As artificial intelligence becomes embedded in everyday business operations, from automated decision-making to predictive analytics, companies are increasingly asking a critical question: who is liable when AI causes harm or makes a wrong decision?
The legal answer is not always straightforward. Liability will depend on how the AI system is used, the contractual framework in place, and the nature of the harm caused. In this article, we explore how liability is approached under English law and what businesses should consider when deploying AI.
Why ai liability is legally complex
Unlike traditional software, AI systems can:
- Operate autonomously
- Learn and adapt over time
- Produce outcomes that are not always predictable
This raises key challenges:
- Who is responsible – the developer, the user, or both?
- Can liability arise even without fault?
- How do existing legal frameworks apply to evolving technologies?
In most cases, existing legal principles still apply, but they must be carefully interpreted in the context of AI.
Potentially liable parties
1. The business using the AI system
In many scenarios, the end user (i.e. the business deploying the AI) will bear primary responsibility.
This is particularly likely where:
- The AI is used to make commercial or operational decisions
- The business has control over how the system is implemented
- There is a failure to monitor or validate outputs
For example, if an AI tool incorrectly rejects customer applications or generates flawed financial data, the business relying on that output may be liable.
2. The AI developer or software provider
Liability may also arise on the part of the developer or vendor, particularly where:
- The system is defective or not fit for purpose
- There is a breach of contract (e.g. warranties or service levels)
- Misrepresentations were made about the system’s capabilities
Well-drafted software agreements often include:
- Limitations of liability
- Exclusions for indirect loss
- Defined performance standards
These provisions are critical in determining where risk ultimately sits.
3. Multiple parties (shared liability)
In many AI-related disputes, liability is shared across multiple parties, for example:
- A developer provides flawed software
- A business deploys it without adequate oversight
- A third party suffers loss as a result
In such cases, the courts may consider contributory negligence or apportion liability between parties.
Legal frameworks that may apply
Although there is no single “AI liability law” in the UK, several existing legal frameworks are relevant:
🔹 Contract law
Most disputes arise from contracts governing AI systems, including:
- Software development agreements
- SaaS agreements
- Licensing terms
Liability is often determined by what the parties agreed at the outset.
🔹 Negligence
A party may be liable if they owe a duty of care and fail to meet the required standard.
For example:
- Inadequate testing of an AI system
- Failure to supervise high-risk automated decisions
🔹 Product liability
If AI is embedded in a product, liability may arise where it is defective and causes damage.
🔹 Regulatory considerations
The UK is developing an evolving regulatory approach to AI, with increasing focus on:
- Transparency
- Accountability
- Risk management
Failure to comply with regulatory expectations may strengthen liability arguments.
Common scenarios where disputes arise
We are increasingly seeing disputes involving:
- Automated decision-making errors (e.g. credit scoring, recruitment tools)
- Faulty AI outputs in financial or operational systems
- Bias or discrimination claims linked to AI models
- Failure of AI-driven products or platforms
- Misrepresentation of AI capabilities in contracts
These disputes often combine technical complexity with commercial impact, requiring a strategic legal approach.
The role of ADR in AI and tech disputes
Given the complexity and sensitivity of AI disputes, Alternative Dispute Resolution (ADR) is often the preferred route.
Benefits include:
- Speed and efficiency compared to court proceedings
- Confidentiality, particularly important for proprietary technology
- Flexibility, allowing tailored technical solutions
- Preservation of commercial relationships
Methods such as mediation and arbitration are particularly well-suited to tech disputes, where specialist expertise may be required.
At IMD Corporate, we regularly advise clients on resolving technology disputes through ADR as part of our broader Alternative Dispute Resolution offering.
Practical steps to manage AI risk
To reduce the risk of liability, businesses should:
- Carefully draft and negotiate AI-related contracts
- Define clear allocation of risk and responsibility
- Include robust limitation of liability clauses
- Implement human oversight and validation processes
- Conduct ongoing monitoring and testing of AI systems
- Keep records of decision-making and system outputs
Prevention is often significantly more cost-effective than resolving disputes after the event.
Conclusion: who is liable?
There is no single answer, but in most cases, liability for AI-related harm depends on control, contractual terms, and the specific facts of the case.
Often, responsibility will fall on the business using the AI, the developer, or both.
As AI continues to evolve, so too will the legal landscape. Businesses that proactively manage risk and seek early legal advice will be best placed to navigate this area.
How we can help
At IMD Corporate, we advise businesses on technology disputes, including issues arising from AI systems, software contracts, and automated decision-making.
Visit our Alternative Dispute Resolution page to learn more about how we can help resolve disputes efficiently, or contact our team for tailored advice.
This article is for general information only and does not constitute legal or professional advice. Please note that the law may have changed since this article was published.
