AI In the Workplace: Employment Law Risks Employers Should Be Aware Of

AI-powered recruitment tools are often seen as a way to make decisions that are objective and based on data.

There are many situations where AI can help during the recruitment phase, such as:

• Identifying the skills, qualifications and experience needed for a role and creating job descriptions to source suitable candidates.
• Screening candidates.
• Scanning and sifting CVs and application forms to prioritise applications using certain keywords.
• Administering assessments and analysing results. 

For an employer receiving high volumes of applications, this is very attractive. However, the further AI moves from administrative assistance towards selection and decision-making, the greater the legal risk. 

AI algorithms can easily pick up and perpetuate biases. For instance, if a company’s workforce lacks diversity, that will be reflected in the data used to feed into the AI software.

This lack of diversity will then be reproduced in the results, creating historic inequality without anybody programming an instruction to discriminate.

Additionally, a recruitment tool that attaches significance to career gaps may disproportionately disadvantage women who have taken maternity or caring-related career breaks or candidates that have had periods of sickness due to a disability. 

Under equality legislation, even if a decision is automated, liability still rests with the employer. 

Employers will also need to consider risks from a data protection point of view. 

AI tools may process considerable amounts of candidate information, including special category data.

Employers need to think about:

• What data the AI system actually needs?
• How long will this data be retained for?
• Whether the data is being used for a different purpose from that for which it was collected?
• Whether the privacy information provided actually explains what is happening with this data?
• What is the legal relationship between the employer and the supplier? 
• Whether solely automated decisions are being made?

To manage and mitigate these risks, there needs to be meaningful human involvement. AI can be incredibly useful but it needs to be used appropriately.

Any recommendations given by an AI tool, need to be considered by someone who genuinely understands the purpose of the tool, the data it is using and how, is able to test it, monitor its outcomes and use their own judgment to challenge the conclusion where necessary.

Simply relying on AI to explain why a candidate was not offered a job will not be sufficient.  

AI is being used a lot more frequently in everyday life and, as a result, this is spilling into the workplace. 

When AI is being used by employees for work purposes, it creates risks relating to confidentiality, data protection, intellectual property, information security and misconduct that employers have not historically had to consider. 

However, it’s clear that AI is here to stay and so businesses will need to move with these developments in a carefully controlled and managed way. 

Employers allowing AI, including Generative AI, use in the workplace should:

• Assess AI use by their employees in the workplace and consider what they can, should or should not be using it for.
• Create a clear policy setting out the rules around its use. In particular, this should address:
  
  • The circumstances in which AI can be used;
  • Restrictions on sharing data;
  • Which AI tools are approved or prohibited;
  • What outputs require human checking and who remains accountable for the end product;
  • If applicable, how the employer will be developing Generative AI technology itself or incorporating AI into its own services or products; and
  • How any contractual arrangements with third parties will address liability arising from the use of AI. 
• Train employees on the use of AI. It’s important that training is practical and not theoretical, focuses on real workplace scenarios, is regularly updated (particularly in light of the speed that AI is developing) and emphasises the need for human oversight and review.
• Make staff aware of the consequences of a breach of the AI policy and rules.
  Employers also need to be mindful that employees are increasingly turning to AI tools to draft formal grievances, disciplinary submissions and appeals. This can result in lengthier correspondence which, in some cases, lacks clarity or focus and becomes time consuming to deal with. In these circumstances, it is often beneficial to invite the employee to a meeting to ensure their real concerns and objectives are clearly understood.