Recruitment Tools and the Legal Exposure They Carry

AI is increasingly used to screen job applications, score candidate resumes, and even predict hiring success. For a small business, recruitment tools that use AI can reduce the time spent reviewing applications. However, employment law, data protection law, and discrimination law create significant legal exposure. A recruitment tool that screens out candidates based on patterns in its training data might screen out candidates because of their age, gender, race, disability, or other protected characteristic. When this happens, the business faces discrimination complaints and regulatory investigation. The tool does not protect you. You are liable.

Before you use any AI recruitment tool, you must understand what the tool is actually doing. Many vendors describe their tools as “objective” or “bias-free.” These claims are often not true or are not well-tested. A tool might claim to be objective because it does not explicitly look at gender or age, but it might look at variables that are correlated with these characteristics. For example, a resume screening tool might down-score resumes with gaps in employment, which correlates strongly with career breaks for parenting. The tool might down-score candidates who learned their skills outside formal universities, which correlates with older workers who trained through apprenticeships. These patterns create discriminatory outcomes even though the tool did not explicitly consider age or family status.

Before you use a recruitment tool, ask the vendor for evidence that the tool has been tested for discrimination. Ask what variables the tool considers. Ask whether the vendor has evaluated the tool on candidate populations from different demographic groups to ensure the tool treats them fairly. Many vendors cannot provide this evidence. This is a red flag. A vendor who cannot show that they have tested for discrimination bias is asking you to accept discrimination risk you cannot verify. The safest approach is to avoid tools from vendors who cannot demonstrate fair testing.

For employment screening, consider whether AI adds value over human judgment. If you have three candidates for a junior role and it takes forty-five minutes to review their applications, is an AI tool really saving you time? If you have a hundred candidates, the time saving is real. At scale, the efficiency gain justifies deeper evaluation of the tool’s fairness. For small businesses hiring occasionally, human review is often both faster and legally safer.

If you do use an AI recruitment tool, you must disclose this to candidates. Data protection law requires that you tell job applicants that their data is being processed by an automated system and that you tell them about any automated decision-making that significantly affects them. If an AI tool screens out a candidate’s application without human review, the candidate has a right to know this. You must provide this information in your job posting or in your recruitment process. Many candidates are not comfortable having their applications screened entirely by machines, and transparency about this may cost you some good candidates. But it also protects you legally and demonstrates good practice.

Your recruitment process should not rely entirely on AI. Even if an AI tool screens applications first, a human must review the top candidates or all candidates who pass a certain threshold. The AI is a tool to reduce volume. The human makes the decision. This is not just better practice. It is legally necessary. Article 22 of the UK Data Protection Act and similar provisions in other data protection laws prohibit entirely automated decisions about individuals in certain contexts. Employment decisions are a context where you need human involvement.

Document your recruitment process. Write down what tool you use, how it is configured, and how you validate its decisions. If you use an AI tool and receive a discrimination complaint, your documentation will show whether you took steps to test the tool for bias and whether you involved humans in decisions. This documentation is your defence. Without it, you look like you delegated hiring decisions to a black box, which is the worst possible narrative legally.