Briefing

The next frontier: AI tokens, quantum computing and the infrastructure behind AI

The debate about AI and legal risk has so far concentrated on the models themselves: who is liable when an AI system makes a decision that causes loss, and how that liability is allocated across a development chain. That debate is live, and it is being resolved now. But three adjacent categories are forming immediately behind it, and each of them will produce disputes of scale. They are AI tokens and the token economy, the physical infrastructure of data centres and semiconductors on which AI depends, and quantum computing. None of them is speculative. The businesses exposed to them are, for the most part, the same businesses already exposed to AI liability today.

The next frontier

Three adjacent dispute categories forming immediately behind AI liability.

AI token The token economy	Infrastructure Data centres and chips	Quantum Computing’s next leap
Misrepresentation on utility Governance and dilution Cross-border enforcement	Data-centre SLA failures Chip supply, export controls Foundry yield and quality	Performance shortfalls Broken encryption exposure Export-control disputes

Same profile as AI liability: cross-border, multi-party, proprietary. Arbitration is the forum.

AI tokens and the new token economy

AI tokens are digital tokens issued by AI platforms, AI protocol developers, or AI-focused decentralised applications, typically to fund development, reward data contributors, govern the protocol, or represent computational resources. It is a category that has received insufficient attention in legal literature, which is unlikely to remain the case for long.

The disputes arising in this space mirror in certain respects the cryptocurrency disputes that occupied the legal community over the past several years. Where an AI token is sold to investors on the strength of representations about the platform’s capabilities, the token’s utility, or the anticipated returns, and those representations prove incorrect or misleading, fraud and misrepresentation claims will follow. The governance dimension compounds the exposure: where token holders are given voting rights over model updates, data policies, or revenue distribution, disputes over the proper exercise of those rights, or the improper dilution of holders through the issuance of new tokens, create a novel and complex class of corporate-law-adjacent claim in what is, for the most part, an unregulated or lightly regulated space.

The cross-border enforcement challenges that characterised cryptocurrency disputes apply with equal force here. Token holders may be anonymous, distributed across dozens of jurisdictions, and holding assets that can be transferred instantaneously to any location in the world. The strategies developed in acting for investors across the major collapses of the last crypto winter, the Terraform Luna crisis, the Three Arrows Capital insolvency and the FTX collapse, are likely to be directly relevant to the AI token claims now taking shape.

Data centres, semiconductors and the infrastructure of AI

AI is not only a software phenomenon. Behind every large language model and every autonomous trading system is a physical infrastructure of considerable value and complexity: data centres housing thousands of servers, and semiconductors, particularly graphics processing units and specialised AI chips, that provide the computational power to train and run these models. Disputes about that infrastructure are contractual, high-value, and often multi-tier.

Data centres that host AI operations are typically the subject of complex power purchase agreements, cooling infrastructure agreements, and service-level commitments around uptime, security and performance. When a data centre fails to deliver agreed levels of service, the downstream impact on AI-dependent businesses can be catastrophic. Construction disputes around the commissioning of new facilities are becoming more common as demand for AI compute capacity outstrips the supply of purpose-built sites and developers attempt to bring facilities online under aggressive timelines.

The chip layer carries a different risk again. The global AI chip market is dominated by a small number of manufacturers, and the supply of advanced AI chips is subject to significant geopolitical risk, including export controls imposed by the United States on the supply of advanced chips to certain countries, including China. Disputes arising from those controls, whether they excuse contractual obligations under force majeure provisions, whether buyers have been induced by misrepresentations about the ability to supply, and whether sanctions regimes have been violated, are already a live category. Companies in the foundry and chip manufacturing sector face warranty disputes, supply chain claims and quality disputes, because the margins for error in precision semiconductor manufacturing are extremely narrow.

The foundry sector is subject to intense contractual disputes around yield rates, delivery timelines and the quality of wafers produced. Where foundry production failures cause delay or disruption to AI system development programmes, the downstream liability claims can involve multiple tiers of the technology supply chain, which is precisely the profile of claim that national courts handle least comfortably.

Quantum computing: the disputes of tomorrow

Quantum computing, the use of quantum mechanical phenomena to perform computations of a scale and speed unattainable by classical computers, is beginning to transition from the laboratory to commercial application, and it is approaching faster than many appreciate. The disputes it generates will be distinct from, but closely related to, AI disputes. Performance claims will be prominent: businesses that procure access to quantum computing services through cloud quantum platforms, and find that the advertised computational advantages do not materialise, will bring breach of contract and misrepresentation claims.

More significantly, quantum computing has the potential to undermine currently used encryption standards. If quantum computers can break the encryption protecting financial transactions, communications and data storage, the consequences for those relying on that encryption could be catastrophic, and the liability exposure of those who represented that encryption as secure, or who failed to implement quantum-resistant alternatives, could be enormous.

Export control disputes are already emerging. The United States Bureau of Industry and Security has imposed controls on the export of advanced quantum computing hardware and software, and companies engaged in cross-border quantum technology transactions are navigating the same kind of regulatory risk that characterised the AI chip supply chain disputes. Finally, the development of quantum-resistant AI algorithms, and the question of who owns the intellectual property in those algorithms, will generate IP and licensing disputes requiring expertise spanning quantum physics, computer science and international commercial law.

Where will these disputes be resolved?

Each of these categories shares the characteristics that make AI liability disputes so difficult to place. The data may have been collected in one country, processed in another, and fed into a model trained in a third. The model may be hosted in a data centre in Singapore, accessed via an API developed in the United States, and deployed in a commercial context governed by English law. Parties may be incorporated in multiple jurisdictions with assets scattered across the globe. National courts are, by definition, often bound by the mandatory limits of jurisdiction prescribed by the forum, even when we take into account bilateral treaties and conventions on the recognition and enforcement of judgments, which continue to expand the reach of court judgments across borders. International arbitration nonetheless retains a wider and more consistent scope of enforcement: an award rendered by an SIAC tribunal can be enforced in over 160 contracting states under the New York Convention 1958.

Confidentiality points the same way, though the comparison is one of degree rather than an absolute distinction. Businesses that compete on the basis of proprietary technology have a profound interest in ensuring that their disputes do not become public spectacles, and arbitration, which proceeds in private and imposes confidentiality obligations on the parties, the arbitrators, and the institutional staff, offers a structural protection that court proceedings often cannot match by default, even though courts in England and other jurisdictions can and do order confidentiality or private hearings in appropriate cases. The organising principle holds across every category described here: AI disputes are fundamentally about liability, not technology, and when that liability is disputed across borders, in complex supply chains, with proprietary technology at stake, international arbitration is where those disputes belong.

The practical conclusion for boards and general counsel is unglamorous but decisive. If a business uses, develops, deploys, or provides services connected to AI systems, it will wish to consider including an international arbitration clause in its contracts. The absence of one does not prevent a dispute; it simply means that when the dispute arrives, the parties do not spend the first stage of it arguing about jurisdiction and applicable law, a preliminary battle that can take years and millions of dollars to resolve before the substance is addressed. The machine does not stop. But the law must keep pace.