Artificial intelligence (AI) has outgrown its original job description. It no longer simply supports human decision-making from the sidelines. It now executes transactions, manages workflows, monitors performance, engages with customers and, increasingly, makes decisions with real commercial and personal consequences. That shift has and will continue to not only change how organisations operate, but importantly, will change how disputes arise, and who is left holding the risk when something goes wrong.
The legal issues themselves are not new. A customer challenges a decision that affects their access to a product or service. An employee alleges that an automated recruitment or performance management system produced a discriminatory outcome. A business claims that an AI system supplied by a technology provider failed to perform as promised. A regulator investigates the processing of personal data or the impact of an AI-enabled system on consumers or competition.
What is different is the part AI played in producing the outcome is now at the centre of the dispute. As AI becomes more deeply embedded in business processes, organisations will increasingly have to answer some uncomfortable questions, the crucial one being: when an AI-enabled system gets something wrong, who is responsible?
When the dispute is about more than the decision
Traditional disputes turn on relatively familiar questions. Who made the decision? What information did they have? What did they know at the time? Did they act reasonably? Did they comply with their contractual, statutory or other legal obligations? AI complicates each of those questions.
An AI-enabled decision typically involves several different actors. An organisation may have procured the technology from one provider, relied on data from another source, and configured the system internally. Employees may oversee some decisions, while others are automated or subject to only limited human intervention.
When the outcome is challenged, responsibility is often spread across a complex chain of parties. The dispute may concern the organisation that deployed the system, the technology provider, the quality or use of the underlying data, the way the system was configured or monitored or the extent to which a human was expected to intervene. In practice, several of these issues frequently arise at once.
This is the accountability gap that lies at the heart of AI-related disputes. AI systems can make decisions, or materially influence them, while legal and contractual responsibility for those outcomes remains far less clear.
Familiar legal principles, unfamiliar facts
AI is not creating an entirely new body of law. Most AI-related disputes will continue to arise within established areas of law, including contract, delict, employment, competition, data protection, intellectual property and consumer protection.
What is changing is the factual complexity that courts, regulators and other decision-makers will be asked to work through.
Take an AI system that produces a discriminatory outcome. The dispute will rarely turn simply on whether discrimination occurred. It will raise questions about the data used to train or operate the system, the assumptions built into its design, the way it was deployed and whether the organisation had appropriate mechanisms in place to audit the system and identify and address problematic outcomes.
Similarly, where an AI system fails to perform as expected, the dispute will involve more than a straightforward question of contractual breach. Parties will need to examine how performance was defined, what representations were made about the system's capabilities, whether its limitations were properly disclosed and how responsibility was allocated between customer and technology provider.
In competition matters, AI-enabled pricing or market conduct raises difficult questions about the role of algorithms in producing outcomes that would traditionally have been assessed through human behaviour.
The legal principles are familiar. The technology, the evidence and the allocation of responsibility are not.
The evidence problem
One of the most significant challenges in AI-related disputes is evidential. In a conventional dispute, it is usually possible to identify the individual who decided and examine the information available to them at the time. With AI-enabled systems, understanding how and why a particular outcome was produced requires a far more detailed examination of the technology and the environment in which it operated.
The relevant questions are practical ones. What data informed the decision, and which version of the system was in use? How was the system configured and what instructions, parameters or limitations were applied? Was the output subject to human review and were concerns or risks identified before the decision was made? What monitoring or governance processes were in place, and can the organisation reconstruct how the decision was actually reached?
The ability to answer those questions depends on records created long before any dispute was contemplated. This is where AI governance and dispute preparedness begin to intersect.
An organisation may have policies governing the use of AI but if it cannot demonstrate how a particular system was selected, deployed, monitored and overseen, those policies will provide limited assistance once a decision is actually challenged. Documentation, auditability and clear governance structures are becoming as important from a disputes perspective as they are from a compliance perspective.
The importance of knowing where responsibility sits
AI systems are often introduced through complex commercial arrangements involving multiple suppliers, platforms, data sources and internal stakeholders. When something goes wrong, the contractual framework becomes critical.
Organisations need to consider whether their agreements clearly address system performance, data quality, security, responsibility for errors and the consequences of failure. They should also consider what happens when an AI system produces an outcome that neither party specifically anticipated, a scenario that arises more often than most contracts are drafted to accommodate.
The question is not simply whether a contract contains an AI clause. The more important issue is whether the contractual allocation of risk reflects how technology actually operates.
A contract that assumes a clean distinction between a supplier's product and a customer's use of that product sits uneasily with a system whose performance depends on data, inputs, configuration and ongoing interaction between multiple parties. As AI becomes more sophisticated and more deeply integrated into business operations, the gap between contractual assumptions and operational reality is itself becoming a source of dispute.
Dispute readiness starts before a dispute
Organisations typically start thinking about disputes once something has gone wrong. By that point, the relevant decisions have been made, the contracts have been signed, the systems have been deployed and the records either exist or do not. AI does not leave room for that approach.
The organisations best placed to manage AI-related disputes will not necessarily be those deploying the most sophisticated technology. They will be those that can demonstrate how their systems were chosen, assessed, used, governed and monitored, and that can clearly explain the role human oversight played in every aspect of consequential decisions.
This does not mean treating every AI system as a potential claim. It does mean treating AI-related disputes as part of technology's lifecycle from the outset. Governance, contractual protection, data management, oversight and documentation all shape what happens if a dispute later arises. In that sense, dispute readiness is becoming an important component of responsible AI deployment, not a separate exercise from it.
A new accountability challenge
The growing use of AI does not remove responsibility from organisations. If anything, it sharpens the question of where that responsibility lies.
When a decision is challenged, affected parties, regulators and courts are unlikely to be satisfied by the explanation that "the system made the decision". The more difficult question will be who was responsible for the system, how it was used and whether appropriate steps were taken to manage the risks associated with it.
As AI becomes more deeply embedded in commercial and operational decision-making, disputes will follow. The challenge for organisations is to ensure that they are prepared not only to adopt the technology but also to explain, defend and, where necessary, account for the decisions it produces.
Ultimately, the next generation of AI-related disputes will not turn on whether artificial intelligence was involved. It will turn into a more familiar question, asked of a less familiar technology: when the decision is challenged, can anyone clearly explain how it was made and where responsibility lies?
Webber Wentzel's
AI offering guide sets out how the firm assists clients to navigate these issues in practice.
