The Ethics of Code – Bias, Fairness, and the Workplace

As SMEs transition from viewing AI as a “shiny toy” to a core business engine, the focus shifts from what the AI can do to how it makes decisions. In Singapore, this isn’t just a philosophical debate; it is a fast-approaching legal requirement. With the Workplace Fairness Act (WFA) slated to come into force by the end of 2027, the era of “the black box made me do it” is coming to an end.

For the SME owner, the goal for 2026 is simple: build a “Fairness First” framework now, before the binding penalties of the WFA—which could reach up to $50,000 for serious breaches—become a reality.

The “Automation Bias” Trap

One of the greatest risks identified in Singapore’s 2026 Model AI Governance Framework for Agentic AI is “automation bias”—the human tendency to over-trust an automated system. If your HR team uses an AI screening tool that consistently filters out candidates from a specific demographic because of biased training data, your company is legally liable.

Under the WFA, “protected characteristics” include age, nationality, sex, race, religion, and disability. If an AI tool creates a discriminatory outcome in hiring, promotions, or performance reviews, the Tripartite Alliance for Fair and Progressive Employment Practices (TAFEP) will not look at the software vendor; they will look at you, the employer.

Navigating the Workplace Fairness Act (WFA)

The WFA represents Singapore’s first comprehensive legislative framework for workplace equality. Key implications for AI-adopting SMEs include:

• Merit-Based Proof: Organisations must be able to demonstrate that employment decisions are objective and documented. If an AI recommends a “Top 10” list of candidates, you must be able to explain the job-related criteria that led to that result.
• Internal Grievance Handling: The WFA mandates that every employer must have a written grievance process. If an employee feels an AI-driven appraisal was unfair, you must have a clear pathway for them to challenge it internally before it escalates to mandatory mediation.
• The 25-Employee Threshold: While the WFA initially focuses on firms with 25 or more employees, smaller firms are strongly encouraged to adopt these practices now, as the threshold is expected to be reviewed and potentially lowered as the ecosystem matures.

Implementing “Human-in-the-Loop” Oversight

The 2026 Governance Framework emphasises that humans must remain meaningfully accountable. This means SMEs should avoid “Autonomous HR.”

Instead, implement “checkpoints” where a human must review and sign off on high-stakes AI recommendations. For example, an AI might suggest a shortlist for a promotion, but the final decision—and the documentation of the reasoning behind it—must be executed by a manager who has audited the AI’s output for potential bias.

A Compliance Checklist for Responsible AI

To prepare for the WFA while maximising the Budget 2026 incentives, SMEs should:

• Vet Your AI Vendors: Don’t just ask if an AI tool works; ask for documentation on its “bias testing” and whether it was trained on datasets representative of Singapore’s diverse workforce.
• Conduct a DPIA: A Data Protection Impact Assessment (DPIA) is often the best way to uncover hidden risks in AI deployment, especially when the tool makes decisions that significantly affect individuals.
• Train Your Team on AI Ethics: Use the TechSkills Accelerator (TeSA) subsidies to ensure your HR and management teams understand how to spot and mitigate algorithmic bias.

Conclusion: Fairness as a Competitive Edge

In a tight labor market like Singapore’s, fairness isn’t just a legal obligation—it’s a talent magnet. By ensuring your AI tools are transparent and equitable, you build trust with your workforce and avoid the reputational and financial sting of a discrimination claim.

As we move into Article 5, we’ll see how these ethical and legal principles play out on the shop floor as we explore AI’s impact on specific sectors like the service sector.