If you’re running a business in Queensland today, you’re undoubtedly considering – or already using – some form of Artificial Intelligence (AI). From sophisticated generative models that draft your marketing copy to algorithmic systems that manage your supply chain, AI is transforming how we operate. It offers incredible efficiency, but as a commercial lawyer, my priority is ensuring your business is legally protected when dealing with this complex technology.
The legal landscape governing AI is still developing globally, and Australia is no exception. This means that if something goes wrong – if the AI makes an error, breaches confidentiality, or even signs a bad deal – pinpointing who is responsible isn’t as straightforward as it used to be.
Standard contracts simply aren’t robust enough for AI. If you are procuring an AI system, or if your business relies on an automated process, you need a new approach to contracting. Let’s demystify the core challenges of AI liability and outline the practical steps you can take to shield your business, drawing on insights from Katie Koo’s 2024 article, Navigating Legal Liability in AI-Driven Contracts.
What Are AI Contracts?
AI contracts are agreements that are drafted, negotiated, or executed with the help of artificial intelligence. These can range from simple automation – like filling in standard terms – to more advanced systems that use generative AI to create or modify contract clauses based on data inputs and learned patterns.
The appeal is clear: AI can save time, reduce human error, and streamline repetitive tasks. But it also introduces uncertainty, especially when the AI behaves in unexpected ways.
Key Risk Areas in AI Contracts
Here are some of the main legal risk zones you need to think through:
- Warranties & performance guarantees
The AI provider might warrant that the system will perform according to specification or standards. But overpromising is dangerous – you may end up chasing compensation for underperformance or errors. - Limitation of liability and indemnities
It is common in commercial contracts to limit liability (e.g. cap damages, exclude certain types of losses) and to include indemnity clauses (one party agrees to compensate the other for specific third-party claims). But in AI contexts, these limits must be well thought through – for example, you may want carve-outs for gross negligence, regulatory fines, or data breaches. - Liability for faulty outputs or decisions
Suppose an AI tool generates a report that leads to a business loss. Who bears responsibility? You need to clarify whether the provider is liable (and to what extent), or whether the user (you) bears risks. - Intellectual property, data rights and training data
AI is often trained on data sets, which may include copyrighted or sensitive materials. If training data is misused, or the AI generates infringing content, you may face copyright or data-use claims. - Transparency, audit and explainability
You may want the contract to require access to logs, explanations of decisions, and audits, so you can understand how the AI reached an outcome. - Regulation, compliance, and liability to third parties
If your AI system violates privacy laws, consumer protection laws, or causes harm to third parties (e.g. negligence), you need contractual mechanisms to allocate those risks.
The Problem: Who’s Liable When AI Makes a Mistake?
One of the biggest challenges with AI in contracts is liability. If an AI tool inserts the wrong clause, omits a key term, or generates a contract that doesn’t reflect the parties’ intentions, who’s responsible?
Traditionally, contract law relies on the idea of intention – that both parties understood and agreed to the terms. But AI, especially generative AI, can act in unpredictable ways. It learns from data, adapts over time, and may produce outcomes that even its creators didn’t foresee. This “black box” nature makes it hard to trace errors back to a single source.
The Operator’s Role: Are They Always to Blame?
Some legal scholars, like Eliza Mik, argue that the operator of the AI – typically the business or developer deploying the system – should bear responsibility. After all, they control the system and choose how it’s used.
But Koo’s analysis challenges this view. She points out that as AI becomes more autonomous, especially with machine learning and generative capabilities, the operator’s control diminishes. In these cases, it may be unfair to hold them solely liable for every error.
Instead, liability should be assessed based on who had control at each stage of the process:
- Design phase: Did the operator or developer set appropriate parameters?
- Training phase: Was the data accurate and relevant?
- Deployment phase: Were users properly informed and trained?
This approach allows for a more nuanced, fairer allocation of responsibility.
Why Strict Liability Might Not Work
Some have suggested applying strict liability – where the operator is automatically responsible for any harm caused by the AI. While this might protect users, it could also stifle innovation. Businesses may be reluctant to adopt AI tools if they risk being sued for every glitch or unexpected outcome.
Strict liability also raises ethical concerns. If a user misuses the AI or ignores obvious errors, should the operator still be on the hook? Probably not. A more balanced approach considers reasonable care – did each party take appropriate steps to prevent harm?
The Role of Explainable AI
One promising solution is the development of explainable AI. These systems are designed to make their decision-making processes transparent and traceable. This helps users understand how a contract was generated and makes it easier to identify where things went wrong.
Explainability also supports fairer legal outcomes. If we can see which party influenced a particular clause – whether it was the operator, the user, or the AI itself – we can better determine who should be liable.
What Happens If a Contract Goes Wrong?
If an AI-generated contract contains errors, parties may seek remedies under common law or equity.
Under common law, a contract might be voided if there’s a serious breach, misrepresentation, or mistake. But these remedies can be limited, especially when the error stems from the AI rather than either party’s actions.
Equity offers more flexibility. For example, the doctrine of undue influence could apply if one party relied heavily on the AI and was disadvantaged by its output. Courts may also consider whether the parties acted in good faith and whether the AI’s role created an imbalance in bargaining power.
Australia’s Legal Landscape: Flexibility Over Regulation
Australia currently has no AI-specific legislation governing contracts. While this may seem like a gap, it also allows for flexibility. Businesses can negotiate terms that suit their needs and adopt AI tools without being bogged down by rigid rules.
That said, the government has issued AI Ethics Principles – voluntary guidelines that promote fairness, transparency, and accountability. Following these principles can help businesses demonstrate that they’ve taken reasonable steps to manage risk, which may be relevant if a dispute arises.
Practical Tips for Businesses
If you’re using or considering AI in your contracting process, here are some steps to protect yourself:
- Understand the Tool: Know what your AI system can and can’t do. Don’t rely on it blindly.
- Set Clear Parameters: Limit the AI’s scope where possible – especially in high-risk contracts.
- Review Outputs Carefully: Always have a human review the final contract before signing.
- Train Your Team: Ensure staff understand how to use the AI responsibly.
- Document Everything: Keep records of how the AI was used, including inputs and outputs.
- Include Liability Clauses: Clearly outline who is responsible for errors in your contracts with developers, users, or clients.
Final Thoughts
AI is reshaping the way we do business, and contracts are no exception. While the technology offers exciting opportunities, it also raises complex legal questions. By understanding the risks, adopting best practices, and staying informed, businesses can harness the power of AI while protecting themselves from liability.
As the law evolves, we’re likely to see more guidance on these issues. In the meantime, a thoughtful, balanced approach – combining human oversight with smart technology – is your best bet for navigating the AI contract landscape.
If you’re unsure how AI fits into your contracting process, or if you’ve encountered a dispute involving AI-generated terms, it’s worth seeking legal advice tailored to your situation. The future is digital, but the law still matters.
