Page principale Navigation Contenu Contact Sitemap Recherche

Transparency on the use of artificial intelligence for eIAM, including AGOV (eiam.admin.ch/ai)

For eIAM, including AGOV, artificial intelligence (AI) is used in areas such as software development and the creation of documentation and communication materials. It can also support organisational workflows. Wherever AI is used, the people and organisations involved remain responsible for their tasks and the results.

Key points

  • According to the documented position as of August 2026, AI is not used to process personal data or to assess and carry out personal login processes in the eIAM core system, including AGOV, as described below.
  • AI supports activities including the creation and review of source code and the creation of texts, translations, images, audio and video.
  • AI can also support workflows, for example by sorting support enquiries by topic and forwarding them to the responsible staff members. Staff handle and answer the enquiries. AI does not make decisions on the matters raised.
  • The use of AI in these workflows is designed to ensure that end users are not disadvantaged. Human oversight and means of correction form part of this approach.
  • eIAM acts as an intermediary for electronic identities and provides functions for managing access rights. The responsible bodies retain responsibility for substantive decisions on access rights and for the use of AI within connected applications.

Scope

eIAM connects applications to internal and external identity providers. These include AGOV for the e-government context and the identities available through FED-LOGIN for the enterprise context. In personal login processes, the natural person is authenticated by the relevant identity provider. eIAM passes the designated identity information to the connected application.

The eIAM core system referred to here comprises the functions provided by eIAM for personal electronic identities, including AGOV, the connections to applications, and the mediation and execution of personal login processes.

The information on AI use also covers software development, official eIAM and AGOV content produced by the Swiss Federal Chancellery, and supporting organisational workflows, such as support.

Independent external identity providers, external identity verification services, supporting security services and connected specialist applications are considered separately. The statements on AI use in the eIAM core system do not automatically apply to their internal processing. Legal responsibilities and supervisory duties remain in place irrespective of this distinction.

Personal data and personal login processes

eIAM and AGOV process personal data using software, cryptographic procedures and automated processes. Such procedures are not AI simply because they operate automatically.

According to the documented position as of August 2026, AI is not used to process personal data within the core system described here. AI is also not used there to assess and carry out personal login processes.

One possible future use could involve analysing technical signals from a login process to detect anomalies. If such an AI assessment influenced whether a login was allowed or an additional check was required, AI would be directly involved in delivering the service.

Before such a use is introduced, its legal permissibility, effects on the people concerned, data protection, information security and operational dependencies are assessed. This also includes traceability, proportionality and potential risks of discrimination.

Software development and review

AI can support software development for eIAM and AGOV. It can suggest code, analyse existing code, support testing and documentation, and help identify errors and security vulnerabilities.

Code created or revised using AI is subject to the same quality and security requirements as other code. Responsibility for development, delivery, review and acceptance rests with the relevant specialists and organisations. Using an AI tool does not change this responsibility.

AI can also assist with authorised security assessments, such as penetration tests or vulnerability research under a bug bounty programme. Since attackers can also use such tools, technological developments affect both the protection of the systems and the threats they face.

When handling source code and other information, the requirements relating to information security, data protection, confidentiality and third-party rights apply just as they do when working without AI.

Texts, translations, images, audio and video

AI is used to create and edit communication materials relating to eIAM and AGOV. This includes drafting and revising texts, translating and aligning content across languages, and creating and editing images, speech, audio and video.

Individual items of content may be created largely or entirely using generative AI. The Federal Chancellery is responsible for the content it publishes, irrespective of the tools used to create it.

AI outputs may appear convincing yet be factually incorrect, misleading or unsuitable for their intended purpose. Substantive and editorial review therefore also forms part of publishing AI-assisted content. Particular attention is paid to factual accuracy, currency, clarity, equivalent language versions, accessibility and respect for third-party rights.

The use of AI is labelled where this is required by law or the applicable rules, or where the absence of a label would create a misleading impression.

AI in supporting workflows

AI can support organisational workflows surrounding eIAM, including AGOV, for example by sorting support enquiries by topic and forwarding them to the responsible staff members or units.

Staff handle and answer the enquiries and make any decisions themselves. AI does not make decisions on the matters raised by end users.

The use of AI is designed to ensure that end users are not disadvantaged. This also applies to preparatory steps such as sorting and forwarding enquiries. Human oversight and means of correction form part of this design. They serve to correct misclassifications and prevent enquiries from remaining unaddressed or their handling from being unjustifiably delayed or restricted.

Data protection, information security and transparency requirements also apply to these steps. Human handling of an enquiry does not remove these requirements from the preceding steps. Under the Federal Act on Data Protection, data subjects have the right to request information about the processing of their personal data.

The legal and organisational framework for eIAM is described on the eIAM website. Information on data processing by AGOV, including AGOV support, is provided in the AGOV privacy statement.

Access rights, technical users and AI agents

Depending on the integration, eIAM passes on information about roles and access rights in addition to identity information. The bodies responsible for an application determine which access rights a person receives. These rights may be managed in eIAM's access management, in the specialist application or in a distributed arrangement. Where access rights are technically checked and enforced depends on the chosen integration.

eIAM thus supports the technical implementation of access rules. A person's substantive authority, whether they may act on behalf of an organisation, and the legal effect of an action are determined by the applicable rules and the relevant procedure.

eIAM also supports technical users, for example for access to web services. These forms of access are distinct from personal logins. A personal login confirms the identity of the human user; it does not assign a separate identity or authority to act on someone's behalf to an AI agent used afterwards.

After a personal login, end users may use AI tools or AI agents to assist them in using a specialist application, where this is technically possible and permitted under that application's rules. Whether such automated actions are allowed and which technical means of access are provided for them are determined by the rules governing the application and the responsible bodies.

Further information on identities, access rights and technical means of access is provided in the eIAM service description.

Operational dependencies and digital sovereignty

For digital sovereignty, the key questions are what information an AI service receives, what control exists over that information, and how dependent development or operation becomes on the service.

For example, if source code is created with AI assistance, subsequently reviewed and integrated into eIAM or AGOV, that code does not need the AI model used to create it merely because of how it was produced. Texts and images can also, in principle, be stored, reviewed and edited independently of the original tool.

Using an external AI tool can nevertheless create dependencies. These may concern the tool's availability during development, the handling of confidential information, rights to inputs and outputs, the provider's legal and contractual framework, and the possibility of replacing the tool.

If, however, AI is needed during an ongoing workflow, its availability and outputs can directly affect that workflow. This also applies to a support process outside the eIAM core system. The effects of an outage and whether people or other tools can take over the task depend on the specific design.

The use of an internationally available AI tool alone therefore does not allow a conclusive assessment of the digital sovereignty of eIAM or AGOV. What matters is the flow of data, the actual dependencies and the ability to continue acting. Further information is provided in the article on the digital sovereignty of eIAM and AGOV.

Legal framework

The use of AI is subject to existing law and the rules applicable to the Federal Administration. The Federal Act on Data Protection also applies to the processing of personal data using AI. This includes, in particular, purpose limitation and proportionality, data security and informing the people concerned. Other relevant provisions include the fundamental rights to equality before the law and protection of privacy, as well as the specific legal bases for identity and access management.

Article 21 of the Federal Act on Data Protection governs information and the rights of data subjects in the case of individual decisions based exclusively on automated processing that have legal consequences for the person concerned or a considerable adverse effect on them. The human handling of support enquiries and human decision-making described here constitute the approach chosen for these workflows.

The Federal Administration's strategy, the Federal Chancellery's guidance and the information provided by the Federal Data Protection and Information Commissioner support the application of these rules. Whether a particular use of AI is permissible depends on its task, the information processed and its specific design.

Legal bases and further information

Further development

The technological possibilities and legal framework for the use of AI continue to evolve. This page provides information on significant changes in AI use in the eIAM core system, including AGOV, and in the workflows described here.