AI adoption is creating legal consequences for employers: expert

McKinsey data shows 79% of organisations globally have adopted generative AI, yet only 39% attribute any earnings before interest and taxes (EBIT) impact to it, and just over one-third have scaled AI beyond pilots. Despite this, research across multiple jurisdictions shows that most workers report AI making their jobs more stressful, driven by work intensification, constant upskilling pressure, and anxiety around unverified AI outputs.

Aaron Goonrey, an expert in employment law at Pinsent Masons, said: “Across a number of jurisdictions, employers are confronting the same challenge: significant investment in AI without a clearly demonstrated productivity dividend. For boards and executives, this raises an important governance question of if the benefits remain uncertain, have workforce risks been properly assessed and managed?"

“The conversation about AI has largely focused on capability, efficiency and investment. Increasingly, regulators are asking whether organisations have properly managed the human consequences of deployment,” he said.

“Many organisations have approached AI as a technology project and are now asking whether the AI investment has been worth it. The more urgent legal question is whether they have discharged their obligations in deploying it, particularly given the emerging legal risks and regulators increasingly viewing AI utilisation as a workforce, governance and risk management issue.”

Recently, peer-reviewed research and a study of 1,488 employees found that intensive AI use is creating cognitive fatigue, with the American Psychological Association warning of erosion of professional confidence and independent reasoning.

Goonrey said: “Emerging evidence suggests that intensive AI use may contribute to cognitive fatigue, reduced confidence in independent decision-making and increased psychosocial pressures within the workplace. Employers should be paying close attention to those developments."

“Increasingly, the question is not whether an organisation will adopt AI, but whether it can demonstrate appropriate governance, consultation, oversight and risk management in doing so,” he said.

"The organisations facing the greatest legal exposure are unlikely to be those that adopted AI too quickly or too slowly. They will be the organisations that treated AI as a technology initiative rather than a governance, workforce and risk management exercise.”

Goonrey said employers should be considering psychosocial risk, workforce consultation, role and workforce redesign, training, governance frameworks and performance expectations at the same time they implement new AI tools, not after the fact.

The industrial consequences of AI deployment are also becoming more visible. A September 2026 Unions NSW report, based on a national survey of more than 6,600 workers, found that 88% of respondents regularly performed unpaid overtime, averaging 7.8 hours each week. Among workers whose workplaces had introduced AI, 20% said employers expected higher output as a direct result, while 13% reported an increase in unpaid overtime. More than one-quarter believed AI could replace their role within five years.

The report also found that 22% of respondents were already having work directed or allocated by digital systems, while 15% reported reduced autonomy over how their work was performed. Unions NSW has consequently called for stronger overtime protections, recognition of unpaid overtime as a work health and safety risk, and greater union education concerning AI and workplace rights.

Goonrey said: “These findings demonstrate why AI deployment cannot be separated from working hours, workload management, employee autonomy and the right to disconnect.”

“If AI-generated efficiencies simply become higher targets, tighter deadlines or an expectation that employees remain available for longer, employers may exchange a productivity opportunity for a significant employment and industrial relations risk. Employers tend to focus on whether the technology works. Employees tend to focus on what the technology means for them. Most workforce disputes arise in the gap between those two perspectives," he said.

"Where employees perceive AI as threatening job security, increasing surveillance, diminishing autonomy or intensifying workloads, employers should anticipate greater scrutiny from employees, unions and other workforce representatives. 

"Consultation should not be treated as a procedural step at the end of implementation. It is an essential part of identifying risk, testing assumptions and bringing the workforce with the organisation. Employers that fail to engage early may face workforce distrust, organised employee resistance, disputes and, potentially, industrial action. In many jurisdictions, consultation is not simply good practice. Depending on the circumstances, it may be a legal obligation."

AI is also likely to become an increasingly significant focus of employee activism. Questions about algorithmic management, automated decision-making, surveillance, job security and whether productivity gains are being shared fairly are moving rapidly into mainstream industrial relations.

Regulators, too, have taken notice. For example, Safe Work Australia released specific regulatory guidance in July 2026 confirming that AI and digital technologies can directly introduce or worsen psychosocial hazards. Equivalent frameworks are active in the UK, under Health and Safety Executive (HSE) guidance; across EU member states, under the EU AI Act and existing health and safety directives; and in Singapore and the Hong Kong Special Administrative Region (SAR).

Goonrey said: “The legal consequences for employers extend well beyond technology regulation. They can arise through work health and safety obligations, consultation requirements, discrimination laws, workplace surveillance issues, industrial relations disputes, performance management practices and, in some jurisdictions, AI-specific regulatory regimes."

"As AI becomes embedded in workplaces, boards and executives will increasingly be expected to demonstrate active oversight and governance," he said.

"I was not aware is unlikely to be a persuasive response where risks were foreseeable and information should reasonably have been escalated."