Is AI adoption in manufacturing outpacing governance?

Like 19th century pioneers adventuring west in search of fortune and prosperity, many manufacturers are embarking on their own intrepid AI journeys. There is no doubt that the last 12 months has seen a shift in the needle, and the tentative conversations that circled around the ‘if’ of AI have given way to the ‘how’. Experimentation has given way to expectation, ambition has shifted to delivery, and potential has yielded to proof.

Yet, while bravely heading out into the unknown, there are a number of questions that remain unanswered around AI’s deployment in manufacturing. That’s not to say that the sector is blindly ploughing ahead with technology without due process and diligence. Key considerations are being asked around expected ROI, the most suitable applications where AI can make a difference, workforce buy-in and integration with legacy systems.

Key takeaways

• AI adoption in manufacturing has shifted from experimentation to real implementation: Companies are no longer asking if they should use AI, but how to deploy it effectively and deliver measurable results.
• There is a major gap between AI usage and governance: While 93% of organisations are using AI, only seven per cent have proper governance frameworks in place—creating significant risk as regulations like the EU AI Act come into force.
• Ungoverned AI can amplify existing organisational weaknesses: Issues like poor data quality, weak processes, or compliance gaps can become much worse when scaled through AI systems.
• Key risks of AI include bias, privacy breaches, misinformation, and over-dependence: AI can produce inaccurate or biased outputs, mishandle sensitive data, reduce human critical thinking, and even create legal risks around intellectual property.
• Strong governance frameworks are essential for safe and effective AI use: Organisations need clear oversight, regular audits, ethical guidelines, and structured standards (like ISO 42001) to manage risk, ensure compliance, and build trust.

FAQs

• What is meant by “ungoverned AI”?
• Why is AI governance important in manufacturing?
• What are the main risks of using AI without proper controls?
• How does the EU AI Act impact businesses?
• How can organisations implement effective AI governance?

These are undoubtedly key issues that need to be part of the conversation. However, the sheer acceleration of the technology and its capabilities has seen it outpace its own governance; the rules, processes and decision-making structures used to oversee and guide how AI is managed, controlled and held accountable.

The most significant provisions of the EU AI Act will come into force later this year, and new data shows that while 93% of organisations are already using AI, a mere seven per cent have embedded governance frameworks in place, creating a major gap as regulatory scrutiny increases.

The AI Act is a European regulation on artificial intelligence (AI) – the first comprehensive regulation on AI by a major regulator. The act assigns applications of AI to three risk categories. First, applications and systems that create an unacceptable risk, such as government-run social scoring of the type used in China, are banned. Second, high-risk applications, such as a CV-scanning tool that ranks job applicants, are subject to specific legal requirements. Lastly, applications not explicitly banned or listed as high-risk are largely left unregulated.

Kirsty Wakefield, Information Security Sector Manager at ISO certification body ISOQAR, has warned that many organisations are effectively operating with “ungoverned AI”, where tools are deployed without clear oversight, accountability or risk controls.

She commented: “While it continues to amaze us, and many organisations are racing to adopt, and work alongside it, in 2026, we remain both hopeful and cautious. Its sub-optimal, yet increasingly human-like intelligence, along with its power to transform traditionally linear technologies into limitless opportunities raises an important question: how far can it go before it slips beyond our control?”

AI is now being developed and deployed across virtually every sector and organisations across manufacturing are leveraging everything from customer service chatbots, to automation tools and analytics models to streamline processes, open new revenue avenues, and gain competitive advantage.

Of course it is also important that AI is developed and deployed safely, ethically, and in line with business priorities and wider societal values. Kirsty added that although AI can create new problems of its own, it is more likely to amplify existing weaknesses within an organisation. Issues like inconsistent data management, gaps in compliance and regulation, or operational weaknesses such as discrepant workflows can quickly amplify and become systematic issues that are much harder to resolve.

She added: “With proper governance in place, businesses can conduct regular audits as part of an AI compliance process to help identify potential risks early, and address them before the implementation of new AI technologies.

“Rarely before have organisations turned to a software or tool that prompts us to question who really controls who. AI and machine learning tools possess boundless capabilities that require a deep understanding of their intended purpose, abilities and functions.

“Machine learning tools in particular, continue to adapt and learn automatically without direct programming and without the need for direct human intervention. In the absence of a clear organisational understanding of how these systems operate, AI tools can become unpredictable, and even with well-intended use, businesses risk deploying tools that can lead to errors, bias, data breaches or even ethical violations.”

Unlike humans, AI tools lack context and awareness of both truth and ethics. And as non-sentient systems, they lack the ability to distinguish between right and wrong and accurate and inaccurate information. In particular, generative AI models (such as chatbots or text/image generators) can produce content based on hallucinated facts, incorrect reports or misleading statements that appear credible. In some cases, ungoverned tools may also be misused to manipulate information, intentionally creating misinformed content, false claims and scenarios.

Concerns about these risks have already influenced regulatory responses including provisions within the EU Artificial Intelligence Act that restrict certain high-risk AI systems. Without governance, small inaccuracies can quickly lead to widespread misinformation, potentially leading to reputational damage, compliance breaches and overall erosion of public and customer trust. Kirsty picks up the story.

Privacy and data protection

Ungoverned AI can pose significant risks to privacy, especially when dealing with sensitive data. AI technology can process, infer, and categorise sensitive personal data at a scale and depth that traditional systems cannot. This includes the ability to infer sensitive attributes, or store and analyse biometric information.

Without clear governance, large datasets may be collected, processed, or retained without a clear legal basis. Worse yet, when managed incorrectly, this data could be unintentionally shared across departments, or to vendors and third parties.

Possessing critically sensitive information without the right governance and safeguards in place could quickly lead to catastrophic data breaches, data leaks or cyber security incidents which can become irreversible if they ever take place.

Bias and unethical categorisation

When collecting sensitive information and data, another widely discussed risk of AI is algorithmic bias. As AI systems learn patterns from the data they are trained on, if that data reflects inequalities, incomplete information, or societal bias, AI models can replicate this and even amplify those patterns when making decisions.

Even when protected characteristics such as gender or ethnicity are removed, these can still be interpreted through proxy indicators such as postcode, education history, language patterns, or purchasing behaviour. Biased outcomes can quickly affect various systems used in recruitment, lending, or risk assessment and could unintentionally disadvantage certain groups.

Bias may not be noticeable immediately, and may emerge gradually as systems continue learning from new data or interacting with users over time. For this reason, regular audits, testing and monitoring are needed to ensure that unfair outcomes do not persist and scale.

Over-dependence and deskilling

While many of us are grateful for the opportunity artificial intelligence has provided the workplace by cutting out routine and time-consuming tasks, it has also introduced an increased risk of over-dependence on the tools. While throughout history technology has regularly replaced skills and job roles, AI tools can lead to reduced critical thinking, decision-making capability and professional judgement within the workforce.

To mitigate this risk, organisations may consider introducing clear AI usage policies dependent on team functions which will ensure that AI tools are used to support human capability rather than replacing it entirely.

IP and copyright

As we know, Generative AI tools are capable of generating new content such as text, images, audio and video, computer code and more. These models are trained on billions of pages of data, and information from every breadth of topic. Much of this material is copyright-protected or privately owned data, leading to disputes around whether the processing, and then output of generative material, breaches privacy, IP, and copyright laws.

Whether it be marketing content, software code, or product developments, AI may produce content that is considered a derivative of copyrighted material, leading to lawsuits, fines, or takedowns. Businesses therefore need clear policies, oversight, and legal review processes to avoid becoming liable for accidental infringement.

Environmental risks 

Artificial intelligence requires substantial computing power, and large-scale data storage creating huge server footprints. AI training alone is highly energy intensive and when running AI tools data centre energy consumption and cooling can skyrocket.

Without governance and strict oversight, businesses may be unable to grasp, track, and report on their AI energy usage. This could undermine sustainability commitments and make it harder to track progress toward business ESG and net-zero targets.

Therefore, clear objectives and governance standards are required from the outset to ensure that organisations prioritise energy-efficient models within their workflows, and integrate AI usage into sustainability goals and reporting.

Conclusion 

As AI adoption grows across all industries, the potential benefits are immense. However, these benefits are partly offset by the risks associated with poorly governed AI systems. As this analysis shows, ungoverned AI carries significant and multifaceted risks where the consequences of deploying Artificial Intelligence without robust governance, are real, tangible, and potentially severe.

The correct tools, oversight, and standards are needed to implement AI responsibly and ethically. Driven by structured frameworks like ISO 42001, organisations must focus on identifying risks before they escalate, and ensure compliance with legal, ethical, and moral obligations for its successful deployment in business.