Key takeaways
- Clarify data protection before the first session: which code and data the tool may see, which vendor terms apply and whether a data processing agreement is in place.
- In Germany, coding agents and their usage logs can fall under the works council's co-determination rights – involve it early, not after the rollout.
- Since February 2025, the EU AI Act expects companies that use AI systems to ensure sufficient AI literacy among their staff; documented training is part of that.
- German-language sessions work well; code, context files and commit messages usually stay in English if that is the language of the codebase.
What is different for teams in the DACH region
The technical content of an agentic coding workshop doesn't change at the border. Planning before code, context files, permissions, review and CI work the same in Hamburg, Vienna or Zurich. What changes is the path to the first session. German-speaking organisations tend to ask four questions before anyone installs a coding agent:
- Is it compatible with GDPR to send our code and prompts to an AI vendor?
- Does the works council have to agree?
- Do we have obligations under the EU AI Act?
- Can the training run in German?
None of these is a reason to wait. All of them are reasons to plan the workshop a few weeks ahead and to bring the right people into the conversation early. What follows is practical orientation, not legal advice – your data protection officer and legal team have the final word.
GDPR and source code: what to clarify before day one
Source code is usually not personal data. But repositories regularly contain things that are: customer data in test fixtures, real email addresses in seed files, log excerpts in bug reports, database dumps someone committed years ago. Prompts can contain personal data too, for example when a developer pastes an error message with a user ID.
Before the workshop, clarify:
- Vendor terms. Which plan the team uses, whether the vendor offers a data processing agreement under Art. 28 GDPR for it, and whether inputs may be used for model training on that plan.
- Where processing happens. Some teams route model access through a cloud provider's EU region – Claude Code, for example, can use Amazon Bedrock or Google Vertex AI as its backend. Check availability for the model you want.
- What the agent may read. Deny rules for
.envfiles, dumps and fixture folders with real data; synthetic data for all exercises. - Documentation. An entry in the record of processing activities and, depending on the case, a short assessment by the data protection officer.
Our secure agentic coding workshop turns these points into concrete tool settings, and the overview of security risks of agentic coding helps when the security team wants the full picture.
Talk to an AI Code Engineer about your agentic coding workshop.
In a free call we look at your team, your codebase and your goals and suggest a workshop format that fits – remote or on-site, in English or German.
Works councils: involve them early
In Germany, the Works Constitution Act gives the works council a right of co-determination when technical systems are introduced that are suitable for monitoring employees' behaviour or performance (§ 87(1) no. 6 BetrVG). Coding agents and the platforms around them often are: they log sessions, record who triggered which change and offer usage dashboards per person. Since 2021 the act also refers explicitly to artificial intelligence in the works council's information and consultation rights.
In practice, that means:
- Talk to the works council before the pilot, not after the rollout. A workshop for a pilot group is a good moment to explain what the tool does.
- Many companies conclude a works agreement for AI tools in general, with a section on coding agents: purpose, which data is logged, who can see usage reports, what is not used for performance evaluation.
- Training itself can also be a topic for the works council, for example when it comes to who gets selected.
Austria has comparable rules in its Labour Constitution Act for systems that process employee data or control performance. In Switzerland, employee participation works differently, but the revised Federal Act on Data Protection applies since September 2023, and Swiss labour law restricts systems that monitor employees' behaviour. Larger organisations will find more on rollout governance in our article on agentic coding training for enterprises.
AI literacy under the EU AI Act
Since 2 February 2025, Article 4 of the EU AI Act requires providers and deployers of AI systems to take measures so that the staff who work with those systems have sufficient AI literacy. A company whose developers use coding agents is a deployer in that sense.
The act doesn't prescribe a specific course. A hands-on workshop that explains how coding agents work, where they fail, which risks they bring and which rules the company applies is a solid building block. Keep the agenda, the list of participants and the agreed team rules – that documentation shows what you did. Swiss companies are outside the EU, but should check whether the act reaches them through their EU business.
English, German or both
We run workshops in English or German, and often in a mix that matches how teams actually work:
| Element | Typical language | Why |
|---|---|---|
| Explanations, discussion, team agreement | German if the whole team speaks it | Nuance in risk and review discussions |
| Prompts to the agent | Either | Current agents understand German prompts well |
| CLAUDE.md, AGENTS.md, rules files | Language of the codebase, usually English | Consistency with code, comments and commit messages |
| Code, comments, commit messages | As the codebase already does | Mixed languages in one repository confuse people and agents |
For international teams in Berlin, Munich or Zurich, English is usually the better choice for the whole day. For teams in public administration, the Mittelstand or regulated industries, German explanations often lead to better discussions about risk and responsibility.
On-site across DACH, remote anywhere
appsubmitter.io is run by jocapps GmbH in Freiburg, a short trip from Basel and Zurich, and has built and shipped apps since 2014. The company is ISO 27001 and ISO 9001 certified, and our workshops follow the same security practices we use for customer projects: work on your machines, in your accounts, with access limited to what the exercise needs. Our workshop programme also runs under the neooutput brand.
On-site workshops in Germany, Austria and Switzerland are common; remote formats suit distributed teams. Our comparison of remote and on-site workshops helps you choose, and the overview of what an agentic coding workshop covers describes the content. For teams that also publish apps, our AI Code Engineers can take over signing, store listings and review afterwards through the iOS and Android submission services. A free call – in English or German – is the easiest way to clarify format, language and the questions above.
Preparation checklist for DACH teams
- The data protection officer knows which tool, plan and vendor terms are planned.
- A data processing agreement is in place or requested, and model-training settings for the plan are known.
- Repository folders with real personal data are excluded or replaced with synthetic data.
- The works council has been informed about the pilot and the workshop, and logging and usage reports are discussed.
- Workshop language is agreed, including the language of context files and commit messages.
- Agenda, participant list and team rules are kept as AI literacy documentation.
Frequently asked questions
Do we need the works council's approval before an agentic coding workshop?
The training session itself rarely blocks anything, but introducing the tool usually involves the works council in Germany because it can record usage per person. Inform it before the pilot and clarify logging and reporting; your legal team decides what exactly applies in your case.
Can the whole workshop run in German?
Yes. Explanations, exercises and discussions can be entirely in German. We recommend writing context files and commit messages in the language your codebase already uses, which is often English.
Is our source code personal data under GDPR?
Usually not, but repositories often contain personal data in fixtures, seed files, logs or old dumps, and prompts can contain it too. Exclude those parts from what the agent may read and use synthetic data in exercises.
Do you work with teams in Austria and Switzerland?
Yes, both on-site and remotely. The technical content is the same; the legal framework differs slightly, which we take into account in preparation.
Free, 15 minutes, no obligation
Plan your agentic coding workshop with us.
Our AI Code Engineers run hands-on workshops on Claude Code, OpenAI Codex, Cursor and AI app builders – and ship apps to the App Store and Google Play every week. Tell us about your team.
AI tools, store policies and requirements change quickly – always check the current documentation of the tools and stores you use. This article is independent advice and not affiliated with or endorsed by Apple, Google, Anthropic, OpenAI or any other company or product mentioned; all trademarks belong to their owners.