Can Employees Use ChatGPT at Work? What Irish Employers Should Consider
Artificial intelligence (AI) is becoming increasingly common in Irish workplaces. Employees are using tools such as ChatGPT to draft emails, prepare documents, summarise information, analyse data and assist with everyday tasks.
For employers, the question is no longer simply whether employees are using AI. In many companies, they already are!
The more important question is whether employees understand how AI can be used safely and appropriately in the workplace and whether employers have appropriate controls in place.
This issue has become particularly relevant following the publication by the Workplace Relations Commission (WRC) of specific guidance on the use of AI in WRC cases.
What is the WRC saying about AI?
In recent months, the WRC published Guidance for Parties on the Use of AI before the WRC.
The guidance applies to anyone involved in a WRC case who uses AI tools to prepare written submissions or evidence. The WRC recognises that AI can be useful for organising ideas, improving grammar and assisting with drafting. However, it also highlights the risks associated with relying on AI generated material.
In particular, the WRC warns that AI tools can produce information that is incorrect or misleading, including in relation to Irish employment and equality law. Parties using AI remain responsible for ensuring that material submitted to the WRC is accurate and reliable.
This is an important message for employers:
Using AI does not transfer responsibility for the information produced by AI to the technology.
If an employer, employee or representative relies on AI generated information in a workplace dispute, they remain responsible for checking that information before relying on it.
The WRC's guidance is available on its website: Guidance for Parties on the Use of AI before the WRC
What are the risks of employees using AI at work?
AI can offer significant benefits to employers and employees. However, unrestricted use can create a number of HR, legal and data protection risks.
For example, an employee might enter information about a colleague into an AI tool to help draft an email or report.
That information could include:
- Personal information.
- Salary or payroll information.
- Medical or absence information.
- Performance concerns.
- Disciplinary or grievance details.
- Investigation notes.
- Customer or commercially sensitive information.
Employees may not always understand what happens to information once it is entered into an AI tool or whether they are permitted to use the company's information in that way.
Employers therefore need to consider not only whether employees can use AI, but also what they can use it for and what information they can enter into AI systems.
Can an employee use ChatGPT to draft an email?
In many circumstances, there may be no issue with an employee using an AI tool to assist with a routine piece of work.
For example, an employee may use AI to improve the wording of a general business email or help structure a presentation.
The risk increases where the employee provides confidential or personal information to the AI tool in order to generate the content.
Employers should therefore consider establishing clear rules around acceptable AI use rather than taking an approach of either allowing everything or prohibiting AI completely.
A practical AI policy might distinguish between:
Low risk uses, such as:
- Brainstorming ideas.
- Improving general grammar.
- Creating generic templates.
- Summarising non confidential information.
and higher-risk uses, such as:
- Processing employee information.
- Making recruitment decisions.
- Assessing employee performance.
- Drafting disciplinary or grievance documentation.
- Analysing confidential business information.
- Making decisions about employees.
What about HR using AI?
HR teams may have particularly strong reasons to use AI, but they also handle some of the most sensitive information within a company.
For example, HR professionals may be tempted to use AI to help draft:
- Disciplinary letters.
- Investigation reports.
- Grievance responses.
- Performance improvement plans.
- Recruitment communications.
- Policies and procedures.
- Meeting notes.
While AI can assist with drafting, employers should ensure that confidential employee information is not entered into an AI tool without appropriate consideration of data protection, security and company policy requirements.
AI should support HR decision making rather than replace professional judgement.
An AI generated disciplinary letter, for example, should not simply be copied and issued to an employee without an HR professional reviewing the content, checking the facts and ensuring that the appropriate process has been followed.
What happens if AI gets it wrong?
One of the biggest risks with generative AI is that it can produce information that sounds convincing but is incorrect.
This is particularly important in HR, where an inaccurate statement about employment legislation could have significant consequences.
The WRC's recent guidance specifically highlights this issue. AI generated material may contain inaccurate information, including incorrect references to legislation or legal authorities. Remember, parties remain responsible for the accuracy of what they submit.
This means employers should never assume that information generated by an AI tool is legally correct simply because it is presented confidently or appears professional.
The same principle applies outside the WRC.
If an employer uses AI to help draft an HR policy, employment letter or disciplinary correspondence, the employer remains responsible for ensuring that the final document is accurate and appropriate.
AI and employment decisions
There is another important issue for employers to consider, using AI to make or support decisions about employees.
The EU AI Act identifies a number of AI systems used in employment, recruitment and worker management as high risk. These can include systems used to screen applications, rank candidates, monitor performance or support decisions relating to employment whether it be offering employment, promotion and career progression.
This means employers need to be particularly careful if AI is being used to influence decisions such as:
- Who is shortlisted for interview.
- Who receives a promotion.
- How employee performance is assessed.
- How work is allocated.
- Who may be selected for redundancy.
- Whether an employee is considered suitable for a particular role.
An employer should not assume that it can avoid responsibility simply because an AI system made the recommendation.
Transparency, human oversight and appropriate governance remain critical.
Should employers ban ChatGPT?
For most companies, a complete ban may not be the most practical approach.
Employees are increasingly familiar with AI tools, and preventing their use entirely may simply result in employees using them without informing the employer.
A better approach may be to establish clear expectations around acceptable use.
An AI policy could explain:
- Which AI tools employees are permitted to use.
- What information must never be entered into an AI tool.
- Whether employees need approval before using AI for certain tasks.
- Whether AI generated content must be checked before being used.
- When human review is required.
- Who is responsible for monitoring compliance.
What should employers do now?
Employers should consider reviewing how AI is currently being used within their company.
This does not necessarily require a major technology project. A good starting point is simply to understand what employees and managers are already doing.
Employers should consider:
- Identify where AI is already being used across the company.
- Establish clear rules around acceptable AI use.
- Protect confidential and personal information from inappropriate use in AI tools.
- Review HR processes where AI may be influencing employment decisions.
- Train employees and managers on the appropriate use of AI.
- Ensure AI-generated information is checked before it is relied upon.
- Consider data protection requirements before introducing new AI tools.
- Maintain appropriate human oversight of employment decisions.
- Keep records of important employment decisions and the reasons for those decisions.
Employers should also keep their AI policies under review as the technology and regulatory environment continues to develop.
The WRC has now established a role under the EU AI Act in relation to AI systems used in the Irish labour market, including systems used in employment and worker management.
AI is a tool – not the decision maker
The increasing use of AI does not change the fundamental responsibility of employers to manage employees fairly, lawfully and appropriately.
AI can be an extremely useful tool for improving efficiency and supporting HR teams. However, employers should be cautious about allowing AI to make decisions that affect employees without appropriate human oversight.
The recent WRC guidance provides a useful reminder of the principle:
If you use AI, you remain responsible for the outcome.
As AI becomes more embedded in the workplace, employers who establish clear expectations, provide appropriate training and put sensible safeguards in place will be better positioned to benefit from the technology while managing the associated risks.
How MSS The HR People can help
AI is developing rapidly, and many employers are still working out how it should be managed within their company.
MSS The HR People can assist employers with:
- Developing an AI workplace policy.
- Reviewing existing HR policies in light of AI use.
- Advising on the use of AI in recruitment and HR processes.
- Developing guidance for managers and employees.
- Supporting employers with WRC cases and employment related matters involving AI.
If your company is already using AI, or you are considering introducing AI into your HR processes, now is a good time to review your approach and ensure appropriate safeguards are in place.
For support on any of the above, don’t hesitate to contact MSS The HR People at info@mssthehrpeople.ie or call 1 887 0690.
