Published on 18 August 2026

Transparency on the use of artificial intelligence for AGOV (agov.ch/ai)

Status August 2026: Artificial intelligence (AI) is now a widely used tool in software development and content creation. For a government authentication service in particular, it is important to distinguish between AI being used as a tool to create software or content and AI forming part of the live system, processing personal data or influencing login procedures. This article provides transparency on how AI is used for AGOV, where it is currently not used, and the principles according to which any future use would be assessed.

Key points

  • Within the AGOV core system considered in this article, AI is currently not used to process personal data.
  • AI is currently not used to assess or carry out AGOV login procedures.
  • AI may be used as a tool in software development, quality assurance and security testing. Responsibility remains with the people and organisations involved.
  • AI may be used extensively to create texts, images, audio and video content relating to AGOV. The Federal Chancellery is responsible for the content it publishes, regardless of the tools used to create it.
  • AGOV authenticates natural persons, not AI agents. These persons may act towards a target system on their own behalf or in an organisational context. AGOV does not determine the resulting authorisations or legal effects.
  • The essential distinction is between AI as a tool used during creation and AI as a runtime component of the AGOV system.

AI is used within the framework of applicable law and the requirements governing the Federal Administration. In this article, the term AI is used in accordance with the Federal Administration’s AI terminology. An AI system is a machine-based system that infers from the input it receives how to generate outputs such as predictions, content, recommendations or decisions. Automation and algorithms are therefore not, in themselves, synonymous with AI systems. At the date of this article, Switzerland does not yet have overarching legislation specifically governing AI. This does not mean that AI is used in a legal vacuum: existing law applies regardless of the technology used. In particular, data protection legislation is technology-neutral and applies directly to AI-supported processing of personal data. Depending on the specific use case, additional information, transparency or other legal obligations may apply. For the Federal Administration, the strategy on the use of AI systems also provides the relevant direction. AI systems are to be used responsibly and the applicable requirements, in particular those concerning the law, information security and data protection, must be observed. Through this article, the Federal Chancellery also provides transparency on the specific situation concerning AGOV. This general overview does not replace any information or other obligations that may arise in a particular use case.

Scope of this article

To ensure that statements about the use of AI are unambiguous, the scope considered in this article is clearly defined.

For the purposes of this article, the AGOV core system includes in particular:

  • AGOV accounts and the functions associated with them;
  • connections between AGOV and the authorities and their applications;
  • the mechanisms used for authentication and for carrying out the actual login procedures.

The article also covers official AGOV-related content created or published by the Swiss Federal Chancellery. This includes texts, images, audio and video content. The scope does not include, in particular, target systems operated by authorities or third parties that are reached after a successful AGOV login. Nor do the statements below concerning the AGOV core system cover upstream, downstream or supporting systems and services provided by the Confederation or third parties, such as DDoS and bot protection or external identity verification systems. Such systems and services may themselves use AI. Whether and how they do so is governed by the legal, contractual, security and organisational frameworks applicable to them.

This boundary solely defines the subject matter of this article. It does not alter any legal responsibilities, competences or supervisory obligations.

AI and the processing of personal data in the AGOV core system

Within the AGOV core system as defined above, AI is currently not used to process personal data. This statement is deliberately precise. AGOV does, of course, process personal data and uses software, algorithms, cryptographic methods and automated processes. Automation and algorithmic processing are not, however, the same as the use of artificial intelligence. If AI were to be used in the future within the AGOV core system to process personal data, the legal requirements and, in particular, the requirements relating to data protection, information security, transparency and proportionality would have to be assessed before its introduction.

AI in the AGOV login process

AI is currently not used for the authentication itself or for assessing an AGOV login procedure. This may change as technology develops. For example, procedures could conceivably be used to combine different technical signals from a login procedure in order to enhance security, identify anomalies or assess their plausibility. Such a use would need to be assessed fundamentally differently from the use of AI as a development tool. If an AI system were to make an assessment during a live login procedure and that assessment influenced whether a login was permitted, prevented or made subject to additional checks, the AI would become a runtime component of the AGOV system. Before such a system was introduced, the relevant legal and data protection requirements, information security, traceability, proportionality, possible risks of discrimination and the implications for digital sovereignty would in particular have to be assessed. Depending on the specific design, data protection provisions governing automated individual decisions could also become relevant.

AI in the development of AGOV code

AGOV is developed under the responsibility of qualified specialists. AI may be used to assist them, just like other development tools.

Potential uses include:

  • generating code suggestions;
  • analysing and reviewing existing code;
  • supporting testing and documentation;
  • identifying errors or vulnerabilities;
  • other development and quality assurance tasks.

Code generated or suggested by AI does not acquire any special status as a result. The same requirements regarding quality, security, verifiability and third-party rights apply as to code created in other ways. Responsibility is not transferred to the AI system used. It remains with the people and organisations responsible for development, delivery, review and acceptance. AI is also relevant to IT security. AI-supported tools may be used in authorised security testing, for example penetration testing or bug bounty activities. At the same time, attackers may also use such tools. Security measures must continue to evolve in response to these technological developments. Regardless of the tool used, the applicable requirements relating to information security, data protection, confidentiality and third-party rights apply when handling information and code.

AI in texts, images, audio and video relating to AGOV

AI is an important tool for creating and editing communication content relating to AGOV.

It may be used, for example, to:

  • draft, structure and revise texts;
  • translate and linguistically align multilingual content;
  • create and edit illustrations and images;
  • generate and edit speech and audio;
  • create and edit videos.

Individual production steps may be heavily supported by AI. Content may also have been created largely or entirely using generative AI. Regardless of the production process, responsibility for the content and its publication remains with the Swiss Federal Chancellery and the responsible individuals.

Official AGOV content is expected in particular to meet the following requirements:

  • compliance with the law;
  • factual accuracy;
  • timeliness;
  • comprehensibility;
  • consistency;
  • correct and equivalent multilingualism;
  • accessibility;
  • respect for third-party rights;
  • suitability for the intended communication purpose.

What matters is not whether every individual production step was performed manually, but whether the published result meets these requirements, has been reviewed and can be responsibly published. Generative AI in particular can produce linguistically or visually convincing results that are factually incorrect. The use of AI therefore replaces neither subject-matter review nor editorial review. Where disclosure of the use of AI is required by law, provided for under applicable rules or necessary to avoid creating a misleading impression, such disclosure is made.

AI use by end users and AI agents

AGOV authenticates natural persons. These persons may act towards a target system on their own behalf or in an organisational context. An AGOV account represents the natural person. AGOV does not determine whether that person is entitled to act for a particular organisation, what role or authorisation the person has, or what legal effect an action produces. These matters are determined by the relevant legal and business context and by the target system. AGOV does not provide separate authentication for autonomous AI agents. The authentication means required for an AGOV login are assigned to the natural person. The situation may be different after a successful AGOV login. Where technically and legally permitted by the relevant target system, end users may use AI tools or AI agents to process content or use functions provided by an e-government service. This generally takes place after the AGOV login and within the relevant target system. Whether automated actions, delegations or AI agents are permitted there, and under what conditions, is a matter for the relevant target system and the authority or organisation responsible for it. In this situation, AGOV authenticates the person. It does not determine what significance the target system assigns to the subsequent use of AI.

AI as a tool or as part of the system

One fundamental distinction is particularly important when assessing the use of AI in connection with AGOV. Today, AI is mainly used for AGOV as a tool to create or review products and artefacts. These may include code, tests, texts, images, audio or video. The result of such work may subsequently become part of AGOV without the AI system used to create it itself becoming part of the live AGOV system. A simple example illustrates the distinction: if a piece of code is created with the support of an AI model, then reviewed and integrated into AGOV, AGOV does not require that AI model to run the resulting code. If, by contrast, an AI system were used during a login procedure and its assessment influenced the subsequent login flow, that AI system would form part of the operational delivery of the service. At present, AI is not a runtime component of the AGOV core system within the scope considered in this article. This distinction is important in particular for data protection, information security, responsibility and digital sovereignty.

AI and digital sovereignty

Artificial intelligence is not, in itself, digitally sovereign or digitally non-sovereign. What matters is how a specific AI system is used, what dependencies arise from that use, and to what extent control, verifiability, ability to switch and ability to act are retained. For a general discussion of AGOV’s digital sovereignty, see also the article “To what extent is AGOV digitally sovereign?” at agov.ch/ds.

Dependency during creation and dependency during operation

The use of AI also requires a distinction between dependency during creation and dependency during operation. If an externally provided generative AI model is used, for example, to assist in creating a text, an image or code, this does not automatically create an operational dependency of the AGOV core system on that model. In principle, the resulting output can be reviewed, stored, further edited and operated independently of the original tool.

Such use may nevertheless create other dependencies and risks, for example:

  • dependency on specific tools or skills in the development process;
  • possible disclosure of information to external providers;
  • confidentiality and data protection issues;
  • questions concerning rights in inputs and outputs;
  • requirements concerning traceability of the creation process;
  • dependency on a provider or its legal and contractual framework;
  • restrictions on the ability to replace one tool with another.

The use of AI as a creation tool is therefore not automatically sovereign or non-sovereign. The specific setup is what matters. An operational dependency would be much more direct if an AI system were indispensable during live operation, for example if login procedures depended on an AI assessment obtained on an ongoing basis. Such a use would raise questions in particular concerning availability, control, data sovereignty, substitutability and resilience. Using an internationally available AI service merely as a creation tool does not, in itself, create an operational dependency of the AGOV core system. What matters is the information actually processed, the dependencies created and the ability to continue providing the services required for AGOV in a controlled and secure manner independently of the AI service concerned.

Future developments

AI and the ways in which it can be used are developing rapidly. The legal framework is also evolving. As of August 2026, the Confederation is preparing a consultation draft to implement the Council of Europe Convention on AI. Legislative measures are envisaged in particular in the areas of transparency, data protection, non-discrimination and supervision. The role of AI in relation to AGOV may therefore also change. The essential distinction remains clear: there is a difference between AI supporting people in development, review and communication and an AI system itself becoming a production component, processing personal data or influencing a login procedure.

This article describes the situation as of August 2026. Significant changes in the use of AI within the AGOV core system should be made traceable in this overview.