AI litigation: From emerging risk to commercial reality

Artificial intelligence is no longer a theoretical disruptor for litigators. It is already shaping claims, altering evidential landscapes and influencing the way disputes are fought and resolved.

As part of London International Disputes Week, we brought together practitioners across litigation, IP, arbitration and regulatory disciplines to explore where AI disputes are heading. The discussion made one point abundantly clear: AI is not creating entirely new areas of law, but it is accelerating and amplifying existing legal risks (and opportunities) in ways that businesses are only just beginning to confront.

The result is a rapidly evolving litigation landscape where traditional causes of action are being applied to new, technology-driven fact patterns, often at scale and under increasing political scrutiny.

The emergence of AI-driven claims

Perhaps the most striking development is the shift from theoretical to real-world claimant activity.

In the United States, we are already seeing claims framed around algorithmic harm, which range from addiction-style cases linked to recommendation systems through to allegations involving severe behavioural impact caused by interactions with AI chatbots. These cases are often packaged as personal injury claims, drawing comparisons with earlier waves of litigation against industries such as tobacco or opioids.

For UK practitioners and businesses, this is significant not because the underlying legal principles are novel, but because they are not. Claims in negligence, product liability and consumer law are highly adaptable. Where there is a compelling narrative of harm the courts already have the tools to respond.

This raises a real prospect of similar litigation in the UK, particularly as political attention increases. Recent debate around restricting child access to social media, coupled with broader concerns about algorithmic influence, suggests that this is becoming an increasingly active policy space.

If software is brought more clearly within the scope of product liability regimes (a question currently under consideration) the barrier to entry for claimants may reduce significantly. In that scenario, AI-related disputes could begin to resemble life sciences or consumer product litigation, with the potential for group claims and significant damages exposure.

Intellectual property risk: focusing on outputs

From an intellectual property perspective, much of the debate to date has centred on inputs: specifically, whether training large language models on copyrighted material constitutes infringement. That question remains unresolved, and the UK government has so far been reluctant to take a definitive position, seeking to balance the economic advantages offered by AI-driven innovation while still protecting the creative industries.

However, for most businesses, the practical risk lies in AI-generated outputs. It is increasingly easy for teams to produce content that inadvertently infringes third-party rights. Marketing teams may generate visuals or campaigns that closely resemble existing brands; internal prompts may encourage replication of third-party material; and the resulting outputs can create both actionable infringement and problematic disclosure in subsequent litigation.

The risk of infringing materials being published or used commercially, is compounded by the risk of incriminating prompt histories emerging during the disclosure phase of subsequent litigation.

The speed and accessibility of AI tools mean that these issues are no longer confined to specialist teams, which are likelier to be more aware of potential attendant risks. They are playing out across businesses at all levels.

AI as a litigation amplifier

AI is not only creating new disputes; it is also changing how disputes unfold. One practical issue highlighted is the growing impact of AI on volume and cost in litigation. Litigants in person, for example, increasingly use AI tools to generate extensive and persistent correspondence. While this may improve perceived access to justice, it can also place a disproportionate burden on defendants required to respond.

The implications for disclosure are equally significant. AI systems often generate large volumes of underlying data, including prompts, iterations and outputs. Managing, preserving and analysing this data is resource-intensive. A failure to do so may give rise to a Court drawing adverse inferences.

Liability for AI-generated statements

A key emerging principle (already visible in international case law) is that businesses remain responsible for what their AI systems say.

A recent case involving a chatbot providing incorrect information demonstrates how courts are likely to approach these issues. The fact that a statement is generated by AI does not absolve the business of responsibility. Instead, it is treated in the same way as any other representation made to a customer.

In the UK, this aligns closely with consumer protection principles, including the prohibition on misleading actions. Crucially, a statement does not need to be false to be actionable. It is sufficient that it is misleading in a way that affects consumer decision-making. Given the increasing enforcement powers available to the Competition and Markets Authority, including the ability to impose substantial financial penalties, this is likely to become an area of heightened risk.

AI in International Arbitration

International Arbitration is already established as an excellent means by which to enforce contractually agreed allocation of risk across borders. We do not see AI changing that, although as with litigation the way in which arbitration is practiced and its attraction for certain sectors is changing.

Arbitration is a confidential process. Consequently, it is often preferred by technology companies, including AI / AI driven companies: no disclosure of parties, arbitrators or substance; often no disclosure of the existence of the case itself. Arbitration also permits selection of the Tribunal, which is attractive for parties in the technology and AI spaces: your case will be decided by someone who understands the technology and current trends. Arbitration agreements already feature more detail on the technical expertise of arbitrators, trend that sems likely to develop. As a process, arbitration remains flexible and efficient, as it can be tailored (by party agreement) to the parties’ needs. Again, in a sector as fat-moving as AI, this has obvious advantages.