Investigators conducting workplace investigations are increasingly experiencing the many uses of generative AI (“GenAI”) in investigation processes.

With many using GenAI applications such as ChatGPT or Copilot as in-palm life coaches and personal assistants, it is no wonder that they are also using GenAI to draft their responses to investigations.

Some leading technology organisations have mandated the use of GenAI and shifted the dial from employees asking “can AI be used?” to requiring employees to ask “why can’t AI be used?” The mandatory use of GenAI is not limited to technology organisations, with retailers rolling out GenAI applications to improve in store experiences for both consumers and staff.

How is GenAI being used by employees in investigations?

Against this backdrop of personal GenAI usage, it is not surprising that employees are using GenAI during workplace investigations including at the very earliest stages of drafting a complaint and in the later stages of drafting responses to allegations.

Employees may also use GenAI to give them guidance on the approach they should take to interviews with investigators.

What are the issues with employees using GenAI?

Because GenAI applications can have a tendency to “hallucinate”, investigators need to be aware of the potential for recollections or assertions to be fabricated. The use of GenAI in the context of workplace investigations is particularly problematic due to the need for any workplace investigation to be accurate and seek out reliable information.

Despite these warnings about GenAI hallucinations, we have even seen lawyers list cases in submissions to the Court, where the cases were hallucinated by GenAI applications.

Various Courts including the Federal Court of Australia, Supreme Court of New South Wales, Supreme Court of Victoria and NSW Industrial Relations Commission have developed, or are in the process of developing, guidelines or practice notes on the use of GenAI in Court proceedings. Generally, the approach taken by the Courts is that they have:

• identified a potential issue with confidential information being put into GenAI platforms without knowing where it could end up;
• prohibited or advised particular caution on GenAI use for witness statements or affidavits because those documents should reflect the witness’ own knowledge;
• prohibited GenAI use to paraphrase a witness’ evidence because it could embellish, strengthen or dilute a witness’ recollections or evidence; and
• in some jurisdictions made it a requirement that witnesses declare that they have not used GenAI in the preparation of their witness statement or affidavit.

What does this mean for investigators?

In the context of investigations, investigators need to be aware of the potential for GenAI hallucinations, embellishments, dilutions or alternations. Similarly, investigators also need to be mindful of the potential effect of GenAI on both written and oral evidence given by witnesses to an investigation.

How can organisations mitigate the use of GenAI in workplace investigations?

Practical steps which investigators may take either at the outset of an investigation, or once it has been detected that GenAI may have been used by any participant, include:

• warning participants that GenAI has the potential to hallucinate and to materially change an individual’s recollection;
• making it clear to participants that they should only rely upon their own knowledge; and
• warning participants that they are ultimately responsible for the statements which they provide to the investigation and if they have been found to have fabricated any statement or misled the investigation, the use of GenAI will not be a valid excuse for this serious misconduct.

Depending on an organisation’s policy on GenAI use, some more stringent steps include:

• directing participants not to use GenAI to produce any part of any written materials that they submit about their own knowledge;
• making it clear to participants that they may use GenAI to proof for spelling and grammar, but not to make any alterations to or paraphrase the content of any written statements; and
• requiring participants to confirm in any written statement they provide that they have not used GenAI to produce any of the content (other than for spelling and grammar review).

What if an organisation requires the use of and promotion of GenAI?

Where possible, it would be optimum to exempt workplace investigations from any mandatory use of GenAI. This would protect the purpose of an investigation, which is to make reliable findings about whether allegations are substantiated. The reliability of findings is heavily dependent upon the reliability of participants’ information.

If an exemption is incompatible with an organisation’s policy on GenAI use, it will be critical to set out at the beginning of any investigation process the potential issues with GenAI use, and that ultimate responsibility for the content of any response or information will rest with the employee producing the document.

If you are interested in an overview of investigation processes generally see PCS’ contribution for Australia to the most recent International Employment Lawyer’s Guide to Workplace Investigations.