AI in the workplace: who is liable when technology fails to keep workers safe?

Artificial intelligence (AI) is reshaping the workplace, especially in high-risk industries such as manufacturing and construction. From automated machinery and intelligent monitoring systems to predictive safety tools, employers are increasingly using technology to reduce the risk of accidents.

Used well, these systems could make workplaces safer by detecting hazards and warning workers before an accident occurs.

But what happens when the technology gets it wrong?

When AI malfunctions or an employer relies too heavily on automated safety processes, the consequences could be serious injury, permanent disability or even death.

As adoption grows, employers must ensure that innovation does not come at the expense of worker safety.

How is AI being used to improve workplace safety?

AI is already being tested and deployed across construction, manufacturing, warehousing and logistics.

Examples include:

• Intelligent monitoring systems: cameras and computer vision software can identify potential hazards, such as workers entering restricted areas or failing to wear appropriate protective equipment.
• Predictive maintenance: AI can analyse equipment data to identify patterns that may indicate a developing mechanical fault.
• Automated machinery and robotics: robotic systems can undertake hazardous, repetitive or physically demanding tasks, potentially reducing workers’ exposure to certain risks.
• Wearable technology: connected devices can monitor environmental conditions or provide alerts when workers enter potentially dangerous areas.
• Data analysis: AI tools can analyse workplace data to identify patterns and highlight areas where additional safety measures may be needed.

Sophisticated technology does not, however, make a workplace safe by default. Systems can malfunction, data can be incomplete and alerts can be missed. A tool intended to control risk may introduce new hazards if it is poorly implemented or relied on without adequate safeguards.

Employers must therefore understand both the capabilities and the limitations of any technology they introduce.

What happens when AI fails?

Imagine a warehouse where an AI-enabled camera system is designed to detect workers entering the operating area of automated machinery.

The system is intended to trigger a shutdown when someone enters a dangerous zone. However, under certain conditions, it could fail to recognise a worker. The machinery continues operating, and the worker suffers a serious crush injury.

Although hypothetical, the scenario exposes the questions that any investigation would need to answer.

Was the system suitable for the environment in which it was being used? Had its limitations been properly assessed? Were other protective measures in place? Had employees received adequate training? Was the machinery designed and maintained safely?

An AI-related workplace accident may have no single cause. The technology may have failed, but weaknesses in risk assessment, system design, maintenance, supervision or working practices may also have contributed.

What does the law say about AI and workplace safety?

The Health and Safety Executive (HSE) addressed this issue directly in its policy statement on 12 June 2026, setting out its regulatory approach to AI.

The HSE confirmed that existing health and safety legislation applies regardless of the technology being used. It also emphasised that risks arising from workplace uses of AI must be assessed and appropriate controls put in place.

The practical message is clear: AI may create unfamiliar hazards and technology should form part of a safe system of work, not become a substitute for one.

An employer cannot assume that installing an automated safety system removes the need for other precautions. Nor should workers be expected to compensate for unreliable technology by constantly anticipating its failures.

Under the Health and Safety at Work etc. Act 1974, employers have duties to ensure, so far as is reasonably practicable, the health, safety and welfare of their employees.

In practice, employers should assess the safety implications before deployment and keep those risks under review as the technology, data and working environment change. Employers should consider whether the system is reliable in the conditions in which it will operate, whether workers understand its limitations, what happens if it fails and whether effective safeguards are available.

Who is responsible if an AI-related workplace accident occurs?

AI can make responsibility less immediately obvious than in a conventional workplace incident, but complexity does not remove accountability.

Depending on the circumstances, potentially relevant parties could include:

The employer

An employer may have failed to carry out a suitable risk assessment, provide appropriate training, maintain equipment or implement adequate safeguards. Relying on an AI system does not, by itself, discharge the employer’s health and safety responsibilities.

The manufacturer or supplier

If a machine or safety system was defective, the circumstances may raise questions about the manufacturer’s or supplier’s responsibilities. Whether a product was defective, and whether any resulting liability can be established, will depend on the available evidence and applicable law.

The organisation responsible for integrating or maintaining the system

Where different technologies are combined, the way the system was installed, configured, tested or maintained may be relevant to understanding why an accident occurred.