Beware of AI-Generated Safety Data Sheets
As Artificial Intelligence rapidly transforms industries across Australia, many businesses are beginning to explore AI-generated Safety Data Sheets (SDSs) as a faster and cheaper alternative to traditional chemical compliance processes. While AI can assist with administrative tasks and document drafting, relying on AI alone to create legally compliant Safety Data Sheets may expose companies to serious compliance, safety, and liability risks.
Safety Data Sheets are not ordinary business documents. In Australia, SDSs are highly regulated technical and legal documents that must comply with strict requirements under Work Health and Safety (WHS) legislation and the Globally Harmonized System (GHS) for classification and labelling of chemicals. A single error in classification, exposure controls, dangerous goods transport information, toxicological data, or emergency response instructions can place workers, emergency responders, and businesses at risk.
One of the major concerns with AI-generated SDSs is accuracy. AI systems are designed to predict language patterns — not validate chemical science. Unless an AI platform is connected to verified chemical databases, current regulatory frameworks, and expert-reviewed rule engines, it can generate information that appears technically correct but is actually inaccurate, outdated, incomplete, or entirely fabricated. This phenomenon, commonly referred to as “AI hallucination,” is particularly dangerous in the chemical compliance sector where precision is critical.
Another key vulnerability is jurisdictional compliance. SDS requirements differ between countries and regulatory bodies. Many generic AI tools have been trained on mixed international content and may unknowingly blend U.S., European, and Australian regulatory requirements into a single document. This can result in Australian workplaces receiving SDSs that fail to meet local Safe Work Australia requirements, exposure standards, or dangerous goods classifications.
Chemical classification itself is another high-risk area. Proper classification requires detailed interpretation of toxicological, ecological, physical hazard, and ingredient concentration data. AI systems without validated classification engines may incorrectly assign hazard statements, signal words, or precautionary phrases. An incorrect classification can affect workplace storage requirements, PPE recommendations, placarding obligations, transport rules, and emergency response procedures.
There are also growing concerns regarding the source material AI systems use to generate SDS content. Some AI platforms may unknowingly rely on outdated SDSs, non-compliant templates, publicly scraped internet data, or overseas documents that do not reflect current Australian legislation. If incorrect information enters the AI model, that misinformation can be reproduced at scale across multiple Safety Data Sheets.
Emergency services and workplace responders also depend heavily on accurate SDS information during incidents involving fires, spills, chemical exposure, or toxic releases. Incorrect first aid advice, incompatible firefighting media, or inaccurate storage segregation guidance generated by AI could have severe real-world consequences during emergencies.
Despite these risks, AI still has an important role to play in the future of chemical management. When combined with verified regulatory databases, expert validation, structured chemical data, and compliance-focused software platforms, AI can significantly improve efficiency, assist with drafting, automate repetitive processes, and reduce administrative workloads. However, AI should support chemical compliance professionals — not replace them.
Australian businesses should exercise caution before relying on fully AI-generated Safety Data Sheets. Before adopting any AI-based SDS solution, companies should ask critical questions:
- Is the system connected to verified and current chemical databases?
- Does it comply specifically with Australian WHS and GHS requirements?
- Are SDSs reviewed or validated by qualified chemical compliance experts?
- Can the platform provide audit trails and evidence of classification decisions?
- How frequently are regulatory updates and exposure standards maintained?
- Does the software include structured rule-based classification rather than relying solely on generative AI?
As AI adoption accelerates, businesses must remember that compliance responsibility ultimately remains with the chemical manufacturer, importer, or supplier — not the AI tool. Inaccurate Safety Data Sheets can lead to regulatory penalties, workplace injuries, reputational damage, and significant legal exposure.
In the chemical industry, speed should never come at the expense of safety or compliance. AI can be a powerful assistant, but when it comes to Safety Data Sheets, expert oversight and verified compliance systems remain essential.