GermanyDigitale Medien-Staatsvertrag Teil 2 (Discussion Draft), Rules for Providers of AI Information Systems (Proposed Sections 77a to 77g of the Medienstaatsvertrag)
Diskussionsentwurf der Rundfunkkommission der Länder für einen Digitale Medien-Staatsvertrag (DMStV) Teil 2 Medienstaatsvertrag §§ 77a bis 77g (neu)
A published draft that has not reached a legislature, dated , as of .
An AI transparency rule binding private bodies.
As of .
Where it has got to
The text described here is Diskussionsentwurf der Rundfunkkommission für einen Digitale Medien-Staatsvertrag (DMStV) Teil 2, published .
Locally, this stage is Diskussionsentwurf (Anhörung).
The stage above is recorded at rundfunkkommission.rlp.de.
Comments are invited until . The draft is not before any state parliament and needs the consent of all 16 states.
What it requires
- This measure is proposed and binds nobody yet; what follows is what it would require if enacted in this form.
- Name an authorized recipient of service in Germany and point to that person in an easily recognizable and directly accessible way in your offering.
- Take responsibility for the content your AI information system generates.
- Keep easily perceivable, directly accessible and permanently available information on how your AI information system works, in understandable language.
- For news and political information content, state how likely the statements in newly generated content are to be accurate.
- Do not discriminate against third-party journalistic and editorial offerings without an objective reason.
- Make sure reproduced news and political content reflects the sources used without distorting their meaning.
- Hand the documents the state media authority asks for to that authority.
What this law does
The Rundfunkkommission der Länder deliberated on proposals titled Digitale Medien-Staatsvertrag Teil 2 for a reform of the Medienstaatsvertrag and the Jugendmedienschutz-Staatsvertrag on and put them up for discussion. The discussion draft states that its texts are not finally agreed. Comments on the draft are invited until .
The draft would add a new category of telemedia, the KI-Informationssystem (AI information system), which provides new content to an undetermined circle of users, individualized and automated, also on the basis of third-party journalistic and editorial offerings. Its proposed rules would not apply, apart from the first two paragraphs of section 77b, to an AI information system that reaches fewer than one million users per month in Germany on a six-month average.
A provider of an AI information system would be responsible for the content its system generates. A provider would have to keep easily perceivable, directly accessible and permanently available information on how the system works, in understandable language. For news and political information content, a provider would also have to state how likely the statements in newly generated content are to be accurate.
An AI information system would not be allowed to discriminate against third-party journalistic and editorial offerings. A provider would have to ensure that reproduced news and political content reflects the sources used without distorting their meaning.
When LexLint raises it
When your app profile says your app deploys a chatbot or generates content with AI.