Germany's government bill against digital violence would criminalise sexualised deepfakes and add an offence for harmful deepfakes

Key takeaways

• This is a government bill. The Ministry page says "Entwurf" and the draft text leaves the entry-into-force date blank.
• Proposed section 184k StGB would punish intentionally or knowingly producing or making accessible, without authorisation, an image that appears to show sexual acts or intimate body parts of a specific person, including computer-made content.
• Proposed section 201b StGB would punish making accessible, without authorisation, computer-made content that appears to show a real event involving another person and is suited to damage that person's reputation considerably.
• The explanatory memorandum says captions visible at first glance that indicate manipulated content, and are suited to rule out the impression of a real event from the outset, would take it outside section 201b, while invisible digital signatures would not.

What does the German digital violence bill propose, and where does it stand?

The Ministry describes a bill to improve civil and criminal protection against digital violence. Its page states the aim as closing gaps in criminal liability for sexualised digital violence and covering the area more clearly. The same page lists the bill as a draft ("Entwurf"), with a last update of 7 October 2026 and first publication on 17 April 2026. It lists a ministry draft (Referentenentwurf, 17 April 2026), association statements (3 June 2026) and a government bill (Regierungsentwurf, 7 October 2026).

The page does not report a Bundestag vote, and the draft text leaves the entry-into-force date blank. Article 17 of the draft provides that Article 10 would enter into force the day after promulgation and that the rest of the Act would enter into force on the first day of the quarter after promulgation. The explanatory memorandum says the Act should be evaluated, as to its civil part, no later than five years after entry into force. Nothing in this article describes current German law.

How would the sexualised deepfake offence in section 184k work?

The Ministry page says the bill is concerned in particular with the unauthorised production and distribution of sexualised deepfakes: "Es geht dabei insbesondere um das unbefugte Herstellen und Verbreiten von sexualisierten Deepfakes". The draft text of section 184k(1) no. 2 StGB covers image content that, because it was created or altered with a computer program, gives the impression that sexual acts, or the unclothed genitals, buttocks or female breast of a specific other person, are shown. It would apply to whoever, intentionally or knowingly and without authorisation, produces it or makes it accessible to a third person. The draft sets a penalty of up to two years' imprisonment or a fine.

The same subsection also covers ordinary image recordings of sexual acts or intimate body parts, so deepfakes sit in one provision with voyeuristic and intimate images. Under the draft, subsections 1 and 2 would not apply to acts that serve overriding legitimate interests, and the text names art, science, research, teaching and reporting on current events. Prosecution would generally need the victim's application, unless the authorities see a special public interest. Our article on the EU AI Act ban on intimate image abuse systems covers a separate EU-level prohibition on AI systems that generate intimate imagery without consent, which the Commission's AI Act service desk says applies from 2 December 2026. This bill does not replace it.

What would the new offence for harmful deepfakes in section 201b cover?

The Ministry page says that making other reputation-damaging deepfakes accessible without sexual content should be covered by a further new offence: "Das unbefugte Zugänglichmachen sonstiger ansehensschädigender Deepfakes". The draft section 201b StGB would apply to whoever, without authorisation, makes accessible to a third person content created or changed with a computer program that appears to show an actual event involving another person and is suited to damage that person's reputation considerably. The draft sets up to two years' imprisonment or a fine, applies only if no other provision carries a heavier penalty, and extends to deceased persons.

The explanatory memorandum gives examples and limits. It names an AI-generated video that makes a well-known physician appear to advertise products, and content aimed at politicians and candidates in political debate. It says content that is plainly not a real event falls outside the offence even if passed on to deceive. It also says that captions visible at first glance, which indicate the content is manipulated or artificially generated and are suited to rule out the impression of a real event from the outset, would take it outside the offence, while invisible digital signatures and barely perceptible captions would not. It says fabricated voice recordings and film sequences are covered, and real images combined with false text quotes are not. The memorandum contrasts the draft with an earlier Bundesrat proposal on Bundestag document 21/1383 and says the draft requires that the content be objectively suited to damage reputation considerably.

What else does the bill add that platforms and service providers should read?

The Ministry page lists a new offence on identity abuse, covering reputation-damaging communication under a fake profile of a real person (draft section 201c StGB), and a new offence on repeated or constant digital surveillance, for example through hidden GPS trackers (draft section 202e StGB). It also describes a civil side. Victims would get a new right to information: after a court order, online platforms and internet access providers would have to disclose the identity of rights violators. Courts could also issue evidence-preserving orders, and victims could ask for removal of unlawful content and, for serious violations where there is a risk of repetition, a temporary block of the violator's account. Operators of social networks would be subject to new duties to name a domestic agent for service of process.

The Ministry page and the explanatory memorandum put the civil measures under court order. The memorandum says the judge-ordered route is meant to check that the underlying statement is criminal before identity data is released. Our article on Azerbaijan's deepfake crimes law shows a different approach, pairing offences with AI labelling duties.

What should platforms, media teams, employers and lawyers do now?

This bill itself imposes no current obligations, so the useful work is preparation and watching the text as it moves.

• Platforms and AI image service providers: map how user reports of deepfakes are handled today, and who would answer a German court order for user identity data or a removal order.
• Media and communications teams: record approval steps for any synthetic image, voice or video of a real person, and whether a visible label travels with the file.
• Employers: check whether staff policies cover creating or sharing synthetic media of colleagues, since the draft offences would apply to individuals.
• Lawyers: follow the Bundestag stage for changes to sections 184k, 201b, 201c and 202e before advising on exposure, and read the government bill itself rather than a summary.

What are the key dates and penalty thresholds?

What we did not verify

What we opened: the Ministry's legislative procedure page and the government bill PDF it links (Gesetzentwurf der Bundesregierung, with explanatory memorandum), both on 2026-10-25.

What we did not open: the cabinet decision itself, press coverage of it, the synopsis and Q&A document, the earlier ministry draft, the association statements, and any Bundestag or Bundesrat record. The government bill PDF carries no cabinet date, so we state the 7 October 2026 date as the Ministry page gives it for the bill document. We also did not test how courts would read the terms "considerably" or "unauthorised".

What we refuse to claim: that the bill has passed or will pass in this form, a date it would take effect, any fine or sentence beyond the penalty ranges in the draft text, or that it does or does not reach AI model developers beyond the provisions described above.

Related briefings

• AI Regulation News hub
• Europe AI regulation tracker
• Attorneys and AI regulation
• EU AI Act Article 5 ban on intimate image abuse
• Azerbaijan Law 387-VIIQD on deepfake crimes and AI labelling
• Denmark copyright bill on deepfakes and likeness rights
• EU AI Act Article 50 transparency duties