Emerging AI Standards: What They Mean for AI Governance

How ISO/IEC 42001, ISO/IEC 27090 and other international AI Standards are helping organisations manage Artificial Intelligence responsibly.

Artificial Intelligence is transforming the way organisations operate, from automated customer services and business analytics to generative AI, software development and decision-making systems.  

However, as AI adoption accelerates, so do the risks associated with its use and organisations must consider many aspects of compliance including data protection, cybersecurity, intellectual property, accuracy, transparency, bias and accountability. Customers are asking for more assurance that AI is being used responsibly, while governments and regulators are introducing new legal requirements.

A growing family of international AI Standards provides organisations with recognised frameworks for addressing these challenges.

In October 2026, Robert Clements form Assent Risk Management participated in a webinar organised by the European Union of Associations of Translation Companies (EUATC) and ISO/TC 37/SC 5, exploring emerging ISO Standards for AI and their implications for business.

Although the webinar focused on language services, the standards discussed have much wider applications across sectors including technology, financial services, healthcare, manufacturing, professional services and the public sector.

Since the webinar, the publication of ISO/IEC 27090:2026, a new international standard addressing cybersecurity threats to AI systems, has added another important dimension to the evolving AI Standards landscape.

So, which AI Standards should organisations be considering, how do they relate to legislation such as the EU AI Act, and where should businesses begin when implementing effective AI Governance?

Why AI Governance is becoming a business priority

AI Governance refers to the policies, responsibilities, processes and controls through which an organisation directs, monitors and manages the development, procurement and use of Artificial Intelligence.

It ensures that AI systems are used responsibly, risks are appropriately managed and decisions remain accountable.

Even organisations that do not develop AI themselves may have significant governance responsibilities.

The development, integration and use of AI tools can create risks associated with confidential information, intellectual property, inaccurate outputs and non-compliance with contractual requirements.

Effective AI Governance enables organisations to understand where AI is being used, identify associated risks, establish appropriate controls and demonstrate accountability to customers, regulators and other stakeholders.

Importantly, AI Governance is not simply an IT responsibility. It requires involvement from many parts of the organisation throughout the Lifecycle of the AI system.

ISO/IEC 42001: The foundation for AI Governance

ISO/IEC 42001:2023 is the international standard specifying requirements for an Artificial Intelligence Management System (AIMS).

Like ISO 9001 for quality management and ISO 27001 for information security, ISO 42001 provides a structured management system framework that organisations can implement, maintain, continually improve and have independently certified.

The standard is applicable to organisations developing, providing or using AI systems, regardless of their size or sector.

One of the greatest benefits of ISO/IEC 42001 is that it encourages organisations to identify the AI systems they use and understand their responsibilities in relation to those systems.

Many businesses have adopted AI organically, with employees using applications embedded within Microsoft 365, Google Workspace, customer relationship management systems, design platforms and other business software.

Without appropriate oversight, organisations may have limited visibility of the technologies being used or the information being processed.

ISO/IEC 42001 provides a recognised framework for bringing these activities under organisational control.

It establishes requirements covering leadership and accountability, AI policies, risk and impact assessments, operational controls, performance monitoring, internal auditing and continual improvement.

Importantly, it is not primarily a technical standard for AI developers. It is a management system standard that helps organisations govern AI activities in a structured and consistent way.

For organisations already certified to ISO 9001 or ISO/IEC 27001, the familiar harmonised management system structure can make integration more straightforward.

ISO/IEC 42001 is therefore a logical starting point for organisations wanting to establish effective AI Governance, demonstrate responsible AI management and prepare for independent certification.

Assent Risk Management provides specialist ISO 42001 consultancy and implementation support, helping organisations develop management systems appropriate to their activities, risks and business objectives.

ISO/IEC 27090:2026 – A new standard for AI cybersecurity

One of the latest additions to the international AI Standards landscape is ISO/IEC 27090:2026, Cybersecurity – Artificial intelligence (AI) – Addressing security threats and compromises to AI systems.

Published on 14 October 2026, the standard provides guidance on understanding, detecting and mitigating cybersecurity threats that specifically affect AI systems.

While conventional cybersecurity standards address issues such as unauthorised access, malware and data breaches, AI introduces additional vulnerabilities associated with training data, models, prompts and automated decision-making.

ISO/IEC 27090 helps organisations understand these threats, their potential consequences and the measures that may be used to address them throughout the AI system lifecycle.

How does ISO/IEC 27090 relate to ISO 27001 and ISO 42001?

The three standards address different but complementary aspects of organisational risk management.

StandardPrimary purpose
ISO/IEC 42001Establishes requirements for an AI Management System, covering organisational governance, risks, impacts and continual improvement.
ISO/IEC 27001Establishes requirements for an Information Security Management System, covering the protection of information and associated risks.
ISO/IEC 27090Provides guidance on cybersecurity threats and compromises specific to AI systems, including detection and mitigation measures.

For organisations already operating an ISO 27001 Information Security Management System, ISO 27090 offers an opportunity to strengthen existing security risk assessments and controls by considering AI-specific threats.

Similarly, organisations implementing ISO 42001 can use ISO 27090 to inform their approach to AI-related cybersecurity risks.

However, it’s important to note that ISO/IEC 27090 is a guidance standard rather than a standalone management system that can be certified.

The value of ISO 27090 is in helping organisations understand emerging AI security threats and incorporate appropriate protections into their existing governance and cybersecurity arrangements and building that understanding into their integrated management system..

AI risk management: ISO/IEC 23894

ISO/IEC 23894:2023 provides guidance on managing risks associated with Artificial Intelligence.

It complements ISO/IEC 42001 by offering a more detailed approach to identifying, analysing, evaluating and treating AI-related risks (clause 6.1 of the AIMS standard).

AI risk management needs to consider the potential impacts on individuals, groups and society.  

For example, an AI system used to support recruitment decisions could introduce discriminatory outcomes, while an AI-powered financial assessment tool might produce inaccurate or unfair recommendations.

An AI system used within healthcare could have even more serious consequences if its outputs were unreliable or inappropriate.

As technologies evolve, new risks emerge and previously acceptable controls may become ineffective.

Organisations should therefore regularly review their AI risk assessments, particularly when introducing new systems, changing suppliers or expanding the use of existing technologies.

The MIT AI Risk Repository provides a useful source of information for organisations seeking to identify potential AI risks.

However, any generic risk registers should be adapted to reflect the organisation’s actual activities, systems and operating environment or context in ISO language.

AI impact assessments: ISO/IEC 42005

Another important standard is ISO/IEC 42005:2025, which provides guidance on conducting AI system impact assessments.

While the AI risk assessment considers uncertainty and its consequences, impact assessments help organisations evaluate how AI systems may affect individuals, groups and society.

This distinction is particularly relevant where AI is used to support decisions affecting people.

For example, an organisation introducing AI into recruitment, employee monitoring, credit assessment or customer profiling should consider whether the technology could create unintended consequences or disproportionately affect particular groups.

An AI impact assessment can help identify affected interested partners, evaluate potential harms and benefits, and establish measures to reduce negative impacts.

ISO/IEC 42005 complements the governance requirements of ISO/IEC 42001 and provides a useful framework for documenting these assessments.

For organisations operating internationally, AI impact assessments can also support wider regulatory compliance activities, although they do not automatically satisfy every legal requirement.

AI system lifecycle management: ISO/IEC 5338

ISO/IEC 5338:2023 addresses processes throughout the AI system lifecycle, from initial planning and development through to operation, maintenance and retirement.

During the EUATC webinar, Robert highlighted a frequently overlooked aspect of AI Governance: the end of an AI system’s operational life.

Organisations often focus on selecting and implementing new technologies but give less consideration to what happens when those technologies change or are withdrawn.

For example, what happens if an AI supplier changes its underlying model and the outputs no longer meet the organisation’s expectations?

Can the organisation identify the change, assess its consequences and restore an earlier configuration where necessary?

If an organisation decides to migrate to another AI platform, how will it transfer information, preserve appropriate records and ensure that data held by the previous supplier is securely managed?

These questions demonstrate why lifecycle management should begin before an AI system is introduced.

Organisations should consider how systems will be maintained, monitored, updated, replaced and ultimately decommissioned.

ISO/IEC 5338 provides guidance that can help organisations establish a more systematic approach to these activities.

AI bias, transparency and explainability

AI systems can produce biased, misleading or difficult-to-explain outputs.

These issues may arise from training data, model design, system configuration or the context in which the technology is used.

For organisations relying on AI to support decisions, understanding and managing these risks is essential.

ISO/IEC TR 24027 provides guidance on bias in AI systems, while ISO/IEC TS 12791 addresses the treatment of unwanted bias in certain machine learning tasks.

Other standards, including ISO/IEC TS 6254 and ISO/IEC 12792, address explainability, interpretability and transparency.

These standards help organisations consider whether AI outputs can be appropriately understood, evaluated and challenged.

However, businesses should recognise that some existing AI Standards were developed around particular machine learning techniques and may not fully address every characteristic of modern generative AI.

During the webinar, Adam Leon Smith, Chair of the AIQI Consortium and Deputy Chair of the UK’s national AI standards committee,  highlighted the ongoing work to develop more comprehensive approaches to assessing bias and evaluating generative AI systems.

For organisations, the practical lesson is that AI outputs should not automatically be treated as reliable simply because they are generated by sophisticated technology.

Appropriate validation, monitoring and human oversight remain essential.

Data quality: The ISO/IEC 5259 series

The ISO/IEC 5259 series addresses data quality for analytics and machine learning.

Data quality is fundamental to the performance and reliability of AI systems.

A common misconception is that larger datasets necessarily produce better results.

In reality, data must be relevant, representative, sufficiently complete, appropriately current and suitable for its intended purpose.

For example, an AI model trained on outdated or unrepresentative information may produce inaccurate predictions or discriminatory outcomes.

Similarly, an organisation using AI to analyse customer feedback or business performance data must consider whether the underlying information is sufficiently reliable to support meaningful conclusions.

During the webinar, the panel highlighted the importance of not focusing exclusively on training data.

Testing datasets and the information processed during ongoing operation are also critical to assessing AI system performance.

The ISO/IEC 5259 series provides guidance on data quality characteristics, measurement and management, helping organisations establish more systematic approaches to these issues.

For businesses developing or customising AI models, these principles can be particularly valuable.

Emerging AI Standards under development

As of October 2026.  

International standardisation continues to evolve alongside AI technology.

During the EUATC webinar, several developing standards were highlighted as potentially significant for organisations using or developing AI.

These included projects within the ISO/IEC 42119 series, addressing aspects of verification, validation, red teaming and the quality assessment of AI systems.

Other developing standards, including ISO 23281, ISO 23282 and ISO 26320, focus on natural language processing and the evaluation and management of language-related technologies.

Although some of these projects are particularly relevant to software developers and AI providers, they may also influence future expectations for organisations procuring or integrating AI services.

It is important to distinguish between published International Standards and projects still under development.

Draft standards may change before publication, and their requirements should not be treated as established obligations.

Nevertheless, organisations investing significantly in AI should consider monitoring relevant standards development so that future governance, assurance and procurement requirements can be anticipated.

AI Standards and the EU AI Act: Understanding the relationship

The European Union’s Artificial Intelligence Act introduces legal obligations for organisations developing, supplying and using AI systems.

Although ISO Standards provide recognised frameworks for AI Governance, implementing or achieving certification to ISO 42001 does not automatically demonstrate compliance with the EU AI Act.

The legislation establishes obligations according to the nature of the AI system, its intended purpose and the role of the organisation concerned.

For example, a business using a third-party AI application to assist with administrative tasks may have different obligations from an organisation developing an AI system used for recruitment, credit assessment or other regulated activities.

The EU AI Act also introduces transparency obligations for certain AI-generated content and requirements relating to the governance, documentation and monitoring of particular AI systems.

For organisations operating internationally, the legislation may be relevant even where the business is established outside the European Union.

ISO/IEC 42001 can support compliance by providing a structured approach to identifying legal requirements, allocating responsibilities, managing risks and maintaining evidence of governance activities.

Other standards, including ISO/IEC 23894, ISO/IEC 42005 and ISO/IEC 27090, can provide more detailed guidance on particular aspects of risk, impact assessment and cybersecurity.

However, organisations must separately determine which legal obligations apply to their activities and whether additional controls are necessary.

Frequently asked questions about AI Governance and AI Standards

What is the main ISO Standard for AI Governance?

ISO/IEC 42001:2023 is the principal international management system standard for AI Governance. It specifies requirements for establishing, implementing, maintaining and continually improving an Artificial Intelligence Management System.

Other standards provide complementary guidance on AI risk management, impact assessment, cybersecurity, data quality and technical assurance.

Is ISO/IEC 42001 certification mandatory?

No. ISO/IEC 42001 certification is generally voluntary, although customers or contractual arrangements may require it.

Organisations can implement the standard without pursuing independent certification.

Certification can nevertheless provide additional assurance to customers and stakeholders that an organisation’s AI Management System has been independently assessed.

What is the difference between ISO/IEC 42001 and ISO/IEC 27090?

ISO/IEC 42001 specifies requirements for managing AI activities across an organisation.

ISO/IEC 27090 provides guidance on identifying, detecting and mitigating cybersecurity threats specific to AI systems.

The standards are complementary. ISO/IEC 42001 establishes the overarching management framework, while ISO/IEC 27090 provides more detailed information to support AI cybersecurity risk management.

Does ISO/IEC 27090 replace ISO/IEC 27001?

No. ISO/IEC 27090 complements ISO/IEC 27001 and ISO/IEC 27002.

ISO/IEC 27001 establishes requirements for an Information Security Management System, while ISO/IEC 27090 addresses security threats and compromises specific to AI systems.

Organisations should continue to apply established information security practices while considering whether additional AI-specific protections are necessary.

Can organisations be certified to ISO/IEC 27090?

ISO/IEC 27090 is a guidance standard, rather than a standalone management system certification standard.

Organisations can use it to inform their cybersecurity arrangements, but it should not be confused with certifiable standards such as ISO/IEC 27001 and ISO/IEC 42001.

Do organisations need AI Governance if they only use third-party AI tools?

Yes, proportionate AI Governance remains important even where an organisation does not develop AI systems itself.

Third-party AI applications can introduce risks involving confidential information, data protection, intellectual property, inaccurate outputs and supplier dependency.

Organisations should understand how these systems are used, what information they process and what responsibilities remain with the organisation.

Can ISO/IEC 42001 be integrated with existing ISO certifications?

Yes. ISO/IEC 42001 uses the harmonised management system structure found in standards such as ISO 9001 and ISO/IEC 27001.

This allows organisations to integrate common processes, including internal auditing, management review, competence, document control and continual improvement.

AI-specific requirements must still be addressed.

Who should be responsible for AI Governance?

Senior leadership retains overall accountability for organisational governance.

Operational responsibilities may be assigned to an AI Governance Manager, compliance manager, information security specialist, risk management function or multidisciplinary committee.

Where organisations lack the necessary internal expertise, an outsourced AI Governance service can provide specialist support while retaining accountability within the organisation.

How Assent Risk Management can help

At Assent Risk Management, we believe that AI Governance should enable responsible innovation rather than create unnecessary administrative burdens.

Our consultants work with organisations across a wide range of sectors to develop practical management systems that address their specific activities, risks and business objectives.

We provide specialist ISO 42001 consultancy, helping organisations establish Artificial Intelligence Management Systems, integrate AI Governance with existing ISO certifications and prepare for impartial third-party certification.

Our services include gap analysis, AI risk and impact assessments, management system development, internal auditing, training and certification readiness.

For organisations already certified to ISO 27001, we can help review how emerging AI cybersecurity risks should be incorporated into existing information security arrangements, drawing on guidance such as ISO/IEC 27090.

We also provide an Outsourced AI Governance service and a Fractional AI Governance service, enabling businesses to access experienced specialists without employing a dedicated full-time AI Governance Manager.

This can include maintaining AI risk registers, assessing new technologies, supporting governance committees, reviewing legal and contractual requirements and helping organisations respond to AI-related concerns.

Whether your organisation is beginning its AI Governance journey, implementing ISO/IEC 42001 or reviewing the security of existing AI systems, Assent can provide a bespoke solution.

For more information contact our expert team!

____________________________________________

AI Transparency: This blog was written with the help of Chat GPT AI from a transcript of the live, human webinar discussion at the EU ATC Webinar “Emerging ISO Standards for AI and What They Mean for Language Services” On 13th October 2026.

ARMAdmin
ARMAdmin
Articles: 127