French employment Law: AI week

Risks of using AI in the recruitment process

In recent years, AI has become a popular tool in the recruitment process. While it offers undeniable advantages, such as streamlining initial candidate interactions and automating certain aspects of the hiring process, it is essential to consider the potential risks associated with its use.

Using AI in the recruitment process could damage the diversity, equity and inclusion efforts made by companies.AI models are trained on large datasets which may inadvertently contain biases from historical hiring practices. As a result, it can inherit these biases and make unfair judgments or recommendations.

In addition, the AI system may inadvertently favour candidates with certain demographic characteristics, leading to discrimination in the hiring process.

One good example to illustrate this point is the AI tool built by Amazon, which was supposed to help sort through mass applications and find top candidates, which it did. However, its references were based on resumes submitted to the company over a 10-year period, which were largely from men, as was common in the technology industry at that time. Hence, Amazon’s system taught itself that male candidates were preferable and it automatically penalised any resume that included the word “woman”.

Despite Amazon’s efforts to edit the tool, there was no guarantee that it was completely gender neutral as long as the datasets themselves were historically biased.

It is important to remember that, in France, article L.1132-1 of the French Labour Code prohibits discrimination during the hiring process. If there is discrimination on the basis of origin, gender, age, sexual orientation, religious, philosophical or political convictions, employers could face up to 3 years’ imprisonment and a €45,000 fine (article 225-2 of the French Criminal Code).

The biggest issue with discrimination resulting from an AI system is that it is automatic and often unintentional on the part of the employer. This leads to a potential large-scale number of claims and easier means of proof, without the employer even being aware of such risk beforehand.

Candidates who believe that they have not been hired because they were discriminated against due to a biased AI model can also ask the judges to put an end to such measure and seek compensation for the damage they have suffered.

On top of this issue, using AI tools during recruitment raises the matter of the processing of personal data using AI. It may result in the leak of confidential information of the company, as well as of the candidates, as discussed in our previous article in this series.

The significant benefits of AI tools are undeniable in terms of efficiency and scalability. They should nevertheless be managed with caution, in order to avoid any litigation related to their use.