From Warning to Prevention: AI and the Reasonable-Person Standard in Digital Negligence

by

Kapila Premarathne

CyJurII Scholar

on 20 August 2026

Introduction

The law of negligence asks “what would a reasonable person have done in the circumstances?”   The answer for this question is typically expected after something done wrong. But with the development of AI-assisted digital platforms, a further question is becoming just as important as “can we expect more rationality from a digital system when it is increasingly capable of recognising risks that the human user may not recognise?”

Existing literature has examined how negligence law can apply to AI, what duties may arise for AI developers and deployers, and whether AI itself should be assessed against a reasonable standard. Ryan Abbott has argued that computers could eventually become safer than humans and potentially provide a new benchmark for reasonable conduct. ¹ Mihailis Diamantis has similarly proposed a negligence standard for evaluating AI itself.² Other scholars have considered reasonable care across the AI value chain and have highlighted the difficulties of allocating responsibility between different actors.³

This article asks a somewhat different question by extending the same arguments further as” can AI operationalise aspects of the reasonable-person standard itself?” If AI can increasingly recognise risk, understand context and infer apparent intention, could reasonable care move from being mainly a retrospective legal test towards becoming a more prospective mechanism for preventing foreseeable digital harm?

From Warning to Intelligent Prevention

Digital platforms have relied heavily on warnings to reduce foreseeable risks for decades. Users are expected to recognise the risk and make the appropriate decision. But the warning does not necessarily mean that the risk has been properly understood. AI changes this possibility because platforms may now be capable of analysing contextual information and identifying risks that the user may not recognise. The question is therefore no longer only whether the platform warned the user, but whether it could reasonably have identified and proportionately prevented the risk.

AI as Part of the Reasonable-Person Standard

The reasonable person is not the most intelligent person, and the standard is not simply a prediction of what most people normally do. It is a normative legal standard used to determine what reasonable care requires in particular circumstances. However, the practical content of reasonable care has never been completely separated from technological development. In Blyth v Birmingham Waterworks Co, the standard was expressed through the conduct expected of a reasonable person in the circumstances.⁴  In Bolton v Stone, a foreseeable risk did not automatically require every possible precaution. The probability of harm and the burden of avoiding it remained the operative considerations, much as the Learned Hand formula would later make explicit in American law.⁵ AI therefore does not need to replace the reasonable person for technology to influence reasonable care. But it needs only to shift what a reasonable actor, weighing the same probability and burden, would now be expected to do.

My argument differs from the “reasonable computer” proposition. Abbott asks whether computers may eventually become the benchmark against which human conduct is judged.⁶ Diamantis asks how AI itself should be judged under negligence law.⁷  The question present in this article here is different as “ What if AI becomes an instrument through which the reasonable person exercises reasonable care?”. The logical answer is that the reasonable person would remain the legal standard . AI would not become the legal comparator. Instead, AI could provide the reasonable actor with capabilities that were previously unavailable.

When Does AI Capability Become Reasonable Care?

AI capability does not automatically create a legal duty, particularly where systems may misunderstand context or produce biased outputs. Medical-AI research illustrates that existing negligence principles can accommodate AI, while implementation, reliance, monitoring and supervision may influence the applicable standard of care.⁸ ⁹ The UK Jurisdiction Taskforce similarly recognises that existing English private law can address many AI-related harms.¹⁰ The key question is therefore whether a sufficiently reliable and reasonably available AI capability should form part of reasonable care when addressing a foreseeable risk.

The Importance of Apparent Intention

This becomes particularly interesting when AI systems can examine behavioural patterns rather than individual requests. A user may divide a harmful objective into a series of apparently harmless requests. Each request may appear acceptable when considered separately. However, the sequence may reveal a different objective. An AI system capable of considering the broader interaction could potentially ask “What is this user apparently trying to accomplish?”

AI can literally read a person’s mind. The issue is whether apparent intention can reasonably be inferred from the context, sequence and pattern of behaviour. If AI becomes sufficiently capable of making such assessments, a new legal question emerges as “ when does knowledge of probable harmful intention create a reasonable expectation that the system should intervene?”

It is particularly relevant where users deliberately attempt to circumvent AI safeguards through prompt engineering. A user who misunderstands a warning is not necessarily in the same position as someone who deliberately restructures a sequence of requests to avoid a safeguard. At the same time, a platform should not automatically become negligent simply because a more sophisticated system might theoretically have detected the behaviour. Foreseeability, reliability and proportionality must remain important.

The Reciprocal Responsibility Problem

This does not mean that responsibility should simply move from users to AI developers or digital platforms. Suppose an AI-enabled system identifies a significant risk and gives the user an opportunity to stop the activity. If the user deliberately circumvents those safeguards, that conduct may become relevant to contributory negligence. The Law Reform (Contributory Negligence) Act 1945 provides for reduction of damages where the claimant’s own fault has contributed to the harm.¹¹

The opposite situation is also possible. If a platform can reasonably identify a foreseeable risk but only provides a generic warning, its failure to prevent the risk may be relevant to negligence.

Fraser and Suzor recognise that AI-related harm may involve different actors and forms of reasonable care.¹² This article goes further by asking whether AI itself can become part of the mechanism through which reasonable care is exercised in real time.

Conclusion

AI does not require negligence law to abandon the reasonable-person standard. Rather, AI may become part of the system through which reasonable care is exercised before harm occurs. The key question is therefore when reasonable care should require the use of AI capabilities for risk recognition and prevention.

Footnotes

1.     Ryan Abbott, ‘The Reasonable Computer: Disrupting the Paradigm of Tort Liability’ (2018) 86 George Washington Law Review 1, 1–45.

2.     Mihailis E Diamantis, ‘Reasonable AI: A Negligence Standard’ (2025) 78 Vanderbilt Law Review 573, 573–615.

3.     Henry L Fraser and Nicolas P Suzor, ‘Locating Fault for AI Harms: A Systems Theory of Foreseeability, Reasonable Care and Causal Responsibility in the AI Value Chain’ (2025) 17 Law, Innovation and Technology 103, 103–138, DOI: 10.1080/17579961.2025.2469345.

4.     Blyth v Birmingham Waterworks Co (1856) 11 Exch 781, 784.

5.     Bolton v Stone [1951] AC 850, 858.

6.     Ryan Abbott, ‘The Reasonable Computer: Disrupting the Paradigm of Tort Liability’ (2018) 86 George Washington Law Review 1, 1–45.

7.     Mihailis E Diamantis, ‘Reasonable AI: A Negligence Standard’ (2025) 78 Vanderbilt Law Review 573, 573–615.

8.     Gary Kok Yew Chan, ‘Medical AI, Standard of Care in Negligence and Tort Law’ in Gary Kok Yew Chan and Man Yip (eds), AI, Data and Private Law: Translating Theory into Practice (Hart Publishing 2021) 173–198, DOI: 10.5040/9781509946860.ch-008.

9.     Gary KY Chan, ‘AI in healthcare: Regulatory guidelines and judge-made negligence principles for AI implementers’ (2026) 26 Medical Law International 144–168, DOI: 10.1177/09685332251362405.

10.  UK Jurisdiction Taskforce, Legal Statement on Liability for AI Harms under the Private Law of England and Wales (7 July 2026).

11.  Law Reform (Contributory Negligence) Act 1945, 8 & 9 Geo 6 c 28, s 1(1).

12.  Henry L Fraser and Nicolas P Suzor, ‘Locating Fault for AI Harms: A Systems Theory of Foreseeability, Reasonable Care and Causal Responsibility in the AI Value Chain’ (2025) 17 Law, Innovation and Technology 103, 103–138, DOI: 10.1080/17579961.2025.2469345.