July 16, 2026

How ISO 42001 strengthens third-party AI compliance

Third-party artificial intelligence (AI) tools can introduce several risks to your organisation. That means it’s an area where you may suddenly find yourself non-compliant with regulatory requirements or legislation.

You might not build the AI model, own the code, or even fully understand how it works. But if that AI influences your decisions, services or customer outcomes, the risk still sits with you.

What you need, then, is a framework for managing and controlling third-party AI risks. That’s where ISO 42001 comes in. It provides a structured way to manage AI across your organisation, including systems supplied by external vendors.

ISO QSL is a team of ISO consultants. We partner with businesses around the UK to help them develop and implement ISO-compliant management systems as well as certificating them. Here’s how we recommend approaching external AI in your artificial intelligence management system (AIMS) under the new ISO 42001 standard.

Are you prepared for ISO 42001?

We’re here to help you review how your organisation uses AI, identify gaps against the standard and prepare for ISO 42001 certification.

Book a discovery call
Jodie Turner, Marketing Team Leader at ISO QSL

About the author

Jodie Turner – Marketing Team Leader

During my time as ISO QSL, I’ve developed extensive knowledge of digital marketing alongside a strong understanding of the ISO standards that help organisations improve.

Why third-party AI is now a compliance risk

AI is taking the world by storm. But your organisation most likely isn’t developing its own model from scratch. Instead, you’re more likely relying on one or more third-party AI systems as the foundation of your solution, or even using them end to end. Even if you’re an AI developer, you still likely rely on external AI tools and systems, and integrating them into your existing workflows. These could include:

  • Customer-facing AI chatbots
  • Recruitment screening tools
  • Fraud detection systems
  • SaaS platforms with embedded AI features
  • Internal tech support chatbots

This introduces a challenge, somewhat similar to the GDPR-related requirements under ISO 27001, namely that you inherit the outcomes of these systems without having full visibility into how they operate. That creates a significant gap between use and control. External AI may introduce biased outputs, inconsistent performance or unusual decision-making, and because your system is outsourced, there may be little you can directly do to reduce the risks.

Despite that, you’re still accountable for those risks, and expected to manage them.

This is already reflected in legislation such as the EU AI Act, which shares responsibility across organisations involved in developing and using AI. Similar regulations are likely to follow in other regions.

So, what can you do? This is where ISO 42001 becomes invaluable.

How ISO 42001 approaches third-party AI systems

ISO 42001 applies to your organisation if you develop, provide or use AI. That’s a wide scope, and it explicitly includes externally-provided systems.

Rather than treating third-party AI as a procurement issue, the standard brings it into your AI management system. This means you must identify, assess, govern and monitor your external AI in the same structured way as your internal AI.

In ISO language, ‘control’ and ‘influence’ refer to what your organisation can directly manage versus what it can shape indirectly. Where direct control is limited, as is often the case with third-party AI, ISO 42001 expects you to manage risk through influence, using supplier selection, better defined requirements and ongoing oversight.

Importantly, using third-party AI doesn’t reduce your responsibilities. Your organisation remains accountable for how it’s selected, implemented, used, monitored and reviewed over time.

What you need to implement

So, what does complying with ISO 42001 look like in practice? At its core, it comes down to building an AI management system with the right components in place from the start:

AI inventory

The first step is visibility. You need a clear inventory of where your organisation uses AI, including third-party systems. This should capture what each system does, the role it plays, and which processes or decisions it affects.

Risk assessment

From there, move on to the risk assessment. An ISO 42001-specific risk assessment goes beyond traditional IT or security checks (although it’s certainly related in many ways to ISO 27001 risks). You need to assess how your AI behaves in your specific context, including potential impacts on individuals, decision quality and operational outcomes. You’ll specifically consider bias, over-reliance, lack of explainability, misuse, and so on.

AI supplier due diligence

After that, carry out due diligence on your AI supplier. You don’t need to interrogate your provider or learn the ins and outs of AI development. However, you should understand your model’s intended use, limitations, performance expectations and how they manage changes and updates. You need enough information to make informed decisions about whether and how you should use the system.

Develop governance

With the previous steps in place, it’s time to develop your governance. That means defining how the AI should be used within your organisation, where you require human oversight and what controls apply to its outputs. Embed these controls into your workflows.

Training

With governance defined, you need to ensure your employees understand how to use AI systems correctly in practice. Your controls are only effective if your staff consistently apply them. Training should reflect the level of risk and ensure all individuals understand what the AI is designed to do, its limitations, and when human judgement must take priority, especially for third-party systems, where reliability is more likely to be assumed.

Monitoring

AI systems are designed to adapt and evolve. When they’re controlled by external providers, there may be little you can do to influence how they change. Outputs can shift, features can change and performance can drift over time. Implement an ongoing review process to ensure that your third-party AI risks remain understood and controlled rather than assumed.

Change management

When a supplier updates an AI system, that update may alter the risks to your organisation. Thus, you need a structured process for reviewing and approving changes to ensure that you maintain control as the system evolves.

What auditors will expect to see

During your audit, auditors will scrutinise your third-party AI systems. Relying on supplier assurances and policies isn’t enough. Auditors will expect to see that your organisation has identified where external AI is used, assessed the associated risks, defined the appropriate controls and implemented monitoring.

You must demonstrate this through evidence. That typically includes an AI system inventory, documented risk assessments, supplier evaluations, records of decisions and controls and evidence of ongoing monitoring and review.

The key point auditors will focus on is, can you demonstrate and prove that your governance is followed in your actual operations?

How ISO 42001 supports wider compliance

ISO 42001 is a governance framework. It provides the structure you need to align with legal or regulatory requirements, but it doesn’t replace them.

Embedding your risk assessment, lifecycle controls and monitoring into a formal management system makes it much easier to demonstrate accountability. This is particularly valuable where you rely on third-party AI. While you can’t control all the risks here, you can demonstrate due diligence and compliance through consistent, documented, evidenced, and repeatable governance.

Strengthening your third-party AI compliance with ISO 42001 isn’t just about policies, although those are, of course, fundamental. You need a working system that identifies risk, applies the right controls, and stands up to audit.

That’s where ISO QSL can help. Our ISO consultants work with you to build and implement an ISO 42001-compliant AI management system that reflects how your organisation uses, or plans to use, AI, including third-party tools and suppliers. So if you’re relying on external AI and need to demonstrate control, now is the time to act. Get in touch with ISO QSL today to get started.