AI Ban in Workplace Rules Can Lead to Dismissal, But Not Automatically

A growing number of Polish companies are writing bans on public artificial intelligence tools into their workplace regulations, and employment lawyers warn that breaking such a ban can, in extreme cases, lead to summary dismissal. At the same time, they caution that simply using ChatGPT against a supervisor's instructions is not, on its own, enough for that dismissal to hold up in labor court.

The starting point is a simple question that more and more HR departments are asking: can an employer even ban staff from using tools like ChatGPT, Claude, or Gemini. Lawyers answer unambiguously: yes, under the managerial authority employers hold over the organization of work as a whole.

Legal basis for the ban

Article 100 paragraphs 1 and 2 of the Labor Code obliges employees to follow supervisors' work-related instructions, provided they don't violate the law, and to look after the employer's interests and keep confidential any information whose disclosure could cause the employer harm. On this basis, a company can introduce a ban in its workplace regulations or issue a direct work order covering specific tools or an entire category of publicly available language models.

The problem starts once an employee breaks that ban and the employer wants to act on it. The harshest response is termination of the employment contract without notice through the employee's fault, the well-known summary dismissal under Article 52 of the Labor Code.

When dismissal holds up

Under Supreme Court ruling I PK 94/15, the concept of a serious breach of basic employee duties consists of three elements that must all be present together: unlawfulness of the employee's conduct, harm to or serious threat against the employer's interests, and fault, meaning intent or gross negligence. Missing even one of these elements means the summary dismissal has no legal basis, and the employee can successfully appeal to a labor court.

In practice, this means that simply clicking into ChatGPT against the rules, with no negative consequences for the company, usually isn't enough to justify dismissal. The situation looks different when an employee pastes trade-secret data, customer personal data, or internal strategic documents into an external AI tool.

If an employee used AI assistance to create some kind of internal report, which led to company data leaking online, that could constitute a breach of employee duties - Katarzyna Kamecka, labor law expert, Polish Economic Society (Polskie Towarzystwo Gospodarcze)

Accident versus bad faith

Lawyers stress that what matters is an individual assessment of each case, not automatically applying sanctions for the mere fact of breaking a ban. A one-off, accidental use of a widely available AI assistant for a simple task, with no protected information disclosed, is judged differently from systematic, knowing circumvention of the ban while handling sensitive customer data or the company's trade secrets.

The risk is especially relevant for professions where an AI tool's mistake carries direct legal or financial consequences. Lawyers using language models can run into situations where the tool invents nonexistent laws or court rulings, which, without verification, can expose both the employee and the law firm or company they work for.

What it means for Polish companies

Experts recommend employers take a more practical approach than an outright ban, which is hard to enforce anyway since employees can use AI on a private phone outside the company network. Instead of a ban, they suggest clearly defining which tools are approved for work, what data must never be entered into them, and, where possible, having the company purchase a more secure, corporate version of an AI tool rather than leaving employees on public, free versions.

For employees themselves, the practical takeaway is that workplace regulations or a work order banning AI must be taken seriously, even if the consequences aren't automatic. The key question a labor court will ask in a dispute isn't just whether the ban was broken, but whether that breach actually harmed or seriously threatened the employer's interests.

For simply failing to follow established workplace order and organization, without any resulting harm or threat to the company, the Labor Code provides for milder measures, a warning or reprimand entered into the employee's file, with no possibility of a financial penalty. Only when a breach of the ban is tied to actual harm, a data leak, or gross negligence does the employer gain grounds for the harshest sanction.