An algorithm that decides which news 300 million people see makes more political decisions every day than any parliamentarian.
All AI systems that prioritise, rank or filter content in the public information space must disclose their criteria. State interference with these algorithms is to be constitutionally prohibited.
Social media and search engines determine which information millions of people see -- and which they do not. The algorithms behind these systems are opaque. Meta, Google and TikTok publish no details about the criteria by which content is prioritised, downranked or suppressed [1].
In Switzerland, over 70 per cent of the population use social media as an information source [2]. When an algorithm decides whether a newspaper article about a parliamentary debate is displayed or not, it intervenes more deeply in democratic opinion-forming than any newspaper ever could. Unlike a human editor-in-chief, the algorithm is accountable to no one.
The EU AI Act classifies AI systems that influence public opinion-forming as high-risk systems. They are subject to transparency obligations: providers must disclose that an AI system is being used and document the basic logic of the decision-making process [3]. However, loopholes remain: the specific weighting factors of the algorithms need not be published.
The Digital Services Act (DSA) of the EU additionally requires very large platforms (over 45 million users) to carry out annual risk assessments and independent audits of their recommendation systems [4].
Transparency means not only disclosure by companies. It also means protection from the state. If governments can influence algorithms -- for example to promote or suppress certain content -- a form of censorship arises that is harder to detect than any traditional press control.
Switzerland has a long tradition of media freedom, enshrined in Article 17 of the Federal Constitution [5]. This guarantee must be extended to algorithmic information filtering. A constitutional amendment that explicitly prohibits state interference with ranking algorithms would be the Swiss equivalent of the First Amendment to the US Constitution -- adapted to the digital age.
Disclosure obligation: AI systems that prioritise content for more than 100,000 users must publish their ranking criteria in a comprehensible form.
Independent audits: Annual review by a body designated by the FDPIC (Federal Data Protection and Information Commissioner).
Constitutional prohibition: State bodies may neither directly nor indirectly influence the algorithmic prioritisation of content.
User control: Every user must have the option of switching off the algorithmic feed and choosing a chronological display.
Switzerland stands at rank 3 of 186 countries in the IMF's AI Preparedness Index [6]. This position carries an obligation. A country that is among the world leaders in AI readiness must also lead in AI governance.
The revised Swiss Data Protection Act (revDPA), in force since September 2023, already strengthens the rights of those affected by automated individual decisions [7]. Algorithmic transparency would be the logical extension: not just protection of the individual, but protection of the democratic public sphere.
[1] Zuboff, Shoshana: The Age of Surveillance Capitalism. PublicAffairs, 2019.
[2] Federal Statistical Office: Internet use and social media in Switzerland, 2024.
[3] EU AI Act, Regulation EU 2024/1689, Article 13 (Transparency obligations), Article 52.
[4] Digital Services Act (DSA), Regulation EU 2022/2065, Articles 34--35.
[5] Federal Constitution of the Swiss Confederation, Art. 17 (Media freedom).
[6] IMF, AI Preparedness Index, 2024.
[7] Swiss Data Protection Act (revDPA), in force since 1 September 2023.