When an injury occurs at the workplace workers compensation laws provide that employers accept liability for the injury while employers waive their right to sue the employer for damages. Employees have accepted this liability and legislatures have created workers compensation systems to address this.
More recently, employee injury data have been reported in the context of automation and artificial intelligence systems. This newer data paints a picture of a changing workplace.
Employers need to ensure that an injury was actually work related and sustained at the workplace while employees need to ensure workplace safety while conducting their duties.
Industries are already readjusting; Milliman’s 2025 review of statutory financials found that certain factors are influencing this sector: medical and wage inflation, robotics and AI moving into workplace safety programs, lingering debates about gig worker classifications, and new definitions for mental health injuries and hybrid work accidents.
The national workers’ comp loss ratio rose to 54.4% in 2025 from 53.8% in 2024, with ten of the fifteen largest states experiencing a decline.
Anticipating Work-Related Injury and Illness
An employer’s liability for employees’ injuries has always been reactive in nature. Employers wait for someone to get injured, they file a claim in terms of workers comp laws and if successful, the insurance pays out the claim. The introduction of AI systems in the workplace is beginning to disrupt this pattern.
- 81% of small businesses are willing to implement AI safety tools, according to Pie Insurance’s 2025 State of Workplace Safety Survey.
- 44% say they have already implemented those tools
- 73% of those who have done so have seen measurable improvements and outcomes.
Employers are monitoring their employees differently than in the past. With the help of modern advancements, automated video monitoring systems detect when employees might put themselves in danger before an accident happens.
Wearable tech such as fatigue monitors or in-vehicle driver tracking devices helps ensure that employees put the safety of themselves and others first. This will also lead to long term behavioral changes as the data used from these devices can be interpreted to make the workplace safer.
This has major implications on how workers’ compensation operates. AI inserts a new link into that chain so that danger is being replaced with prediction and prevention resulting in no injury.
Therefore, if a company is able to predict reliably and affordably that injuries may occur on a large-enough scale, the cost-effectiveness of the entire chain changes.
Three Fault Lines Are Emerging
As automation evolves throughout the world, three distinct fault lines are developing; these include:
Automation Is Changing Where Injuries Occur
Companies that use automation for many of the most physically demanding aspects of a job (such as repetitive lifting, machine loading) may find that while the nature of the work is changing and becoming less hazardous, the total number of injuries remain the same.
Instead, injuries start to happen in other parts of the production system: to maintenance workers repairing robots, to operators working in human-robot co-location cells, to line workers working with more reliable robots and higher speeds.
Understanding how workers comp works means the insurance industry will need new systems to classify injuries as claims are very different now than what they were in machines from 2015.
The Gig Economy Continues to be Difficult to Define
Remote and gig workers do not fit into a specific category under current workers compensation laws. This makes defining the relationship between employer and employee that much more difficult.
In turn the liability between employer and employees are blurred and employers are unsure of when they will be liable for injuries if a worker does not work at a specific location or worksite.
Numerous referendums, court cases, and exemptions have tried to address this issue by suggesting a change in workers compensation laws and a complete systematic review of the entire system.
Algorithmic Management Creates Injuries That Aren’t Addressed by Today’s Rules
The injuries that workers sustain are no longer limited to their use of heavy machinery or manual labour. Workers are managed by algorithms that determine where they must be on a worksite or what tasks they need to perform.
Tasks may include repetitive actions such as packing and closing boxes in a delivery hub, typing reports and schedules as well as monitoring screens for compliance related issues.
This may lead to newer injuries such as musculoskeletal discomfort and psychological stress, injuries that differ from those of 20 years ago.
What Happens if Portable Benefits Become Standard?
Should the workers’ comp system change and portable benefits be adopted as standard, the employer and employee relationship will fundamentally change. Liability will increase on the employer’s side as additional risk factors from employees moving to various working locations increases the risk of an injury.
Legislators will also need to make provision for limitations in coverage as employers will not be able to cover every aspect of an employee’s daily life should they choose a non-traditional job.
With employees having flexible working hours, worker comp rules will need to clearly define when an employee is actually performing their duties in order to determine when an employer will be liable for any injuries.
Employers need to be aware of the potential causes for injuries to understand when they will be able to claim compensation. Portable benefits must strike a careful balance between accountability and application of the workers’ comp rules.
The Actuarial Question Nobody Has Fully Answered
Should workers’ compensation continue to reflect the employer-employee relationship in terms of who bears the risks of injury, or is the pool of risk destined to become more portable and attached to the individual worker regardless of where he/she goes? All of the incremental changes are treating the current system as static and adding tweaks around the edges.
However, as both AI creates different hazards conditions for employees and fundamentally reshapes the nature of employment, it is becoming increasingly difficult to patch-up a 100 year old system without asking questions regarding its underlying assumptions.
The companies developing AI safety features, the insurance carriers seeking to optimize their line of coverage, and the legislators working to define a ‘worker’ will represent a real-time experiment on how much a legacy model can be stretched before it needs to be replaced.
