DEV Community

Mohammad Taha Azizi Ahad
Mohammad Taha Azizi Ahad

Posted on

Germany Wants Your Group Chat Treated Like a Public Square

Germany Wants Your Group Chat Treated Like a Public Square

On October 7, Germany's Federal Cabinet approved a draft law that should make every messenger developer stop and read the fine print. The draft Act to Strengthen Civil and Criminal Protection against Digital Violence, known by its German shorthand GgdG-E, started life in April 2026 as a law about deepfakes and image-based abuse. The version now moving through the legislative process does something the April draft explicitly refused to do: it reaches into closed chat groups.

I build an encrypted messenger. So when a government redraws the line between a private conversation and a public platform, that line runs straight through my work.

What the draft actually does

The law has two pillars. The criminal-law pillar adds three new offenses to the German Criminal Code: creating and spreading pornographic deepfakes, image-based sexual abuse, and repeated unauthorized surveillance of someone's location or activities through information technology. That last one gets its own new section, 202e StGB. None of that is controversial in the way people usually mean. Few developers will argue that non-consensual intimate images should stay legal.

The civil-law pillar is the part that changes how messengers operate. It gives people who are targeted by digital violence a set of claims they can take to a German Regional Court against intermediaries and internet access providers, even when the intermediary itself is not liable for the content. The claims include disclosure of a user's identity, evidence-preservation orders, and judicial suspension of an account. Non-EU networks have to appoint a domestic service agent in Germany, with fines of up to 500,000 euros for individuals and 5 million euros for legal entities. The draft also includes temporary retention of IP addresses for up to three months so law enforcement can identify perpetrators.

The draft implements part of the coalition agreement and transposes Articles 5 and 6 of EU Directive 2024/1385 on combating violence against women and domestic violence. That much was public in April, when the Federal Ministry of Justice published the ministerial draft and asked for industry submissions by May 22.

The change that matters for messengers

Here is the part that changed between April and November. The April draft explicitly excluded pure messenger and email services. The current draft reverses that. Larger closed chat groups on WhatsApp, Telegram, Signal, Threema, iMessage, and the chat features of Facebook and Instagram now fall inside the scope. The news broke on November 15, first reported by Apollo News.

The draft does not set a fixed participant number. Instead, a group falls under the law when it is large enough that message sharing escapes any one person's control. That is a functional test, not a threshold, and that is exactly what makes it hard to build against. A developer cannot write if (members > 50) and be done. The question becomes: at what point does a group stop being a conversation and start being a publication?

That question has a real answer in the abuse cases the law is aimed at. A group of five friends is a conversation. A group of five thousand strangers passing around someone's photos is a distribution network that happens to have a chat interface. The law is trying to draw that line. The problem is that the line is drawn in words, and words do not compile.

What end-to-end encryption does and does not change

If you build an encrypted messenger, your first instinct is to say the law does not apply to you because you cannot read the messages. That instinct is half right and the half that is wrong is the expensive one.

It is right in this sense: an end-to-end encrypted service cannot delete a message it cannot see, cannot moderate content it cannot read, and cannot hand over plaintext it never had. Any obligation that assumes the provider can inspect content is technically impossible for us, and a law that demands the impossible just pushes developers toward breaking their own encryption or leaving the market.

It is wrong in this sense: almost none of the draft's obligations are about content. Identity disclosure is about who holds the account. Evidence preservation is about metadata, logs, and account records. Account suspension is about the account, not the message. IP retention for three months is about network records. Every one of those can be enforced against an encrypted service without touching a single ciphertext. The draft does not need a backdoor to bite. It needs your registration data, your logs, and your willingness to suspend an account on a court order.

That distinction is the whole debate, and it is the debate most coverage skips. The question is not whether the state can read your messages. The question is how much of the machinery around the messages the state can reach, and whether that machinery is enough to identify and stop an abuser without identifying everyone else in the group.

The design questions this forces on builders

I keep coming back to four questions, and I do not have clean answers to any of them.

First, identity. My messenger uses ML-DSA identity keys, and I show a safety number so two people can check they are talking to the right keys. That is identity in the cryptographic sense: this key belongs to this device. The law wants identity in the legal sense: this account belongs to this person, and a court can find out who. Those are different systems, and bridging them is a design choice with consequences. A messenger that ties accounts to verified real-world identity is easier to police and harder to use safely for the people who need it most.

Second, group size. If the law's test is whether sharing has escaped individual control, then the honest engineering response is to make the boundary visible. A group that grows past a certain size could get a warning that it is now effectively public, or lose features that make mass redistribution easy, like unrestricted forwarding. Signal already limits some of this. But the law does not give a number, so every developer picks their own, and different picks mean different legal exposure.

Third, metadata. Three months of IP retention is a concrete requirement, and it is a reminder that the most sensitive thing a messenger holds is often not the message. It is the record of who talked to whom, when, and from where. My own design keeps as little of that as possible. A retention rule that assumes you have it creates an incentive to keep it, which is the opposite of what privacy engineering tries to do.

Fourth, jurisdiction. The domestic service agent requirement means a non-EU service operating in Germany needs a legal address there. That is a paperwork burden for a company and a wall for an independent developer. I run a small project. A fine of up to 5 million euros is not a compliance cost; it is an existential threat. Laws written for platforms the size of Meta land on projects the size of mine with the same weight.

Where this goes next

The draft is still a draft. The cabinet approved it on October 7, and it remains under deliberation. It still has to pass through parliament, and the messenger-groups expansion is new enough that the debate about it is only just starting. The April version drew industry submissions. The November version will draw more, and louder ones.

I am not going to pretend the underlying problem is invented. Image-based abuse and coordinated harassment in large groups are real, and the people targeted by them deserve a legal route that works faster than a platform's abuse queue. The draft's civil-law pillar, giving victims a court order for identity disclosure and account suspension, is aimed at a gap that exists today: reporting abuse in a closed group often goes nowhere because nobody with the power to act can see inside.

But the law's success will be measured by what it does to the groups that are not abusive. A good outcome is a law that gives victims a fast route to a court order while leaving the encryption intact and the small private group untouched. A bad outcome is a law whose vague group-size test and metadata demands push small encrypted services out of Germany, leaving users with a choice between the giants that can afford compliance and the services that ignore the law entirely.

The honest position for a builder is somewhere uncomfortable: the abuse is real, the tool is blunt, and the difference between the two outcomes is in details that are still being written. If you build messengers, read the draft. The line between a conversation and a publication is being drawn now, and it will be drawn with or without you.


Sources: Federal Ministry of Justice ministerial draft GgdG-E, published April 16, 2026; Baker McKenzie client alert, May 2026; Digital Policy Alert, cabinet approval October 7, 2026; Apollo News via World Today Journal, November 15, 2026; Directive (EU) 2024/1385, Articles 5 and 6.

Top comments (0)