AI-Driven Corporate Disputes: New Frontiers For Litigators

Artificial intelligence (“AI”) is reshaping corporate litigation, raising novel questions about evidence, liability and predictive analytics. With 65% of organisations now regularly using generative AI – courts, regulators and litigators face pressing challenges.

This article considers how AI-generated material is being used as evidence in court proceedings, corporate liability risks arising from AI use as well as the potential - and limitations - of predictive litigation tools.

AI generated sources and evidence in court

AI tools are increasingly used in legal research, e-discovery and document review, delivering significant efficiency gains. In many cases, AI tools reduce tasks from long hours to minutes. However, the reliability of AI-generated outputs remains a serious concern. This is a matter that requires heightened awareness from lawyers.

The risk of AI "hallucinations" has already caused courtroom consequences. For instance, in Wadsworth v Walmart Inc (D. Wyo. 2025), a lawyer's admission to practise was revoked and attorneys were fined USD5,000 for citing non-existent AI-generated cases.

Similarly, in Kohls v Ellison (D. Minn. 2025), the court struck an expert declaration after finding that fake, AI-generated sources had destroyed the expert's credibility.

A further example is the case of Gauthier v Goodyear Tire & Rubber Co (E.D. Tex. 2024), where an attorney was sanctioned for submitting fabricated cases and quotes generated by AI.

These cases provide a clear message: lawyers bear the same professional responsibility for AI-assisted work product as for any other output. AI-generated material must be rigorously verified before submission to a court.

To remedy the shortcomings in law in relation to AI practices, regulatory frameworks are developing in parallel. In the European Union (“EU”), the Artificial Intelligence Act (Regulation (EU) 2024/1689) classifies AI systems used in the administration of justice as high-risk, requiring fundamental rights impact assessments.

In South Africa, the Protection of Personal Information Act No.4 of 2013 (“POPIA”) imposes obligations regarding data privacy and evidential integrity. The "black box" problem – whereby AI algorithms produce outputs without revealing the reasoning behind them, making it difficult for lawyers and judges to scrutinise or challenge AI-driven decisions – poses a direct challenge to the right to a fair trial under Article 6 of the European Convention on Human Rights, and research into "explainable AI" remains ongoing.

Recently, the South Africa’s Department of Communications and Digital Technologies published the Draft South Africa National Artificial Intelligence Policy (“the Draft Policy”) for comment. However, the Draft Policy was short-lived. Barely two weeks after its publication, the government withdrew the Draft Policy on 26 April 2026 following reports that at least 10% of the academic references in its reference list did not exist. The fiasco raised concerns that the policy had been produced using the same AI tools it was intended to regulate, without adequate human verification. In that respect, the drafters appeared not to have followed the very principle of "vigilant human oversight" promoted in the Draft Policy.

Taken together, these examples show how relying on unverified outputs can weaken legal processes and undermine regulatory credibility, ultimately creating the kinds of corporate risks discussed below.

Liability risks for corporates

AI deployment creates a spectrum of liability risks across contract, delict, data protection and regulatory compliance. An error in an AI system can affect vast numbers of individuals before it is even identified.

Case law is already emerging in this regard. In Leeway Services Ltd v Amazon, the claimant alleged that Amazon's AI systems caused its wrongful suspension from an online marketplace.

In Tyndaris SAM v MMWWVWM Limited, it was argued that an AI-powered system's capabilities had been misrepresented. Furthermore, in Moffat v Air Canada, Air Canada was held to have failed to ensure the accuracy of its chatbot's responses to customers. These cases highlight the breadth of claims AI deployment can generate.

Algorithmic bias is a particularly acute risk. AI trained on historical data may perpetuate existing biases – as seen with the COMPAS system used in US courts, which relied on proxies such as poverty that disproportionately disadvantaged African American defendants. Corporates using AI in customer-facing or employment contexts face analogous exposure under equality and anti-discrimination legislation.

Data protection adds a further layer of risk. In the UK, the Information Commissioner's Office has published a strategic approach to AI, and the Financial Conduct Authority is examining AI's impact on financial services. Corporates must conduct thorough due diligence on AI vendors and implement robust privacy policies.

Regulators are also targeting "AI washing" – misleading claims about AI capabilities. In March 2024, the US SEC settled with Delphia (USA) Inc and Global Predictions Inc over misleading statements about their use of AI.

This heightened scrutiny increases the prospect of private claims following adverse regulatory findings. Group litigation is a particular concern: AI's speed and scale are well suited to the mechanics of collective claims.

In England and Wales, whilst Lloyd v Google [2021] set a high threshold for representative claims, and subsequent decisions such as Commission Recovery Limited v Marks and Clerk LLP have shown that appropriately structured group claims remain viable.

The ongoing appeal in Wirral Council v Indivior PLC / Reckitt Benckiser Group, in which investors seek to recover losses arising from allegedly untrue statements in public documents, may prove pivotal. The Court of Appeal's decision on how that securities law claim may be structured could set an important precedent for future group claims based on alleged misrepresentations about AI capabilities. Collective proceedings in the Competition Appeals Tribunal represent a further avenue, with technology companies increasingly targeted.

These overlapping risks show that AI-related disputes will likely become larger and more complex, meaning that while predictive tools can be incredibly useful for anticipating outcomes, they also come with increasing challenges and uncertainty.

Predictive litigation: Promise and peril

AI-driven predictive litigation tools use historical data to forecast case outcomes. Legal Judgment Prediction systems developed for courts including the ECHR, the Swiss Federal Supreme Court, and the US Supreme Court have achieved accuracy rates of approximately 79%.

The potential benefits are significant: better-informed litigation decisions, improved access to justice and dramatic efficiency gains, with some AmLaw 100 firms reporting that AI reduces certain tasks by a factor of one hundred. The prevailing view is that AI will not replace lawyers but will shift work towards higher-level analysis and strategy – an "80/20 inversion" from information gathering to strategic thinking.

However, inaccurate predictions risk poor decisions and may dissuade individuals from pursuing meritorious claims. Algorithmic bias and opacity raise concerns about fairness and accountability, and the European Commission for the Efficiency of Justice has stressed that AI in the legal system must align with fundamental rights, including the right to a fair trial and judicial independence. There is also a risk that judges may defer too readily to AI recommendations.

For law firms, the competitive implications are clear: those that invest in AI tools and develop proprietary methodologies will differentiate themselves, whilst those that do not risk falling behind.

In short, predictive litigation can offer meaningful strategic benefits, but it only works well when guided by careful human judgement, supported by reliable data and managed with safeguards to prevent bias.

Conclusion

AI is set to reshape corporate litigation, but its adoption raises complex legal, ethical and practical questions. Courts must develop consistent standards for AI evidence. Corporates must manage liability risks spanning bias, data protection, regulatory enforcement and group claims. Predictive technologies offer transformative potential but demand careful oversight. For litigators, understanding AI is no longer optional. Rather, it is essential to competent practice in an era where technology and the law are inextricably linked.