How ISO 17100, ISO 18587 and emerging international AI standards can help translation companies manage disruption, maintain quality and build customer confidence.
Artificial intelligence is transforming the global language services industry. From machine translation and automated quality estimation to generative AI and large language models (LLMs), technologies are changing how translation, localisation and other multilingual services are delivered.
For language service providers (LSPs), this transformation presents both opportunities and challenges. AI offers the potential to improve productivity, expand service offerings and support multilingual communication at unprecedented scale. However, it also introduces concerns about translation quality, confidentiality, data protection, accountability and the continuing role of professional linguists.
At a time when traditional business models are being challenged, international standards provide an important foundation for managing change, demonstrating competence and maintaining customer trust.
On 12 October 2026, the European Union of Associations of Translation Companies (EUATC), in collaboration with ISO/TC 37/SC 5, hosted a knowledge-sharing webinar entitled Emerging ISO Standards for AI and What They Mean for Language Services. See the blog here.
The panel included Carina Balbo from ATC, Robert Clements, adviser to ATC Certification, and Adam Leon Smith, alongside representatives of the EUATC and the international standards community.
The discussion explored the published and developing ISO standards for artificial intelligence, their relationship with the EU AI Act, and their implications for language service providers.
One message was particularly relevant: language companies do not need to abandon established quality standards to embrace AI. Instead, they should build on their existing processes, adapting them to the risks and opportunities of new technologies.
AI disruption and the future of language services
The adoption of AI is changing how language services are commissioned, produced, reviewed and valued.
Traditional translation workflows have generally relied on professional linguists, supported by computer-assisted translation (CAT) tools, translation memories and terminology databases.
Machine translation has already introduced alternative delivery models, but generative AI is now accelerating that change through automated translation, summarisation, content creation, multilingual communication and quality assurance.
This is creating new commercial opportunities, but it’s also introducing uncertainty.
How can an LSP demonstrate that AI-generated translations meet a customer’s requirements? Who is responsible when an AI system produces an inaccurate or misleading translation? How can confidential client information be protected when third-party AI tools are involved?
There are also wider considerations concerning linguistic bias, intellectual property, traceability and the reliability of AI systems operating across different languages and cultural contexts.
These are not exclusively technical questions but also questions of organisational governance, quality management and professional responsibility.
This is where ISO standards and certification become particularly valuable.
ISO 17100 and ISO 18587: The foundations of translation quality
Before considering emerging AI standards, it is important to recognise the continuing relevance of established language industry ISO certification.
Two standards are particularly significant: ISO 17100 for translation services and ISO 18587 for machine translation post-editing.
ISO 17100:2015 – Requirements for Translation Services
ISO 17100:2015 establishes requirements for the processes, resources and competencies necessary to deliver professional translation services.
It covers important aspects of translation service delivery, including translator qualifications and competencies, project management, production processes, revision, customer requirements and quality assurance.
The standard provides a recognised framework through which translation service providers can demonstrate that their processes and resources are capable of delivering services that meet agreed specifications.
Importantly, the current ISO 17100:2015 standard excludes raw machine translation output combined with post-editing from its scope.
This distinction matters as language companies introduce new AI-enabled workflows.
An organisation certified to ISO 17100 should ensure that the services it claims are covered by certification genuinely fall within the scope of the standard. The use of AI tools elsewhere in the business does not automatically invalidate ISO 17100 certification, but machine translation post-editing must not simply be presented as an ISO 17100 translation process.
Nevertheless, many established ISO 17100 processes remain highly relevant to the wider management of AI-enabled services.
These include determining customer requirements, managing competent personnel, maintaining project records, controlling external resources and reviewing the quality of delivered work.
A revision of ISO 17100 is also under development, making it important for certified providers to monitor future changes.
ISO 18587:2017 – Post-Editing of Machine Translation Output
ISO 18587:2017 is particularly relevant to language companies adopting machine translation.
The standard specifies requirements for full human post-editing of machine translation output, including the competencies required of post-editors.
Rather than treating machine-generated content as automatically suitable for delivery, ISO 18587 establishes a structured process through which appropriately competent professionals review and improve machine translation output.
For LSPs, certification can demonstrate that machine translation post-editing is managed through defined and independently assessed processes.
This becomes increasingly valuable as customers seek reassurance that efficiency improvements have not come at the expense of quality.
The standard is also being revised. The proposed replacement, ISO/DIS 18587, is titled Translation services – Post-editing of non-human translation output – Requirements.
This broader terminology is significant because the industry is moving beyond traditional machine translation engines towards generative AI and other non-human translation technologies.
As of October 2026, the draft remains under development, and its final requirements may change before publication. Nevertheless, its direction illustrates how international standardisation is responding to technological developments.
For language companies, ISO 17100 and ISO 18587 provide complementary approaches to assuring quality across different translation production methods.
ISO/IEC 42001:2023 – Managing AI across the organisation
One of the most important standards discussed during the webinar was ISO/IEC 42001:2023, the international standard for artificial intelligence management systems.
ISO/IEC 42001 provides requirements for establishing, implementing, maintaining and continually improving an AI management system (AIMS).
Rather than concentrate on a particular technology or service, ISO/IEC 42001 takes an organisation-wide approach to AI governance, and is not specific to the language industry.
It helps businesses establish responsibilities, assess risks and impacts, introduce controls, monitor performance and continually improve how AI is developed, provided or used.
For language companies, this could encompass AI embedded within CAT tools, machine translation engines, automated quality assurance, terminology extraction, generative AI assistants and internally developed language technologies.
During the webinar, Robert highlighted that ISO/IEC 42001 is not primarily a technical standard. Its purpose is to provide a structured management framework.
This is particularly important for organisations that already operate management systems such as ISO 9001 or ISO 27001, where familiar approaches to policies, risk management, internal auditing and continual improvement can support implementation.
An important starting point is understanding the organisation’s relationship with each AI system.
A language company might be a user of AI technology supplied by another organisation, a provider of an AI-enabled service, or a developer of its own AI solutions. Its responsibilities and associated risks will differ accordingly.
ISO/IEC 42001 encourages organisations to examine these relationships systematically rather than applying the same controls indiscriminately.
For an LSP already certified to ISO 17100 or ISO 18587, an AI management system can provide an additional layer of governance without replacing existing translation quality processes.
ISO/IEC 23894:2023 – Understanding AI risks
AI introduces risks that may not be adequately addressed by traditional translation quality controls.
ISO/IEC 23894:2023 provides guidance on risk management relating to artificial intelligence.
For language service providers, relevant risks might include inaccurate translations, hallucinated content, linguistic or cultural bias, disclosure of confidential information, intellectual property infringement and excessive reliance on automated decision-making.
The consequences can vary considerably depending on the context.
An inaccurate translation of general marketing content may result in reputational damage. An error in medical instructions, legal documentation or safety information could have much more serious consequences.
Risk assessment should therefore consider the intended use of the translation, the sensitivity of the information, the reliability of the technology and the potential consequences of errors.
During the webinar, we emphasised the importance of considering risks not only to the organisation but also to individuals, groups and wider society.
This is especially relevant where language services support healthcare, public services, legal proceedings or communication with vulnerable communities.
ISO/IEC 23894 can help organisations adopt a proportionate, risk-based approach, ensuring that higher-risk applications receive appropriate levels of oversight and human intervention.
ISO/IEC 42005:2025 – Assessing the impact of AI systems
Another relevant standard where there is increased and high-risk AI use, is ISO/IEC 42005:2025, which provides guidance on AI system impact assessments.
While risk assessments examine uncertainty and potential consequences, impact assessments help organisations consider how AI systems may affect individuals, groups and society.
For language companies, this may involve considering whether automated translation systems perform consistently across languages, whether certain communities are disadvantaged by poor-quality output, or whether AI-generated content could create misleading or harmful communication.
Impact assessments can also support decisions about when human review is necessary and whether particular AI applications are appropriate for sensitive content.
Together, ISO/IEC 42001, ISO/IEC 23894 and ISO/IEC 42005 provide complementary approaches to AI governance, risk management and impact assessment.
ISO/IEC 5259: Data quality and multilingual content
Data quality is fundamental to reliable AI systems.
The ISO/IEC 5259 series addresses data quality for analytics and machine learning.
This is particularly relevant to language service providers working with translation memories, terminology databases, multilingual corpora and other language datasets.
These resources can be valuable inputs to AI systems, but their usefulness depends on more than the volume of available information.
During the webinar, the panel highlighted that data must be appropriate for its intended purpose.
For example, a large translation memory containing outdated terminology, inconsistent translations or material from unrelated subject areas may be less useful than a smaller, carefully maintained dataset.
Language companies should consider the accuracy, completeness, relevance, representativeness and provenance of data used in AI-enabled workflows.
The ISO/IEC 5259 series provides general data quality principles, although further standardisation is needed to address some of the specific characteristics of multilingual text corpora.
For LSPs, this reinforces the value of established terminology management, translation memory maintenance and quality assurance processes.
These activities are not becoming obsolete. They may become increasingly important as language data is reused within AI systems.
AI security and the new ISO/IEC 27090:2026
Information security is another important consideration for language companies adopting AI.
The recently published ISO/IEC 27090:2026 addresses cybersecurity threats and compromises affecting artificial intelligence systems.
Published in October 2026, the standard provides guidance on understanding, detecting and mitigating security threats specific to AI throughout its life cycle.
These include threats such as data poisoning, manipulation of AI systems and theft of AI models.
For language companies, the implications extend to protecting confidential customer content, evaluating AI suppliers, controlling access to sensitive datasets and understanding how information may be processed or retained by AI platforms.
ISO/IEC 27090 complements established information security standards such as ISO/IEC 27001.
It also highlights why organisations should not assume that traditional cybersecurity controls automatically address every risk introduced by AI.
AI bias, transparency and trustworthy language services
Language is closely connected with culture, identity and human communication. Consequently, unwanted bias in AI systems can be particularly significant for language service providers.
Several published ISO standards address aspects of AI trustworthiness, transparency and bias.
ISO/IEC TR 24027 examines bias in AI systems, while ISO/IEC TS 12791 addresses the treatment of unwanted bias in machine learning for classification and regression tasks.
ISO/IEC TS 6254 provides guidance relating to explainability and interpretability, and ISO/IEC 12792 establishes a taxonomy for AI system transparency.
These standards can support organisations in understanding and evaluating how AI systems behave, although not all their methods were developed specifically for modern generative AI or large language models.
During the webinar, the panel recognised that further work is needed to address the distinctive characteristics of generative AI.
For language companies, this reinforces the importance of appropriate human oversight, particularly where AI output could misrepresent meaning, introduce discriminatory language or fail to reflect the intended cultural context.
Trustworthy language services depend not only on technical performance but also on transparency about how services are delivered and how quality is assured.
Emerging ISO standards specifically relevant to language technology
In addition to published standards, ISO committees are developing new documents that could have a significant influence on the language industry.
Much of this work involves collaboration between ISO/IEC JTC 1/SC 42, which addresses artificial intelligence, and ISO/TC 37, which covers language and terminology.
Several projects discussed during the webinar are particularly relevant.
ISO 23281 and ISO 23282 – Natural language processing
These projects address AI tasks and functionalities associated with natural language processing (NLP), together with methods for evaluating NLP system accuracy.
They could provide greater consistency in how technologies such as machine translation, named entity recognition, sentiment analysis and automated quality estimation are evaluated.
For language companies, consistent evaluation methods may make it easier to compare systems and demonstrate their suitability for particular applications.
ISO/IEC 42119 series – AI system verification, validation and testing
The developing ISO/IEC 42119 series addresses aspects of testing and evaluating AI systems.
Projects discussed during the webinar included verification and validation analysis, red teaming and quality assessment of prompt-based text-to-text systems.
These developments are relevant to language companies using generative AI for translation, summarisation, terminology extraction and multilingual content production.
They could help organisations adopt more structured approaches to evaluating AI outputs rather than relying exclusively on supplier claims or informal testing.
ISO/IEC TS 26320 – Corpus development and maintenance
This developing project focuses on corpus development and maintenance for natural language processing systems.
Its relevance to the language industry is clear. Translation memories, multilingual datasets and terminology resources are important assets, and the quality of those assets can influence the performance of AI-enabled services.
Future guidance could help organisations adopt more consistent approaches to creating, maintaining and evaluating these resources.
These documents are at different stages of development. Their final content, publication dates and potential applications remain subject to the international standards development process.
Nevertheless, they demonstrate that the international standards community is actively responding to the changing needs of the language services sector.
What does the EU AI Act mean for language service providers?
Although ISO standards are voluntary, regulatory requirements are also influencing how organisations approach AI governance.
The European Union’s AI Act establishes a legal framework for artificial intelligence, with obligations depending on the type of AI system, its intended purpose and the organisation’s role.
For language companies operating internationally, the legislation may be relevant even when the organisation is established outside the European Union.
However, not every use of machine translation or generative AI is automatically classified as high-risk under the Act.
Language service providers should consider whether they are acting as providers or deployers of AI systems, what those systems are used for and whether particular legal obligations apply.
The AI Act entered into force in August 2024, with provisions becoming applicable in stages. Following subsequent amendments, important requirements for certain high-risk AI systems are scheduled to apply in December 2027 or August 2028, depending on the category.
ISO/IEC 42001 and related standards can help organisations establish processes relevant to regulatory compliance, including governance, risk management, documentation and monitoring.
However, ISO 42001 certification does not automatically demonstrate compliance with the EU AI Act.
Legal obligations must be assessed separately, taking account of the organisation’s activities and applicable jurisdiction.
For language companies serving international clients, adopting recognised ISO standards can nevertheless provide a consistent foundation for responding to different customer and regulatory expectations.
Why certification remains important in an AI-driven language industry
As AI technology becomes more widely available, language service providers may find it increasingly difficult to differentiate themselves through access to technology alone.
Many organisations can now access similar machine translation engines, generative AI platforms and automated language tools.
The distinction may increasingly lie in how those technologies are selected, controlled and integrated into reliable service delivery.
This is where independent certification can provide value.
Certification to ISO 17100 demonstrates conformity with recognised requirements for professional translation services.
Certification to ISO 18587 demonstrates conformity with recognised requirements for human post-editing of machine translation output.
ISO 42001 certification, where appropriate, can provide additional assurance that an organisation has established an AI management system.
These standards address different aspects of organisational and service performance. They should therefore be viewed as complementary rather than interchangeable.
For example, ISO/IEC 42001 does not establish that a particular translation meets the requirements of ISO 17100 or ISO 18587. Equally, certification to a translation service standard does not establish that every aspect of an organisation’s AI governance is adequately controlled.
The value comes from combining relevant frameworks to provide assurance over both the quality of language services and the governance of supporting technologies.
The role of independent, specialist certification
ATC Certification specialises in the independent auditing and certification of language service providers.
Unlike general certification bodies, ATC Certification works specifically with the language services industry, using auditors with specialist knowledge of translation processes, professional linguists and industry standards.
This expertise is particularly relevant when assessing organisations operating a mixture of traditional translation and machine translation post-editing workflows.
As AI changes service delivery, independent assessment can help organisations demonstrate that their documented processes are implemented effectively and that their certification claims accurately reflect the services provided.
Certification is not a guarantee that every translation will be error-free. It is independent confirmation that an organisation’s assessed processes meet the applicable requirements of the relevant standard.
For customers selecting language service providers, that distinction is important.
ISO 17100 and ISO 18587 Certification with ATC Certification
ATC Certification provides independent certification services specifically for language service providers, including ISO 17100 Translation Services and ISO 18587 Machine Translation Post-Editing.
Our specialist language industry auditors understand the operational challenges facing translation companies and the importance of maintaining quality as technologies and delivery models evolve.
Whether your organisation is considering certification for the first time, extending its certified services or reviewing existing processes in response to AI, recognised industry standards provide a valuable starting point.
Contact ATC Certification to learn more about ISO 17100 and ISO 18587 certification and how independent assessment can help demonstrate your commitment to quality in a changing language services industry.
This article draws on the EUATC and ISO/TC 37/SC 5 knowledge-sharing webinar, “Emerging ISO Standards for AI and What They Mean for Language Services”, held on 12 October 2026. It was drafted using Chat GPT AI from a transcript of the webinar and edited by humans.