AI in the Workplace: Legal Risks for Employers

By Sophia Radford, Solicitor, Employment Law.

2025 saw an explosion of growth in AI users, both commercially and personally, and that momentum has continued into 2026.

AI is by no means a new concept, but the integration of generative AI models into organisations has accelerated at pace.  Back in 2023, a survey undertaken by KPMG found that 20% of businesses were already using generative AI, and 67% of executives confirmed that budget was allocated towards generative AI technology. By early 2026, KPMG found that 88% of organisations are already embedding AI agents into their workflows, products and value streams. Furthermore, 54% of organisations are actively deploying AI agents across core operations. 

Large language model generative AI holds enormous potential across industries. Whilst organisations consider how best to implement generative AI in the workplace and develop internal AI platforms, employers should also be aware that many employees may already be using freely available and easily accessible AI models, such as ChatGPT, Gemini or Microsoft Copilot, because of the productivity gains they offer both professionally and personally. 

Notwithstanding the widespread awareness of AI, many employers lack AI policies to ensure employees use AI appropriately.  Nor have organisations trained employees to understand the fallibility of AI or to recognise how and where bias may be built into AI models.

Hallucinations

The risks of generative AI have been widely discussed, and inaccuracy, hallucinations and bias were the predominant reasons cited for caution. The plausible and authoritative presentation of AI-generated information means that both users of generative AI and the recipients of that information may have no immediate reason to question its validity or accuracy.  

There are currently 57 reported AI ‘hallucinations’ cases (where an AI model creates inaccurate or false information) in the UK, and this number continues to grow as more litigants in person who use AI as a cost-effective means of assisting with litigation.

The increased use of AI in employment tribunal litigation creates not only a risk of hallucinations but also an increase in producing lengthy AI-generated documents that may contain inaccuracies.

All organisations and their representatives should ensure that, if they are presented with a claim, case citations, legal arguments and summaries are carefully checked and verified. This is critical both to maintain trust with clients and to ensure compliance in tribunal proceedings.

Bias and Discrimination

A fundamental point to be reminded of when using AI models is that AI is only as good as the data it is trained on, and the inputted data can create inherent bias within the models. 

The risk of bias in AI was highlighted in one of the first AI cases to be presented to the Employment Tribunal: Manjang &Raja v Uber Eats UK ltd and others:3206212/2021. The claims related to harassment, victimisation and indirect race discrimination arising from the introduction of the Microsoft facial recognition app, which required drivers to take a real-time photo of themselves to verify their identity when working through the Uber Eats app. The use of this facial recognition app as a verification tool resulted in the Claimants’ relationship with Uber Eats being terminated following repeated failures of facial recognition.

At a Preliminary Hearing, the Claimant asserted that facial recognition software placed people from ethnic minority groups at a disadvantage because false positive and false negative results were more common in those groups. The Claimant had raised these concerns with Uber Eats and requested a human review of the app, but received no response from the Respondent.   No judgment was given; the Claimant reached an undisclosed settlement with Uber Eats. However, the case highlights the risk of assuming AI is infallible and relying on it without human review. 

It was acknowledged by Dr Gideon Christian, Assistant Professor at the University of Calgary, who researches race, gender and privacy impacts of AI facial recognition technology, that: "There is this false notion that technology, unlike humans, is not biased.  That's not accurate.  Technology has been shown to have the capacity to replicate human bias."  This replication of human bias will affect content generation and decision-making such that the use of AI algorithms may create a risk of discrimination.  

Where organisations are seeking to use AI in automated decision-making, human intervention is essential at some stage in the process to check for inherent bias. It is imperative that if an individual claims that any form of AI being used by an organisation is producing incorrect or biased outcomes, the organisation must take immediate action to review the system and its use.

Data Protection and Intellectual Property

The use of generative AI in the workplace has also raised concerns about compliance with data protection legislation, both within and outside the UK. 

To assist in compliance with data protection legislation, employers should:

  • advise employees about the data they are permitted to input into generative AI tools;
  • provide training and awareness of acceptable use;
  • explain and justify the use of AI models when processing personal data;
  • clearly set out information regarding the use of AI models in privacy notices;
  • ensure data protection impact assessments are completed before introducing any new AI model;
  • implement  appropriate security measures and controls regarding acceptable use of personal data and AI; and
  • where personal data is used in AI tools, consider that data can be located, extracted or amended in compliance with a DSAR. 

From an intellectual property perspective, there are risks for organisations both in developing generative AI models and in using AI models in the workplace.   These include questions around ownership of AI-generated output, whether outputs may infringe protected works, and the extent to which intellectual property rights can attach to the content that has not been created by humans. 

The rapid development of AI technologies and the legal questions they raise around intellectual property are beyond the scope of this article.  However, they remain a key consideration for any organisation using AI models, particularly where retaining control of intellectual property rights is commercially important.

Transparency

As with data protection, transparency is a central theme in the regulation of AI. It is vital that organisations are transparent about their use of AI and require the same transparency from their workforce. 

This is especially important when employees use free generative AI models. Given how prevalent and influential AI technology has become across all sectors, a prohibition on generative AI is unlikely to be practical and unhelpful.  It may also discourage the openness needed for organisations to identify and manage risk. 

In a study undertaken by Salesforce in 2023, which involved over 14,000 workers across 14 countries, more than half of those interviewed were using generative AI in the workplace without their employer’s approval. The use of “shadow AI", meaning unapproved AI tools used outside of employer-sanctioned systems, continues to be widespread across organisations and raises concerns.  Employees may generate work using AI without guidance on ethical and legal issues, and without training on effective use or alignment with organisational style and standards. 

Where employers are considering implementing AI in the workplace, they should involve their workforce in that process. If employees are open about their use of generative AI, organisations are better placed to check for hallucinations and inaccuracies, maintain data protection compliance and avoid bias and IP infringement.  That, in turn, helps protect a business from both financial and reputational harm.

Governance 

The EU AI Act Regulation (EU) 2024/1689 was adopted on 13 June 2024 and comes into effect on 2 August 2026. The EU AI Act is a new law to ensure that AI systems placed on the EU market are safe, trustworthy and respect fundamental rights. 

The Act lays down uniform EU-wide rules: it defines which AI uses are banned, sets strict requirements for high-risk AI systems, mandates transparency measures (like labelling AI-generated content), and creates processes for market surveillance and enforcement.

Whilst the EU AI Act is an EU regulation, UK businesses may still fall within its scope if they provide AI systems used within the EU. Therefore, we will need to consider this before the Act comes into force this year. At present, there is no UK equivalent legislation governing AI use in the workplace. Instead, the UK continues to rely on existing laws and sector-specific guidance. The Government has also set out a cross-sector framework for regulating AI underpinned by five principles intended to “drive safe and responsible AI innovation”. 

  • safety, security and robustness;
  • appropriate transparency and explainability;
  • fairness;
  • accountability and governance; and 
  • contestability and redress. 

Organisations should be aware that greater regulation of AI use in the UK is inevitable. In 2023, the House of Lords introduced the Artificial Intelligence (Regulation) Bill.  In 2024, the TUC proposed the Artificial Intelligence (Regulation and Employment) Bill proposed by the TUC intended to regulate employers' use of AI, protect workers, and create enforceable rights in the Employment Tribunals. Whilst the TUC bill did not progress to Parliament, and the House of Lords Bill is making slow progress, the UK government is mindful of addressing AI as a national and global issue.  Employers would, therefore, be wise to prepare now rather than wait for a more comprehensive regulatory regime to arrive.

Evasion is Not an Option

Caution should be exercised by organisations given the infancy of many AI models, the pace of development, the ongoing risk of hallucinations (a risk that is likely to always be present), the concerns around bias and the lack of user understanding in many workplaces.

Where organisations are using AI models, they need to consider the risks in the workplace, such as inaccuracies, algorithmic bias, the impact on personal data and data protection compliance, as well as the potential infringement on intellectual property rights. 

For all employers, whether they have embraced generative AI or are approaching it with caution, the key is transparency.  Employers should be clear about AI use both internally and externally, and should provide employees with a clear direction on the use of generative AI models. 

All employers, regardless of size or sector, should have a policy in place. That policy should address acceptable use of AI, the handling of personal data and confidential information, the potential of unintended breach of copyright, and the need for human review. 

Any AI policy should also be supported by training. Workers should understand how to use AI effectively and appropriately, how to identify risks associated with AI use, and when human intervention is required to review AI-generated material for accuracy.

Because AI will continue to evolve rapidly, organisations will need to keep under review which AI tools are most effective and most suitable for their workplace. Policies and training should be revisited regularly and amended as required. 

In my view, evasion is simply not an option. Organisations should not regard AI models as a standalone option, but as systems operating in partnership with their workforce. The human element of review and balance, as a means of mitigating risk, must not be underestimated and must remain present even as AI services become more advanced and independent. 

Job Displacement

There is understandably caution, and in some quarters concern, about the use of AI in the workplace.  However, organisations would be well advised to focus on how generative AI can support their businesses through processes that remain human-led but technology-enabled. 

There is an ongoing narrative that AI is taking over human roles - a narrative endorsed by the announcement from Amazon that they were making 16,000 layoffs. PWC global chairman Mohamed Kande advised that the group’s five-year plan to hire 100,000 people would no longer happen, given that "When we made the plans (in 2021) to hire that many people, the world looked very, very different."  

This view was affirmed by Andrew Bailey, Governor of the Bank of England, who was recently quoted, “As you saw in the Industrial Revolution, now over time, I think we can now sort of look back and say it didn’t cause mass unemployment, but it did displace people from jobs and this is important,”…….My guess would be that it’s most likely that AI may well have a similar effect. So we need to be prepared for that, in a sense.”

Whilst roles that can be easily automated will be affected by AI, many of the recent losses attributed to AI are in response to slower post pandemic growth, less spending and a need for organisations to take cost control measures. In the case of the Amazon layoffs, it was noted that even prolific AI users were still laid off; however, automation is a more convenient narrative to explain mass workforce reductions and maintain market confidence rather than raising questions of financial stability.

AI is certainly reshaping roles within organisations. PWC has noted that whilst their hiring plans may have evolved since 2021, Kande also noted, "Now we have artificial intelligence. We want to hire, but I don't know if it's going to be the same level of people that we hire - it will be a different set of people."

As organisations reshape roles in response to AI, effective recruitment for those new and evolving roles will become increasingly important. Under the Employment Rights Act 2025, from 1 January 2027, the qualifying period for protection against ‘ordinary’ unfair dismissal will be reduced from two years to six months, and the cap on compensatory awards will also be removed. 

This means that any employee whose effective date of termination is on or after 1 January 2027 and who has six months' service will have the right to claim unfair dismissal at an employment tribunal. Given the enhanced administrative burden and financial risks of an employment tribunal claim, organisations must ensure they have robust recruitment processes and probationary periods in place so they can place the right people in the right roles.

How We Can Help

If you would like to discuss any of these issues or require professional assistance in developing practical guidance, AI policies and training to support your workforce, we are here to help.  Please contact Sophia Radford:

Telephone: 01258 488214

Email: sophia.radford@blanchardsbailey.co.uk

Locations: Blandford, Weymouth

Blanchards Bailey

So, how can we help?

Whatever your requirements, our team is standing by.

Call us today on
01258 459361