AI in the workplace: the legal framework in Georgia
Artificial intelligence at the workplace is no longer a purely technological matter: Georgian companies use such systems to screen candidates, evaluate employees, build schedules and monitor performance. The Labour Code of Georgia contains no dedicated chapter or article on artificial intelligence, but that does not create a legal vacuum. Everything an employer does — regardless of whether a decision is taken by a human or by a system — remains fully subject to the Code's general rules. On this page we explain the four groups of norms that AI adoption in the workplace most frequently collides with: the prohibition of discrimination, the rules on pre-contractual relations and the exchange of information about candidates, the rules on changing the terms of an employment contract, and the obligations surrounding confidential information.
The prohibition of discrimination and algorithmic decisions
The Labour Code defines discrimination as the intentional or negligent differentiation, exclusion or granting of advantage to a person on a protected ground, where the aim or effect is to deny or infringe equal opportunity or equal treatment in employment and professional activity. The list of protected grounds is broad: race, colour of skin, language, ethnic or social affiliation, nationality, origin, property or social status, status under an employment contract, place of residence, age, sex, sexual orientation, disability, state of health, religious, social, political or other affiliation (including trade-union membership), family status, political or other opinion, and any other ground. The concept of indirect discrimination is particularly important: a seemingly neutral provision, criterion or practice that places a person at a disadvantage compared with others because of one of these grounds constitutes discrimination unless it is objectively justified by a lawful aim and the means employed are necessary and proportionate.
This is precisely where the principal legal risk of artificial intelligence arises. A screening or scoring algorithm often looks neutral, yet its criteria — for instance a residential district, an age bracket or a contract status — may correlate with protected grounds and, as a result, produce indirect discrimination. Further statutory requirements apply alongside: where a woman and a man perform the same work, the employer must ensure equal pay for that work; harassment at the workplace, including sexual harassment, is a form of discrimination; and the Code prohibits the termination of an employment contract or any negative treatment of or pressure on an employee because the employee applied or complained to the relevant body for protection against discrimination or cooperated with it. These protections apply equally where the instrument producing the unequal outcome is an algorithmic tool.
The limits on candidate information during recruitment
The Labour Code regulates pre-contractual relations separately, and those rules speak directly to systems that collect and analyse candidate data. The employer is entitled to obtain only information about a candidate that is connected with the performance of the work and necessary to assess the candidate's ability to perform the specific job and to take the relevant decision — collecting anything beyond that frame, including by automated means, oversteps the statutory boundary. For their part, candidates must inform the employer of any circumstance that may prevent them from performing the work or endanger the employer's interests, and the employer may verify the accuracy of the information submitted.
The rules on access to this information matter just as much. Information obtained by the employer and information submitted by the candidate may not be made accessible to any other person without the candidate's consent, except in cases provided for by Georgian legislation. A candidate is entitled to demand back the documents submitted if the employer has not concluded an employment contract. The employer, in turn, must provide the candidate with information about the work to be performed; the form of the employment contract (oral or written) and its term (fixed or indefinite); the working conditions; the employee's legal status in the employment relationship; and the remuneration. The pre-contractual relationship ends when the contract is concluded or when the candidate is informed of the refusal to hire, and the employer is under no obligation to justify a refusal — yet the anti-discrimination rules remain fully in force at this stage: before an employment contract is concluded the employer must acquaint the candidate with the statutory provisions on the principle of equal treatment and the means of its protection, and must reflect the anti-discrimination provisions in internal labour rules, collective agreements and other documents and ensure their observance.
Changing employment terms when new technologies are introduced
When a new technology changes an employee's regime or functions, the rules on variation of the employment contract come into play. By mere notice the employer may only clarify individual circumstances of performing the work that do not change the essential terms of the employment contract; essential terms may be changed only by agreement of the parties, while a change in essential terms conditioned by an amendment of Georgian legislation requires no employee consent. The Code defines exactly what is not deemed an essential change: a change of the place of performance, provided the daily commute by available public transport between the employee's home and the new place of work takes no more than 3 hours and does not demand disproportionate costs; and a shift of the start or end time of work by no more than 90 minutes. A simultaneous change of both circumstances is deemed essential and requires consent. Separately, where the state of health of a pregnant, recently delivered or nursing employee does not, per a medical conclusion, allow the performance of the work envisaged by the contract, she may — within reasonable accommodation — demand work corresponding to her state of health with the same employer. Any restructuring of roles around AI tools must be mapped against these boundaries in advance.
Confidential information and its protection
AI systems tend to multiply the flow of information inside an enterprise, which makes confidentiality norms consequential. Under the Labour Code, employees' representatives and the expert supporting them may not disclose to an employee or a third party the confidential information supplied to them by the employer within the lawful interests of the enterprise; this prohibition operates without limitation in time and remains in force after the term of their authority expires. In parallel, the employer may refuse to provide information or to hold a consultation where doing so objectively and with justification significantly impairs the functioning of the enterprise or causes it damage; such a refusal may be challenged in court, and if the refusal is not objectively justified the court is empowered to order the employer to provide the information or to hold the consultation. Keeping this balance in view is essential when internal policies and training processes connected with AI systems are drafted.
How we can help
Our team assists with the legal assessment of artificial intelligence in the workplace: we explain which decisions must remain with a human, how the criteria built into algorithmic tools should be reconciled with the requirements of the Labour Code, and how to draft internal documents and rules so that the prohibition of discrimination is properly reflected and enforced. We also help streamline procedures for collecting candidate data and, where necessary, prepare the dispute. Contact us for a full review — we will assess your specific situation and propose practical solutions grounded in the operative norms of the Labour Code of Georgia.