AI and agency law: why AI cannot act on your behalf

As artificial intelligence becomes more sophisticated, there is a tendency to describe AI systems as if they were autonomous agents, capable of making decisions or taking actions independently. However, UK agency law is grounded in the concept of a legal relationship between two persons. This article explores why AI cannot be considered an agent in legal terms and what this means for businesses using AI technologies.

Understanding legal agency

Agency is a legal relationship in which one person, the agent, is authorised to act on behalf of another, the principal. It is based on mutual consent and requires the agent to have the legal capacity to act. The agent must also owe duties of loyalty and accountability to the principal.

Legal capacity means that only natural persons (humans) or entities with legal personality (such as companies) can act as agents. This excludes non-human entities like AI systems, which cannot form intent, give or receive consent, or be held accountable in law.

Fiduciary duties include obligations such as acting in the principal’s best interests, maintaining confidentiality, and avoiding conflicts of interest. For example, a human agent negotiating contracts for a company must not disclose sensitive information to competitors or act for their own benefit at the company’s expense.

AI systems do not meet these requirements. They are not legal persons and therefore cannot give or receive consent. This means they cannot act as agents in the eyes of the law. Instead, AI models are considered ‘property’ or bundles of Intellectual Property rights belonging to the AI system’s owner. Access is provided as a licence or a service to the end-user.  Whilst AI do not have the capacity to owe legal duties or form the intent necessary to establish a fiduciary relationship, their developers and owners do. When a business licences or utilises an AI system, instead of a fiduciary relationship forming, a contractual relationship will, under either the online terms and conditions or through an negotiated AI agreement executed by the parties. It is important to review and negotiate the obligations of both the owner and the user to reflect this relationship.

Common misunderstandings

There are also misconceptions when discussing AI and agency. These include:

• Anthropomorphism: Treating AI systems as if they think or behave like humans, rather than recognising them as tools governed by software
• False analogy with agents: Assuming AI can perform the role of an agent simply because it carries out tasks or makes decisions
• Legal personality confusion: Believing that AI systems can function as legal entities with their own rights and responsibilities.

Each of these misunderstandings can lead to errors in contract formation, delegation of duties, and allocation of liability.

Implications for contracting and liability

Where AI is used in business operations, the legal responsibility remains with the human or corporate actors who deploy and control the system. For example, if an AI tool approves a loan or processes customer data, any failure to comply with legal standards will fall on the person or entity overseeing the system – not on the tool itself.

This has practical consequences for contract drafting, particularly where automation plays a role in executing or initiating agreements. Contracts should clearly allocate responsibility for unclear drafting, biased wording or ineffective execution by an AI system. They should set out how the AI system is used, by whom, under what oversight mechanisms (such as regular audits), and specify escalation procedures if issues arise.

Although some legal scholars and policymakers have discussed the idea of granting limited legal personality to AI, there is no serious move toward this in UK law. Recent case law and regulatory guidance emphasise that the use of AI must be framed within the current legal framework, not outside it.

AI may be capable of simulating intelligent behaviour, but it cannot act as an agent as currently defined under law. Businesses should ensure that the use of AI is supported by clear legal structures, with responsibility firmly assigned to human decision-makers. Standard commercial contract negotiations and liability apportionment should occur when implementing an AI tool, to ensure both the owner and end-user are adequately protected. As AI technologies rapidly evolve, the role of legal advisors will be essential in ensuring that innovation proceeds within the boundaries of the law while protecting businesses from unforeseen liabilities.

If you need guidance in relation to an agency agreement, our experienced team can help with this. Contact us to discuss how we can support your contract review and optimisation strategy for 2025 and beyond.

You can read a similar article here.