The take up of AI tools within the workspace continues to grow exponentially. The benefits of these tools are significant, with the ability to massively improve efficiency and, sometimes, the quality of your output. However, there are significant risks for employers in the use of these tools which means it is essential that you understand these risks and how AI is being used in your workplace and develop guidelines to assist in their management.

Common Risks of Using AI

Some risks in using AI in your workplace are:

• ‍Protection of confidential and personal information. How the information you input into an AI tool is used and stored differs from platform to platform. For example, your information may be retained and used to train the platform's large language model. This can result in unauthorised disclosure and use of customer or client information. Consequently, it is important that you conduct due diligence on any platform that is used within your business, carefully reviewing their terms and conditions to verify how your prompts and uploaded information is stored and used or whether any rights are granted to the platform in relation to the output generated.
  ‍‍

• AI Generated Output and Intellectual Property Issues. There has been a raft of litigation in various jurisdictions in relation to whether traditional intellectual property rights can apply to AI generated output. For inventions created using AI, the benefit of patent and registered design protection has, to date, been universally rejected on the basis that such protection can only be afforded to natural persons.
  
  Copyright protection has also largely been rejected for AI generated outputs. However, there are some exceptions, such as in the UK, New Zealand, Hong Kong and Ireland, with these countries having provisions in their copyright legislation which provides for protection of computer generated works (though the extent of this protection is currently unclear).
  
  Trade mark protection, on the other hand, is available to AI generated content.
  
  The key takeaway is that you will likely not have the usual IP protections available for outputs generated by AI. As an example, if you have developed source code for one of your online service offerings using AI, a third party who gains access to this code may be able to use this code without fear of copyright infringement. If it is disclosed, it effectively becomes open source. This could in turn affect the perceived value of this asset if it is ever included as part of a business sale, due to the lack of proprietary and protectable rights subsisting within.
  ‍‍

• Work products are not carefully checked and validated. With increased reliance on, and confidence in, AI generated output, the rigour applied to checking and validating the accuracy of that output can wane. There have been numerous high profile cases of AI generated work products not being carefully checked or validated (for example, lawyers referencing fake cases in court submissions based on output prepared by AI). The obvious embarrassment and reputational risk aside, this can also result in breaches of law and contract giving rise to claims for damages or other losses.
  
  The other side of this issue is when business decisions are based on output generated by AI. Just like with humans, AI can present bias depending on the prompts provided or source material used to generate the content. It is important for staff to be aware that the quality of the output will sometimes only be as good as the prompts and inputs provided, leading to output that reflects the users desired position (as opposed to an objective review) or repeating mistakes resulting from these inputs.

Analyse how AI is being used in your business

A business does not need to have formally introduced AI for employees to be using it. Before creating guidelines or a framework for this use, employers should understand:

• which tools staff are using and what terms and conditions apply to those tools;
• what tasks they are using them for;
• whether personal or confidential information is being entered;
• whether AI-generated work is being checked;
• how AI generated content is being used by employees, whether to make decisions or as work product for customers or clients .

An open discussion with employees can help identify where AI is already being used and where clearer boundaries may be needed.

What should an AI policy cover?

A workplace AI policy should be clear and practical. It may include:

• which tools employees may use;
• appropriate and prohibited uses;
• information that must not be entered;
• when human review is required;
• rules for recruitment and employment decisions;
• accuracy and source-checking requirements;
• copyright and ownership considerations;
• when AI use should be disclosed;
• what to do if information is entered by mistake; and
• training and review requirements.

The policy should be easy to find, understood by employees and reviewed as new tools are introduced. In addition, there should be active training and education introduced around AI use so that your staff understand the risks and they are kept front of their mind when using these tools.

How McVeagh Fleming can help

McVeagh Fleming can assist employers with workplace AI policies, privacy obligations, employment processes, confidentiality requirements and intellectual property concerns.

Getting the right framework in place can help employees use AI confidently while making it clear where human review and approval are still required.