Promises and Perils of Relying on Artificial Intelligence (AI) for Regulatory Compliance
18 Mar 2026
As artificial intelligence (AI) tools become more accessible and user-friendly, it’s essential to understand both their benefits and limitations—some obvious, others less so.
When implemented correctly, AI has the potential to deliver significant benefits in regulatory compliance:
- Automation of Dossier Preparation, improving efficiency and accuracy;
- Regulatory Risk Assessment, predicting potential hazards and regulatory changes; and
- Global Regulatory Harmonization, fostering consistency and alignment across regions.
However, blindly following AI output can have significant negative consequences. Although AI tools are designed to be objective, they learn from existing data and can inherit or even amplify biases present within the dataset. Moreover, while AI is an exceptionally powerful tool for addressing known problems represented in its training data, it struggles to solve open-ended problems it has not previously encountered.
Let’s explore the possibilities and pitfalls of using AI for each of the above regulatory activities.
Automation of Dossier Preparation
Preparing submission dossiers requires multiple team members each contributing to the dossier based on their specific areas of expertise. Clear communication, teamwork, and final review are key to ensuring efficiency, accuracy, and overall consistency. However, with so many moving parts, even a simple 1% adjustment to the treatment rate can have far-reaching impacts within a submission dossier. If the impact of this change is only identified during the final review, the preparation of the dossier can be delayed weeks, if not months.
Promises: Improved efficiency and accuracy
AI tools allow routine or repetitive tasks to be completed automatically, freeing skilled regulatory professionals to focus on higher-value, strategic aspects of regulatory compliance. Automated workflows also enhance efficiency and accuracy while reducing the need for human intervention; in the example above, an automated workflow could update all the relevant dossier forms to reflect a 1% change in treatment rate, notify all team members of the change, and notify any affected downstream groups. Used well, AI tools can improve efficiency and accuracy and support timely delivery of complex submissions.
Perils: Misalignment with precise and context-sensitive language
While AI tools can produce fluent, grammatically correct English, they can also distort messages in subtle but meaningful ways. Referring to the subject of a regulatory submission as an “active ingredient” vs. a “notified substance” vs. a “substance of interest” may seem like a minor difference in wording, but the impact can be significant. If the submission concerns a new inert component of a pesticide, referring to it as the “active ingredient” would incorrectly suggest that the substance plays a toxicological role in the pesticide’s effect, potentially misleading government assessors and raising misguided questions. Experienced regulatory professionals must therefore review AI-generated content carefully to ward off subtle biases or misrepresentations that could undermine the notification strategy.
Regulatory Risk Assessment
Regulatory risk assessment is a key component of developing any new product. It involves critically evaluating hazards to human health, to the environment, and to the company arising from the product, and assessing the risk level at each stage of development and marketing. To ensure timely and relevant assessments, it’s critical to stay informed about public health initiatives, advances in environmental hazard monitoring, and changes in regulatory policy.
Promises: Predicting potential hazards and regulatory changes
AI tools can assess regulatory risk at various development and marketing stages. For example, they can flag substances associated with heightened hazards (e.g., a propellant listed under the Montreal Protocol for ozone depletion, or identified as a Substance of Very High Concern under EU REACH) before significant time and budget are invested. They can also flag novel chemistries as categories of concern, indicating that it may be prudent to conduct specific toxicological testing early to confirm the presence or absence of potential effects. Similarly, an AI tool might indicate that although a product raises no concern in the pharmaceutical or industrial chemical space, it may pose an issue in the cosmetics space due to growing restrictions on animal testing. Even in cases like these, the achievable benefits may outweigh the risks — ultimately, a human must make the final determination.
Perils: Difference between risk and opportunity
AI tools can identify risks but struggle to distinguish acceptable vs. unacceptable risks in context. The challenge lies in judgement, balancing scientific evidence, business objectives, stakeholder risk tolerance, and more. While AI can contribute to this process, business growth often requires measured risk and final decisions based on human skill, knowledge, creativity, and imagination.
Global Regulatory Harmonization
Ensuring compliance is essential for business growth in today’s global marketplace. However, different jurisdictions uphold different regulations to protect public health and the environment, and what is considered a concern in one jurisdiction may not be in another. Despite this, small considerations during product development can have far-reaching benefits when it comes to compliance on a global scale.
Promises: Fostering consistency and alignment across regions
AI tools can help ensure that product development aligns with the most stringent regulatory requirements, enabling a single product to achieve global compliance and reducing the need for separate formulations across jurisdictions. For example, not all regions regulate the content of volatile organic compounds in products, nor do all limit the amount of phosphorus‑containing substances in laundry detergents. Designing products that meet all these requirements reduces overall cost, supports consistent global branding, and enhances profitability.
Perils: Absence of common regulatory language
Approaches to regulating products and chemicals differ across regions worldwide. In Canada and the United States, polymers are treated as distinct chemical substances, whereas under the EU’s REACH regulation, it is the monomers that make up the polymer — rather than the polymer itself — that are subject to regulation. Similar inconsistencies exist for common products such as toothpaste, which can be considered a cosmetic or a non-prescription drug depending on the ingredients and claims. Without a clear understanding of the global regulatory landscape, relying entirely on AI output may lead a human user to the wrong conclusion.
You may ask your AI assistant: “Are polymers regulated under REACH?” It would be technically correct in answering “no” (because the monomers, not the polymer itself, get registered). However, this simple answer would be misleading and could lead to non-compliance in the EU. Experts using AI tools must formulate the right questions and interpret the answers within the full regulatory context.
Successfully Integrating AI
AI is a powerful technology that can improve efficiency and accuracy, identify trends, and support alignment across interconnected regulations and jurisdictions. But AI tools aren’t infallible, and human oversight remains essential. This blog has highlighted only a few potential risks associated with AI use.
Aside from errors, complications, and strategic misalignments, AI tools can introduce issues of governance, confidentiality, security, and transparency. On the technical side, avoiding endocrine‑disrupting chemicals in product development may lead an AI assistant to automatically exclude these substances; yet without target‑specific endocrine disruptors, many breast cancer and prostate cancer treatments wouldn’t exist. Perhaps more importantly, AI conclusions can be influenced by “popular” rather than “correct” answers. Subject matter experts must routinely review the training data to ensure the dataset remains free of bias, and implement other safeguards as needed.
As with any tool, AI is most effective in regulatory compliance when applied by professionals trained to recognize and mitigate risks while maximizing benefits. Intertek’s regulatory affairs professionals can assist you in achieving your regulatory compliance goals while using AI tools efficiently, accurately, safely, and transparently.
Intertek Is Here to Help!
Do you have questions about this topic or related topics? The experts at Intertek Assuris can assist you with navigating your regulatory compliance needs, please contact us at chemicals.sci-reg@intertek.com.
