AI in the Workplace

Employers are turning to algorithmic tools in hiring, performance and workplace monitoring. Under Turkish law the question is no longer whether those tools raise productivity, but whether they can be used defensibly under the equality principle, the KVKK and the mandatory rules of the Labour Code.

Systems trained on historical data can produce indirect discrimination. Labour Code art. 5 and the Human Rights and Equality Institution Act forbid discrimination in employment; the black box makes proof harder. Consent is often a weak legal basis given the imbalance between employer and employee. Biometric timekeeping may fail a proportionality test; the right to object to fully automated decisions requires meaningful human oversight of important employment decisions.

Liability cannot be shifted onto the vendor. Where a discriminatory result or a data breach occurs, the employer is as a rule the addressee. Staff pasting client information or trade secrets into external models also engages confidentiality and the cross-border transfer regime. The EU AI Act treats employment systems as high-risk; firms serving the EU market may already face those standards as a compliance duty.

The note sets out concrete steps on vendor diligence, contractual allocation, human oversight and governance. The full text is below.