This page summarises the limits that Turkish law sets on the use of AI in recruitment, in the management of the employment relationship and in its termination. It is based on the publications of Prof. Dr. Murat Volkan Dülger listed at the end. The legislation cited was checked against official sources on 1 November 2026.

Three areas, one common problem

AI is used at three points in working life: targeting job advertisements, screening applications and scoring candidates; allocating tasks, measuring performance and monitoring behaviour (algorithmic management); and ending the relationship on the basis of an alert or score produced by software. In all three, the common problem is that the employee or candidate does not know how the decision about them was made. They cannot challenge a decision they do not know, nor prove that they were discriminated against.

Algorithmic selection in recruitment

Recruitment software usually learns from the employer's past decisions; if those decisions were biased, the system reproduces the bias without anyone intending it. Under Turkish law, Article 5 of the Labour Act (Law No. 4857) prohibits discrimination in the employment relationship on grounds of language, race, colour, sex, disability, political opinion, philosophical belief, religion and sect and similar grounds, but contains a specific rule for the conclusion of the contract only for sex and pregnancy. Article 6 of Law No. 6701 on the Human Rights and Equality Institution of Türkiye prohibits discrimination in any work-related process, expressly including application, selection criteria and conditions of recruitment. The criteria that software uses to screen out candidates fall within “selection criteria”. Preventing a person from being hired on account of hatred is also an offence (Art. 122(1)(c) TCK).

The real problem is proof. The law eases the burden of proof in discrimination claims: if the employee or applicant shows facts that strongly suggest a violation, the other side must prove that there was none (Art. 5, final paragraph, Labour Act; Art. 21 of Law No. 6701). In algorithmic recruitment, however, the candidate may not even be able to take this first step, since they do not know by which criteria, and together with which groups, they were screened out.

New York City's Local Law 144 of 2021 requires recruitment software to undergo an independent bias audit before use and the results to be published. A 2025 report of the New York State Comptroller found that the city agency responsible for enforcement had identified only one violation among the 32 companies it reviewed, whereas the auditors found at least 17 potential violations in the same companies. This shows that a transparency obligation stays on paper without a body with the technical expertise to enforce it.

Algorithmic management and monitoring of employees

The limits of an employer's power to monitor employees were drawn by the European Court of Human Rights in Bărbulescu v. Romania [GC] (61496/08, 5 September 2017) and López Ribalda and Others v. Spain [GC] (1874/13, 17 October 2019): the employee must be informed in advance and clearly of the possibility and implementation of monitoring, the employer must have a legitimate reason, it must be considered whether a less intrusive method was available, and adequate safeguards must be provided. In E.Ü. [Plenary] (2016/13010, 17 September 2020), the Turkish Constitutional Court held that examining an employee's corporate e-mails without prior notice violated the right to the protection of personal data and the freedom of communication.

Turkish law has no specific rule on algorithmic monitoring, but existing provisions set clear limits. The employer must protect and respect the employee's personality (Art. 417 of the Code of Obligations, TBK). It may use the employee's personal data only insofar as they relate to the employee's suitability for the job or are necessary for the performance of the employment contract (Art. 419 TBK), and must use information obtained about the employee in good faith and lawfully (Art. 75 Labour Act). The employee may object to an adverse outcome resulting exclusively from automated analysis (Art. 11(1)(g) KVKK).

In Dülger's view, the words “insofar as necessary” in Article 419 TBK are the basic yardstick for algorithmic monitoring. A system that analyses an employee's location outside working hours, private correspondence or emotional state, or that seeks to predict trade union activity, clearly exceeds this limit. Because algorithmic monitoring not only collects data but also produces an assessment of the employee, proportionality should be reviewed more strictly than for e-mail or camera monitoring.

Dismissal based on an algorithmic decision

In its judgment of 4 April 2023 (ECLI:NL:GHAMS:2023:793), the Amsterdam Court of Appeal held that Uber's decisions to deactivate drivers' accounts on the basis of fraud alerts generated by software were based solely on automated processing within the meaning of Art. 22 of Regulation (EU) 2016/679 (GDPR), since the drivers had not been heard. According to the court, the fact that members of a risk team took notes on the alert did not show that all relevant data had actually been assessed.

Under Turkish labour law, the employer must give notice of termination in writing and state the reason clearly and precisely; before terminating for reasons related to the employee's conduct or performance, it must hear the employee's defence (Art. 19 Labour Act). For employees covered by job security, termination must rest on a valid reason arising from the employee's capacity or conduct or from the requirements of the business (Art. 18 Labour Act). In Dülger's view, a reason such as “the system has identified you as a low performer” is neither clear nor precise; an employer who cannot submit to the court the criteria of the system used and its output in the case cannot prove that the termination rests on a valid reason. Treating software output as sufficient on its own for dismissal would render the right of defence and the valid-reason requirement meaningless.

EU rules

The EU AI Act (Regulation (EU) 2024/1689) classifies as high-risk systems used for recruitment and selection, for decisions on terms of employment, promotion and termination, for task allocation and for monitoring and evaluating performance and behaviour (Annex III, point 4). Obligations for these systems apply from 2 December 2027. A deployer who is an employer must inform workers' representatives and the affected workers before using a high-risk system at the workplace (Art. 26(7)). Emotion recognition in the workplace has been prohibited since 2 February 2025, except for medical or safety reasons (Art. 5(1)(f)); applications analysing a candidate's emotional state in video interviews fall within this prohibition.

Directive (EU) 2024/2831 of 23 October 2024 on platform work prohibits digital labour platforms from using automated systems to process data on a worker's emotional or psychological state, on private conversations, or while the person is not working, and from predicting the exercise of fundamental rights such as freedom of association (Art. 7). Any decision to restrict, suspend or terminate a worker's account, or of equivalent detriment, must be taken by a human being (Art. 10(5)). Workers may obtain an explanation and request a review of decisions taken or supported by automated systems; the platform must reply in writing with reasons within two weeks (Art. 11). The transposition deadline is 2 December 2026.

Points for employers to consider

The main consequences of these rules are as follows. If AI is used in recruitment or in the employment relationship, candidates and employees should be told in advance. Employee data should be processed only insofar as necessary for the performance of the contract. Decisions such as dismissal, disciplinary measures and promotion should be taken, with reasons, by a person who has the knowledge and authority to depart from the system's recommendation. The employee's defence should be heard after the software output and the criteria behind it have been disclosed to them. The system's effect on different groups should be reviewed regularly and records kept; these records are the basis of proof in a discrimination claim or reinstatement action. Employers operating in the EU must also comply with the AI Act's prohibition of emotion recognition at work and inform workers' representatives.

Assessment

In Prof. Dr. Murat Volkan Dülger's view, AI may be used in working life; but the aim should be to control it rather than ban it, on the basis of four principles. First, the use of AI should be disclosed in advance to candidates, employees and employee representatives, and this duty should be written into the Labour Act. Second, decisions that significantly affect an employee's legal position, such as refusal to hire, dismissal, disciplinary sanctions and promotion, should not be taken without meaningful human review. Third, monitoring should be subject to proportionality; analysis of employees' emotional state, examination of private correspondence, collection of data outside working hours and prediction of trade union activity should be expressly prohibited by statute.

Fourth, the systems used should be regularly audited for discrimination. In Dülger's view, an employer who cannot submit the system's criteria and results to the Human Rights and Equality Institution or to a court in a discrimination case should be deemed unable to prove that it did not discriminate. Article 5 of the Labour Act should also be amended to cover discrimination at the hiring stage on all protected grounds and to attach compensation to it. As the New York experience shows, such auditing will not be effective unless carried out by a body with technical expertise.

The legislation and sources cited and their status for Türkiye

The legislation and sources cited on this page are listed below with their full titles and their legal status for Türkiye. For texts not in force in Türkiye, the list states why they are cited and in which cases they may nevertheless bind persons in Türkiye.

  • Turkish legislation: Labour Act, Law No. 4857 (Official Gazette 10.6.2003, No. 25134); Turkish Code of Obligations, Law No. 6098 (TBK; Official Gazette 4.2.2011, No. 27836); Law No. 6701 on the Human Rights and Equality Institution of Türkiye (Official Gazette 20.4.2016, No. 29690); Turkish Criminal Code, Law No. 5237 (Official Gazette 12.10.2004, No. 25611); Law No. 6698 on the Protection of Personal Data (KVKK; Official Gazette 7.4.2016, No. 29677). In force and binding in Türkiye.
  • European Court of Human Rights, Bărbulescu v. Romania [GC] (61496/08, 5.9.2017) and López Ribalda and Others v. Spain [GC] (1874/13, 17.10.2019): the European Convention on Human Rights is binding on Türkiye and, as a duly ratified treaty on fundamental rights, prevails over conflicting statutes (Art. 90(5) of the Constitution). These judgments against other states do not bind Türkiye directly, but are taken into account by Turkish courts and the Constitutional Court as interpretations of the Convention.
  • Turkish Constitutional Court, E.Ü. [Plenary] (2016/13010, 17.9.2020): binding in Turkish law; decisions of the Constitutional Court bind the legislative, executive and judicial organs, the administration and natural and legal persons (Art. 153(6) of the Constitution).
  • Amsterdam Court of Appeal (Netherlands), 4.4.2023, ECLI:NL:GHAMS:2023:793: a Dutch judgment; does not bind Türkiye. Cited to show how Art. 22 GDPR is applied in a work relationship.
  • New York City Local Law 144 of 2021 and New York State Comptroller Report 2024-N-6 (2.12.2025): a US local law and audit report; do not bind Türkiye. Cited to show the practical results of a transparency obligation.
  • EU AI Act: Regulation (EU) 2024/1689 of the European Parliament and of the Council of 13 June 2024. In force in the EU since 1 August 2024. Since Türkiye is not an EU member, it is not in force in Türkiye and does not bind Turkish authorities or courts. It does, however, bind providers and deployers in Türkiye that place systems on the market or put them into service in the EU, or whose systems' output is used in the EU (Art. 2(1)(a) and (c)). Because Türkiye is an EU candidate country and is required, within the Customs Union, to align its technical legislation on products, the Act is also cited as a model for Turkish legislation.
  • Digital Omnibus: Regulation (EU) 2026/1744 of the European Parliament and of the Council of 8 July 2026, amending the AI Act; in force in the EU since 27 July 2026. Its status for Türkiye is the same as that of the AI Act.
  • Directive (EU) 2024/2831 of the European Parliament and of the Council of 23 October 2024 on improving working conditions in platform work: binds only EU Member States (transposition deadline 2 December 2026); does not bind Türkiye. Cited as a model because it is the first EU text to regulate algorithmic management directly.

Related publications and pages

Status of the law: 1 November 2026. This page provides general information and does not replace legal advice on a specific case.