The intersection between AI and the Workplace Fairness Act

Is it possible to argue that there may be a policy interest to ensure that companies continuously adopt new technologies to renew their workforce?

The discussions amongst human resource practitioners earlier this year may have been dominated by the passing of the workplace fairness bill by Parliament on 8 January 2025.

In brief, the Workplace Fairness Act 2025 (WFA) aims to protect individuals from discrimination on the ground of protected characteristics and to establish fair employment practices. There is also the recognition through the WFA that Singapore citizens and permanent residents form the core of the workforce in Singapore.

However, as 2025 creeps closer to an end and the anticipation with respect to the effective date of the WFA starts to build up, the conversations seem to turn towards how the use of artificial intelligence (AI) may have implications when viewed from the WFA perspective.

Having attended the recent People Behind People Forum 2025, organised by the Institute for Human Resource Professionals, there was substantial focus turned towards the intersection between AI and the impact on employment and workforce which can be broadly viewed from many perspectives.

However, what is of interest though is how the use of AI can trigger broader questions around compliance with the provisions of the WFA.

This article is not intended to be a comprehensive review of the WFA and the implications of AI but hopefully through the points raised in this article, it can assist to raise awareness of the potential issues that may creep up unknowingly.

Generally, we can categorise the issues into two categories – when AI is used in the displacement of the workforce and when employment decisions are made using AI.

Decision to use AI in the displacement of workforce
In the former category, the main concern would arise on the very decision to use AI in place of the existing workforce.

Section 17 of the WFA states that there is discrimination against an individual when an employer makes an employment decision that adversely affects an individual. Nothing in that provision provides that the adverse decision made against an individual must be made in favour of another individual or person.

Applying section 17 literally to a situation where an employer decides that AI can be used to replace the existing workforce on the basis that the current employees are old or will have too many conflicting responsibilities such as caregiving responsibilities, that decision in itself can be adverse to the current group of employees and accordingly may fall foul of Section 17 of the WFA.

In this case, even though the employer has not actually discriminated against any of its current employees in favour of employees who are younger or have lesser responsibilities, this is not of concern since section 17 does not provide for that as a requirement to prove discrimination.

However, is it possible to argue that there may be a policy interest to ensure that companies continuously adopt new technologies to renew their workforce? And should the current group of employees refuse to adopt or learn new skillsets to operate the AI systems in place of their current role, then even though they are of protected characteristics, the decision exercised adversely against them should still be permitted to stand?  

If we look at the exceptions under Section 20 of the WFA, it does provide a list of different exceptions but each of the exceptions seems to fall on whether there is a genuine requirement for another individual that does not have the protected characteristic to perform the role. In this case, it seems that AI, which is not an individual, does not therefore fall within the exception.

When using AI as part of the employment decisions
Another area of concern that we need to grapple with relates to the interpretation of “employment decision.” Employment decision is defined under section 2 of the WFA to mean an employment decision mentioned in sections 5, 6 or 7 of the WFA and each of sections 5, 6 or 7 of the WFA refers to employment decision as the decision made by an employer.

Employer is again defined under section 2 of the WFA to mean any person who employs or intends to employ another person under a contract of service with an employer. If a company decides to rely on AI to make the employment decision or relies substantially on AI to make the employment decision, is that considered a decision of the employer since the AI is not a person?

In a case where the company decides to send certain employees for training and in doing so, uses an AI tool to merely assist the company to come up with an initial list but the company eventually decides the final group of employees to be sent for training. In such a situation, as the company, being the employer, makes the decision, it is clearly within Section 6(1)(d) of the WFA if certain employees were being discriminated against.

Let’s consider another situation where the company decides to deploy an agentic AI that will automatically decide which employees should be considered for training and also send training invites directly to them without anyone from the company making the employment decision.

In this new scenario, if the agentic AI decides that only certain male employees should receive training, can it be said that section 6(1)(d) of the WFA may have been breached but in this case, can it be said that the company has indeed made the employment decision?  

It is arguable that agentic AI is deployed by the company, hence the company, as the employer should remain liable for the decision made by the agentic AI, or does it?

As mentioned, the above is not intended to be conclusive in terms of the issues discussed but hopefully it is a reminder to companies that when they start considering the deployment of AI in their employment process, they should also start to consider its implications under the WFA.