Women and the Labour Market: Navigating Algorithm Decision Making

María Camila Salazar-Larsen — Mon, 06 Jul 2026 11:23:55 +0000

AI is reshaping the labour market, influencing operations across organisations, as Eurostat shows. The ILO has stated that women are more likely than men to be affected by automation. As a matter of fact, the impact of AI in the workplace extends beyond job displacement. It is then suspected that AI can reinforce existing gender gaps. Empirical studies have shown that the bias and stereotypes that arise from the caregiving responsibilities that have been historically assigned to women, hinder their access to the labour market since employers believe women will be more absent from work, as they are assumed to prioritize their family (Ramírez-Bustamante et al, 2024). One could ask, how can existing gender gaps be reinforced by AI? Algorithms now influence multiple aspects of employment relationships, including recruitment. Take for example the case of X, a woman who applied for a job. X has taken several leaves to take care of her family. Now, due to inconsistent gaps in her CV, she was rejected in the automated CV-screening process. Therefore, data‑protection frameworks become essential to ensure that women’s data is not used to hinder access to the labour market The present account argues that the failure to treat “sex”-related information as sensitive data within data‑protection frameworks creates a barrier to women’s access to the labour market. Data‑protection rules must therefore be developed through a law‑in‑context perspective that recognises the historical obstacles women have faced in accessing employment. Comparative examples show that defining sensitive data according to the potential impact of its use, rather than through exhaustive lists, may offer stronger protection for women.

Data-protection regimes should be analysed within the broader legal framework in which they operate. In the European context, both primary law (TEU, TFEU, CFREU) and secondary legislation (Directives 2000/78/EC, 2006/54/EC, 2024/1500/EU) establish gender equality as a core principle of the Union. It is within this landscape that the General Data Protection Regulation (GDPR) operates. Nonetheless, gaps remain: Article 22 GDPR permits automated CV‑screening, and Article 9 GDPR does not classify “sex” as a special category of data, possibly indirectly enabling gender‑based discrimination in algorithmic recruitment.

Under Article 22 GDPR, organisations may rely on automated decision‑making when necessary to enter an employment contract. The provision prohibits basing such decisions on special categories of personal data like race or trade‑union membership, unless explicit consent is provided. However, Article 9(1) GDPR does not explicitly include “sex” within its list of special categories of data. The E uropean D ata P rotection B oard confirms that this list is exhaustive, meaning additional categories cannot be added by interpretation. As a result, “sex”-related information could be processed without explicit consent in automated CV‑screening scenarios.

In this context, an organisation cannot use algorithms to process data on race or trade‑union membership to decide whether to invite a candidate to an interview without explicit consent. Yet it may process “sex”-related data, such as motherhood status, age, or CV gaps, to make the same decision. Research shows that algorithms and generative AI tools reproduce biased language (Jiang, 2021) and outdated gendered skill classifications (Imana, Korolova & Heidemann, 2021; Jungyong, Jungwon & Yongjun, 2022; Gorska & Jemielniak, 2023). For instance, the University of Washington (2024) discovered that when using AI tools to screen CVs, the systems preferred male-associated names 52% of the time versus female-associated names 11% of the time. While Article 6 GDPR principles such as purpose limitation and data minimisation offer some protection, persistent structural gender stereotypes and bias suggest that “sex”-related data should receive the heightened safeguards afforded to special categories.

Other legal systems in the Global South offer alternative approaches. In Colombia, “sex” is not explicitly listed as sensitive data under Statutory Law 1581 of 2012. However, the Constitutional Court in Judgment C‑748/11 confirmed that the list of sensitive data is not exhaustive. Statutory Law 1581 of 2012 defines sensitive data as information that may affect privacy or expose individuals to discrimination. Then the focus is on the impact of the data rather than its formal classification.

In today’s labour market, gender continues to shape access to opportunities. Data‑protection frameworks should therefore adopt a gender‑sensitive approach to mitigate the barriers faced by women. Colombia’s impact‑based model may offer stronger protection, as organisations must assess whether “sex”-related information could function as sensitive data and thus require heightened safeguards. On the other hand, the EU framework excludes “sex” from the outset, reducing the scrutiny applied to such data under the GDPR.

Note: This blog post has been written as a result of the research conducted for the book “Gender, Sexuality and Law: South-South Perspectives” which will be published by Springer in 2026 and contains the full law-in-context analysis.

María Camila Salazar-Larsen | OHRH

Women and the Labour Market: Navigating Algorithm Decision Making