New Technologies and AI in Legal Practice – Ethics, Responsibility and Client Safety

The rapid development of artificial intelligence (AI) and modern digital tools is transforming sector after sector of the economy, including the market for legal services. In response to these challenges, at its plenary session in Rzeszów on 12–13 June 2026, the Polish Bar Council (Naczelna Rada Adwokacka, NRA) adopted a key resolution amending the Code of Advocates’ Ethics and Dignity of the Profession.

The new rules set out precisely the terms on which advocates may deploy modern technology in their day-to-day practice, placing client safety and the protection of professional secrecy (advocate-client privilege) first.

AI as support only, never a substitute (§ 23e)

The central element of the reform is the addition of a new § 23e to the Code. Under its wording, the use of programs for obtaining, processing, analysing, generating, sharing or storing information may only be of a supporting nature. Technology is to make the lawyer’s work easier, but it may under no circumstances replace the advocate’s knowledge, experience and independent judgment.

The resolution expressly identifies the areas of risk and prohibits:

• Breach of professional secrecy — it is impermissible to enter, share or allow the processing of privileged information through tools whose mechanisms or operating principles do not ensure adequate protection of that information.
• Surrendering one’s independence — an advocate may not delegate tasks to technology in a way that would undermine their personal role in conducting the case.
• Uncritical reliance on algorithms — the lawyer has an absolute duty to assess and verify every output produced by a technological tool. Failing to do so and acting to the detriment of the client’s interest constitutes an ethical violation.

Full transparency towards the client

The new rules also govern communication with clients. As a rule, using standard technological tools does not require separately informing the client (unless other provisions of law require it).

However, on the basis of the principle of trust that is fundamental to the profession, at the client’s request the advocate is obliged to indicate which technological tools are being used in the case and for what purpose. The client has the right to know how their information is processed.

The advocate’s personal liability remains unchanged

Introducing new technologies into a law firm requires advocates to raise their digital competence. The lawyer must possess the knowledge needed to assess the limitations and risks of the solutions used — such as so-called “AI hallucinations”, i.e. the generation of fabricated statutes or judgments — and must place particular emphasis on personal-data protection.

Regardless of how advanced the programs used may be, the advocate bears full personal liability for the form and content of the professional actions they take. The wording of § 19(6) of the Code was also amended and now provides that transmitting privileged information by electronic and similar means of communication requires particular caution.

Frequently asked questions

Can an advocate use ChatGPT or other AI tools? Yes, but only as support. Under the new § 23e of the Code of Advocates’ Ethics, technology may make work easier, but it cannot replace the advocate’s knowledge, experience and independent judgment, nor breach professional secrecy.

Must an advocate tell the client they are using AI? As a rule, no, where standard tools are used. However, at the client’s request the advocate must indicate which technological tools are used in the case and for what purpose.

Who is liable for an error generated by AI? The advocate alone. Regardless of how advanced the program is, the advocate bears full personal liability for the form and content of professional actions, including verification of the tools’ outputs.

What are “AI hallucinations”? They are AI-generated false information — for example fabricated statutes or judgments — presented as genuine. The advocate has a duty to verify every such output before relying on it.