AI and Legal Professional Privilege

Legal professional privilege (in Norwegian law: advokaters taushetsplikt) is a fundamental protection. It protects confidential communications between a lawyer and their client from being disclosed in court proceedings or to other parties. 

The core principle is that privilege depends on confidentiality. Once confidential information is shared with a third party in circumstances where control over it is lost, the privilege that attached to it may be waived, often permanently. The Norwegian rules on lawyers' duty of confidentiality, Advokatlovens § 32, rest on the same underlying logic: confidentiality is the foundation on which the client-lawyer relationship is built.

In UK and R (on the application of Munir) v Secretary of State for the Home Department [2026] UKUT 81 (IAC), the Upper Tribunal's Immigration and Asylum Chamber addressed, among other things, the use of AI tools in legal practice. The case arose primarily from two sets of proceedings in which lawyers had submitted fictitious case citations apparently generated by AI tools. As the Tribunal noted, freely available generative AI tools such as ChatGPT "are not capable of conducting reliable legal research" and "can produce apparently coherent and plausible responses to prompts, but those coherent and plausible responses may turn out to be entirely incorrect."

However, it was a separate disclosure during the proceedings that raised the confidentiality issue. One of the solicitors admitted that he had uploaded client emails and Home Office decision letters to ChatGPT in order to summarise them for clients. He subsequently understood that this constituted a data breach and informed his clients and the relevant regulators.

The Tribunal's response was direct. It stated that "to put client letters and decision letters from the Home Office into an open source AI tool, such as ChatGPT, is to place this information on the internet in the public domain, and thus to breach client confidentiality and waive legal privilege." The Tribunal further noted that any regulated legal professional doing so would need to bring this to the attention of their regulator and be advised to consult with the Information Commissioner's Office.

The Guidance for Judicial Office Holders regarding AI states that any information inputted into a public AI chatbot "should be seen as being published to all the world," because public chatbots remember every question asked and any information entered, which is then available to be used in responding to queries from other users.

Even disabling the chat history in public AI chatbots does not fully resolve the risk, as it should still be assumed that data entered is being disclosed.

From a privilege perspective, this matters enormously. If a lawyer uploads a client's correspondence, a draft witness statement, or a legal advice document to a consumer AI platform, that information may be stored, processed, and potentially accessed by the provider or third parties. The lack of control over the data means that confidentiality may be lost at the point of upload, regardless of whether any human being reads the material.

In a federal criminal case in New York, a court held that 31 documents created by a defendant using a consumer AI chatbot were protected by neither attorney-client privilege nor the work product doctrine.

First, the chatbot was not the defendant's lawyer, so there was no privileged communication to protect. Second, there was no confidentiality: the chatbot's terms of service permitted the provider to retain user inputs, use them to train its model, and disclose them to third parties, including government authorities. Third, the defendant had not used the chatbot for the purpose of obtaining legal advice, and the chatbot itself disclaimed any ability to provide it.

The court also confirmed that documents do not become privileged simply because they are later handed to a lawyer. As the judge noted, "AI's novelty does not mean that its use is not subject to longstanding legal principles, such as those governing the attorney-client privilege and the work product doctrine." 

The Norwegian rules are different in detail but equivalent in spirit. The Personal Data Act (personopplysningsloven), which implements the GDPR in Norway, imposes additional obligations regarding the transfer of personal data to third parties. Uploading a client's documents to a US-based consumer AI platform without a proper legal basis and appropriate safeguards could therefore breach both the ethical rules and data protection law simultaneously.

Some practical points to keep in mind:

• Treat public AI tools as public spaces. A useful analogy: using a consumer AI platform with client material is closer to reading aloud in a café than speaking in a private office.
• Check the terms of service. The key question is not whether a tool is described as "open source" or "closed source" in a technical sense, but what the provider does with your data. If the terms permit the provider to retain, process, or share your inputs, privilege and confidentiality are at risk.
• The risk is not limited to legal drafting. The solicitor in the UK case had used AI for what he considered an administrative task — summarising Home Office decisions. Administrative use of AI with confidential material carries the same risks as using it for legal submissions.
• Supervising lawyers bear responsibility. The Tribunal emphasised that a solicitor who delegates work to another fee-earner remains responsible for supervision and for ensuring that staff are aware of the dangers of using non-specialist AI for legal research and drafting.

The UK Upper Tribunal's decision does not create new law. The conclusion is logical: if you upload your client's documents to a consumer AI platform, you may be handing them to a third party with no meaningful confidentiality protections. The decision from New York shows that privilege cannot attach when sharing information with AI, no matter how the material is later used. 

Norwegian lawyers operate under the same fundamental logic. The ethical and legal obligations of confidentiality do not disappear because the recipient of the information happens to be a machine rather than a person. As AI tools become more capable and more tempting to use, the professional risks associated with careless use will only grow.

The sensible approach is cautious and pragmatic: use only AI tools that offer contractually binding confidentiality protections, ensure that client material is never uploaded to consumer platforms, and make sure that everyone in your organisation understands where the lines are drawn.