Robots, Artificial Intelligence and UK Law.

Robots and artificial intelligence are no longer confined to factories or science-fiction films. AI systems now help businesses recruit staff, assess customers, detect fraud, produce documents and make recommendations. Robots are being used in warehouses, hospitals, agriculture, transport and care settings.

The technology may be new, but the fundamental legal position remains familiar: a machine is not above the law, and it is not generally responsible for itself.

In the UK, robots and AI systems do not currently have a separate legal personality. They cannot normally be sued or prosecuted, own property, or enter into contracts in their own name. Where something goes wrong, responsibility usually falls on one or more people or organisations connected with the system.

Does the UK Have an Artificial Intelligence Act?

The UK does not currently have one general law regulating every type of artificial intelligence. Instead, AI is regulated according to where and how it is used.

Existing laws may apply to matters such as:

• Product safety
• Consumer protection
• Negligence and personal injury
• Employment and discrimination
• Data protection and privacy
• Medical devices
• Financial services
• Copyright and intellectual property
• Cybersecurity
• Automated vehicles

This means that an AI recruitment tool, a surgical robot and a self-driving vehicle may be subject to very different legal and regulatory requirements.

The UK's Regulatory Approach

The government has generally favoured a sector-based approach in which existing regulators apply principles relating to:

• Safety, security and robustness
• Transparency and explainability
• Fairness
• Accountability and governance
• Challenge and redress

The precise legal obligations depend on the system, the industry and the risks involved. A business cannot assume that AI is unregulated simply because there is no single UK Artificial Intelligence Act.

Can a Robot Be Legally Responsible?

No general UK law currently treats a robot or AI system as a legal person.

If a warehouse robot injures a worker, a medical system gives unsafe advice, or an automated service causes financial loss, the machine itself will not normally appear in court to explain what happened. Responsibility may instead rest with:

• The manufacturer
• The software developer
• The supplier or distributor
• The business that selected and deployed the system
• The employer operating it
• The person controlling or supervising it
• A maintenance or servicing provider
• A professional who relied on its output

More than one party may be responsible. The answer will depend on what went wrong, what each party knew, and whether the harm was caused by the design, software, data, maintenance, or the way in which the system was used.

Who Is Liable When a Robot Causes Injury?

Defective Products

A manufacturer, producer, importer or supplier may face a claim where a defective product causes injury or damage.

A robot may be defective because of:

• Unsafe physical design
• Faulty sensors
• Defective software
• Inadequate safety controls
• Cybersecurity weaknesses
• Insufficient warnings
• Failure to provide safe operating instructions
• An unsafe software update

The fact that the physical machine operated correctly may not end the matter. It's software or a connected service that's caused it to behave dangerously; those elements may also need to be investigated.

Negligence

A negligence claim may arise where a person or organisation owed a duty of care, breached that duty and caused reasonably foreseeable loss or injury.

Examples could include:

• A manufacturer failing to test a safety-critical system
• An employer allowing staff to use a robot without training
• A business ignoring known faults or safety warnings
• A maintenance company failing to repair defective sensors
• A professional relying unquestioningly on an obviously unreliable AI result
• A user disabling a safety mechanism

Calling a decision "automated" does not remove the duties of the organisation that chose to use the system.

Employers' Health and Safety Duties

Employers must take reasonably practicable steps to protect employees and others affected by their activities. These duties continue when robots or AI-controlled machinery are introduced.

An employer should consider:

• Risk assessments
• Safe areas and physical barriers
• Emergency stopping procedures
• Human supervision
• Staff training
• Maintenance and inspection
• Software and security updates
• How humans and machines work alongside one another
• What happens if communications or sensors fail

A robot designed to stop before striking a person is useful. An employer should still plan for the possibility that the sensor, software or emergency stop will fail.

Must Robots Have a"“Kill Switch"?

There is no universal rule requiring every robot or AI system to have a large red button marked "stop before becoming self-aware".

However, machinery and workplace safety requirements may demand suitable emergency stopping systems, isolation controls or other safety measures where these are necessary to control risk.

The appropriate safeguard will depend on the machine. It may include:

• An emergency stop button
• Automatic shutdown
• Physical guarding
• Remote isolation
• Speed or force limits
• A manual override
• Human approval before an action is completed
• A safe mode following loss of power or communication

A stop mechanism is only useful if it is accessible, regularly tested and capable of overriding the system when needed.

Who Is Responsible for an AI Decision?

Businesses sometimes describe an outcome as "the computer's decision". Legally, that explanation may be insufficient.

An organisation that uses AI should be able to explain:

• Why the system was selected
• What information it uses
• What the system is intended to do
• Who checks its performance
• What human oversight is provided
• How errors and bias are detected
• How a person can challenge the outcome

A developer may be responsible for defects in the technology, but the organisation deploying it may also be responsible for using it in an inappropriate setting or relying on it without adequate checks.

AI and Automated Decisions About People

Data protection rules may apply where an AI system uses personal information to assess, predict or make decisions about an individual.

This could include AI used to:

• Shortlist job applicants
• Monitor employee performance
• Assess applications for credit or insurance
• Detect suspected fraud
• Set prices
• Determine access to services
• Evaluate health or behavioural information

Organisations must have a lawful basis for processing personal information. They should ensure that its use is fair, transparent, accurate and secure.

Solely Automated Decisions

Additional protections can apply where a significant decision is made solely by automated means without meaningful human involvement.

Depending on the circumstances, a person may be entitled to:

• Be told that automated decision-making is being used
• Receive meaningful information about how it operates
• Challenge the decision
• Request human involvement
• Express their point of view
• Contest an incorrect or unfair outcome

Placing a person in the process merely to approve whatever the computer recommends may not amount to meaningful human oversight.

Bias and Discrimination

An AI system can discriminate even though a computer has no personal feelings or intention to treat someone unfairly.

Bias can arise from:

• Unrepresentative training data
• Historic patterns of discrimination
• Incorrect assumptions built into the system
• Using information that acts as a substitute for a protected characteristic
• Poor testing
• Failing to consider disabled users

An employer cannot avoid equality law by saying that an algorithm made recruitment or dismissal decisions. A service provider cannot avoid responsibility for discriminatory treatment simply because its software produced the result.

Organisations should test for unfair outcomes before deployment and continue monitoring the system throughout its use.

Robots and AI in the Workplace

Automation can remove dangerous or repetitive work, increase productivity and assist employees. It can also affect jobs, working conditions and workplace monitoring.

Employers introducing AI should consider:

• Whether employees need consultation
• Whether duties or roles will change
• Training and redeployment
• Redundancy procedures
• Health and safety
• Employee monitoring
• Data protection
• Discrimination and accessibility
• Whether automated performance scores are reliable

Will Robots Replace Employees?

An employer may reorganise its business or automate work, but it must still follow employment law.

If automation removes or reduces the need for particular work, a redundancy situation may arise. The employer may need to consult affected employees, use fair selection criteria, consider suitable alternative employment and make statutory or contractual redundancy payments.

It would not normally be enough to call an employee into a meeting, introduce their robotic replacement and ask them to hand over the keys.

Automated Recruitment and Management

AI may be used to screen applications, arrange shifts, measure productivity or recommend disciplinary action.

Employers should understand the limitations of these systems and should not assume that an automated score is objective or accurate.

Human review may be particularly important where an outcome could affect recruitment, promotion, pay, discipline or dismissal.

AI in Healthcare

AI can support diagnosis, analyse scans, monitor patients and assist with surgery. Healthcare AI may also be regulated as a medical device when it performs a medical purpose.

Responsibility may be shared between:

• The manufacturer or developer
• The healthcare provider
• The clinician using the system
• The organisation purchasing or configuring it
• The body responsible for maintenance and updates

A clinician may use AI as a tool, but should still exercise appropriate professional judgement. An unexplained or obviously inconsistent recommendation should not be accepted merely because a computer produced it.

Self-Driving Vehicles

Automated vehicles are one area in which Parliament has created a dedicated legal framework.

The Automated Vehicles Act 2024 provides for the authorisation and regulation of vehicles capable of driving themselves on roads in Great Britain.

The legislation distinguishes between genuine self-driving technology and systems that merely assist the driver. Where a system is only driver assistance, the human driver generally remains responsible.

Under the automated vehicle framework, responsibility can shift when an authorised self-driving feature is operating. The legal position may involve authorised self-driving entities, operators, insurers and users, depending on the vehicle and the circumstances.

Drivers should not assume that a vehicle is legally self-driving merely because its advertising uses words such as "autopilot", "automated" or "driverless".

Robots Used by Consumers

Businesses selling domestic robots, smart appliances, automated toys or AI-enabled products must consider consumer and product safety law.

Consumers may have rights where a product:

• Is unsafe
• Is faulty
• Does not match its description
• Is not fit for its intended purpose
• Stops working following an update
• Does not provide the promised functionality

Businesses should also consider how long software and security support will be provided. A connected product may become unsafe if known vulnerabilities are left uncorrected.

Cybersecurity and Hacking

A connected robot can create both physical and digital risks. A security weakness could allow an attacker to steal information, interrupt services or take control of the machine.

Manufacturers and operators should consider:

• Secure design
• Access controls
• Encryption
• Software updates
• Vulnerability reporting
• Network separation
• Incident response plans
• Safe shutdown following an attack

A business that continues using a system despite known security risks may face liability if foreseeable harm results.

Can AI Own Copyright?

AI systems do not own copyright or other intellectual property in their own name.

Copyright questions can arise over:

• The information used to train an AI model
• Whether protected work was copied
• Who owns AI-assisted output
• Whether the output reproduces a substantial part of an existing work
• The use of a person's voice, image or likeness
• The ownership of software, prompts and datasets

Businesses should not assume that content is free to use merely because an AI service produced it.

Should Robots Pay Tax?

Robots do not currently pay income tax, National Insurance or corporation tax in their own right.

The businesses that own and operate them remain subject to the normal tax system. Proposals for a special robot tax or universal basic income are matters of economic and political debate rather than current UK legal requirements.

Automation could eventually influence how governments raise revenue and support people whose jobs change or disappear. For now, however, the office printer has no National Insurance number and the warehouse robot will not be completing a tax return.

Do the Laws of Robotics Apply?

The often-quoted rules stating that a robot must not harm a human and must obey human instructions originated in science fiction. They are not UK legislation.

Real-world machines may have safety rules programmed into them. Still, legal compliance cannot be reduced to three simple instructions.

A delivery robot may need to protect pedestrians, comply with traffic rules, preserve personal information and avoid property damage. These obligations can conflict and require proper design, testing and supervision.

What Should Businesses Do Before Using AI or Robots?

Before deploying an AI system or robot, a business should consider:

• Its intended purpose
• The reasonably foreseeable risks
• Who is legally and operationally responsible
• Whether sector-specific approval is required
• Product and workplace safety
• Data protection
• Equality and discrimination
• Human oversight
• Cybersecurity
• Insurance
• Contractual responsibility between suppliers
• Record keeping and audit trails
• How decisions can be challenged
• How the system can be stopped or withdrawn safely

Contracts should clearly address maintenance, software updates, access to data, intellectual property, security incidents, defects and responsibility for losses.

What Should You Do If an AI System Causes Harm?

Preserve evidence as soon as possible. Relevant evidence may include:

• The machine or device
• System logs
• Software versions
• Prompts and outputs
• Photographs and video recordings
• Medical records
• Contracts and operating instructions
• Training records
• Maintenance history
• Risk assessments
• Communications with the manufacturer or supplier

Avoid altering or disposing of the system before obtaining advice, as this could destroy important evidence.

Depending on what happened, advice may be required from a product liability, personal injury, employment, data protection, commercial or intellectual property solicitor.

The Future of Robot Law

Artificial intelligence is developing faster than many conventional legislative processes. Further laws and regulatory requirements are therefore likely, particularly for advanced AI, autonomous systems and AI used in safety-critical or high-impact decisions.

The most important principle is unlikely to be that robots become legally responsible for themselves. It is more likely that organisations must remain accountable for the systems they create and use.

Robots may become more independent. However, for the foreseeable future, they will still need a responsible human or organisation standing behind them. Putting the robots in charge of writing the rules may improve the drafting speed. However, whether they would remember to include a functioning off switch is another question.

Disclaimer: Solicitors.com is not a firm of solicitors. Content on this site is provided for general information about UK law and is not legal advice. The applicable law may differ between England and Wales, Scotland and Northern Ireland and may depend on the particular technology or industry. You should obtain advice from a regulated solicitor about your circumstances. Although we aim to keep information accurate and up to date, AI law and regulation are developing rapidly. Use of this site does not establish a solicitor-client relationship.

Feedback: Is anything on this page incorrect or incomplete? Suggested amendments may be credited. Please email us.