AI chats as evidence in Swiss criminal prosecution - what really counts?

Artificial intelligence has long been part of everyday life. People use chatbots such as ChatGPT not only for work or study, but sometimes also in legally sensitive situations - for example to seek advice on possible misconduct. This raises the question: can an AI chat serve as evidence in Switzerland if a criminal investigation is launched?

No special rules for AI content

The Swiss Code of Criminal Procedure (CCP) is technology-neutral. According to Art. 139 para. 1 of the Code of Criminal Procedure, all evidence that is suitable for establishing the truth and was obtained lawfully may be used. This means that an AI chat can also be used as evidence - just like emails, social media messages or voice memos.

The only important thing is that the chat is authentic, comprehensible and unadulterated.

When an AI chat has probative value

For a chat to be usable, investigating authorities or courts must be able to trace it:
- Who conducted the chat,
- when and under what circumstances it was created,
- whether the content is unchanged.

A simple screenshot is usually not enough. To prove authenticity, additional data is often required - such as chat exports with timestamps, device logs or technical confirmations from the platform operator.

Example 1: Self-incriminating statements

Suppose someone is writing in an AI chat:

„I stole the car yesterday - what should I do now?“

If this chat is later found on the person's laptop, it can be used as evidence. The decisive factor is whether it can be proven that the chat really came from the person and was not manipulated. If so, it can help to substantiate a suspicion.

Example 2: Preparation of a criminal offence

A user asks an AI for technical assistance with a prohibited action:

„How do I programme a virus to paralyse a company?“

Such dialogues can become relevant in investigation proceedings for preparatory acts under Art. 260bis SCC - provided there is a verifiable connection between the chat and the user (e.g. IP address, device ID, login).

Limits of utilisation

Not every chat may be used. If a chat was obtained unlawfully - for example by hacking into another person's account - it is generally unusable under Art. 141 of the Code of Criminal Procedure.
Data protection law and international legal assistance also play a role: many AI platforms operate servers abroad, which makes it more difficult to provide evidence and obtain data.

What courts have done so far

Swiss courts have long accepted digital evidence: emails, chat histories and social media posts are part of everyday life in the criminal justice system. AI chats are not a new category, but a continuation of this practice. Their probative value depends on how well they can be linked to other evidence.

An AI chat can be used as evidence in Switzerland - but not automatically. It only carries weight in court if its origin, authenticity and relevance are proven. An isolated text excerpt is not enough. In combination with other evidence, however, a chat can certainly be decisive - whether to incriminate or exonerate a person.

Digital traces do not simply disappear, not even in a chat with an AI. Anyone who believes that there are no consequences in a private dialogue is mistaken - because even artificial conversations can have very real legal consequences.