Currently, there is no explicit provision in the Swedish Code of Judicial Procedure regarding the use of generative AI-models and the impact on the legal privilege for lawyers who are members of the Swedish Bar Association ("Advokats"). Case law has yet to address this issue in the context of Chapter 36, Article 5 of the Swedish Code of Judicial Procedure. The closest relevant precedent from the Supreme Court concerns the prohibition against seizure under Chapter 27, Article 2 under the same Code, which has been found to include electronically stored information.

However, with respect to professional secrecy, the Professional Code of Conduct for Advokats still applies, which provides for observing secrecy and discretion concerning clients' affairs. It is therefore important that Advokats can ensure that their use of generative AI-models are not in conflict with these obligations.

The Swedish Bar Association has issued general advice on the use of generative AI-models which emphasizes the importance of using common sense, exercising caution and taking appropriate measures to protect client information as well as to maintain professional duties. Advokats should therefore carefully review the terms of use and other relevant information to determine if the model meets adequate security requirements so that obligations of discretion and secrecy are being upheld.

Unless it is fully established that the model meets the necessary requirements for the obligations to be upheld, Advokats should refrain from entering sensitive information or other client-related information when using a generative AI models. Otherwise, there is a profound risk that the client information will be leaked to third parties, as both inputs and outputs could potentially be used for further development of the AI model.