A scope with many AI systems or AI suppliers takes longer. We give you a dated plan after the scoping call.

Yes, as a separate engagement with a different team, so the assessors stay independent of the program. See our ISO 27090 AI security consulting page.

How is a Truvo ISO 27090 assessment different?

Our assessors build and operate management systems, so findings come with fixes your team can carry out.

The Self-Assessment Approach

  • Document-Only Reviews:
    Many AI security reviews confirm that policies exist. Customers ask whether your team follows them.

  • Independence Gap:
    The engineers who built the AI safeguards are often the only people who know them well.

  • Generic Findings:
    "Improve AI governance" gives your team no clear next step.

The Truvo Approach

  • Practitioner-Led Assessment:
    Our assessors build and operate management systems, and every engagement is led by a senior CISSP in the US or Canada.

  • Threat-by-Threat Coverage:
    We review each AI system against the threats ISO 27090 describes, and record the evidence and conclusion for each.

  • Actionable Findings:
    Every finding is rated by risk, tied to an ISO 27090 threat or topic, and paired with a recommended safeguard.

Ready to assess your AI security?

Independent ISO 27090 AI security assessments by practitioners who build and operate management systems.

Book a Strategy Call

From the blog: AI security and governance

Researchers examining a robotic arm, showcasing technology and innovation, illustrating Top ISO 42001 Consultants in Canada (2026): A Buyer's Guide

Top ISO 42001 Consultants in Canada (2026): A Buyer's Guide

Truvo Cyber is a Canadian cybersecurity firm that implements ISO/IEC 42001, the AI management system standard, for SaaS and technology companies ...

Learn More
ISO 42001 Cost in 2026: The 4 Factors

ISO 42001 Cost in 2026: The 4 Factors

ISO 42001 implementation and certification for small organization can land anywhere between roughly US$20,000 and US$55,000 for a first ...

Learn More
An animated infographic titled "AI Governance Crosswalk." A Venn diagram connects Risk Management (ISO 42001), Management System (NIST AI RMF), and Legal Compliance (EU AI ACT) to a central "Shared Core." A man points, and text at the bottom reads "Build Once, Comply With All Three."

ISO 42001, NIST AI RMF, and the EU AI Act: The Complete Control Crosswalk

ISO 42001, the NIST AI RMF, and the EU AI Act overlap on roughly two-thirds of their controls. Design one control set against that shared core and ...

Learn More