Description of the Training

Your development teams already use AI tools, often without binding rules. In projects under IEC 61508, ISO 26262 or IEC 62304 this raises specific questions: Can AI-generated artifacts go into the safety case? Who reviews them, and who is responsible?

The most delicate question concerns your trade secrets. Source code, schematics and specifications go to an AI provider as soon as a developer pastes them into a tool or an agent reads the repository. What happens to them there is governed by the provider's terms: training, retention periods, subprocessors, server location. German trade secret law (GeschGehG) protects information only if you can show appropriate confidentiality measures. Using AI without rules puts that protection at risk.

Then there is the EU AI Act. Since February 2025, Article 4 has required companies that use AI to ensure their staff have sufficient AI literacy.

NewTec uses AI agents in the development of functionally safe systems and faces the same questions: How do we protect our customers' data and our own know-how? In the workshop we apply the answers we have found to your organization. You leave with an approval matrix that defines which data may be processed with which tool, and with a prioritized plan for getting started.

Objectives and Benefits

Assess where AI reliably helps in the development process today, and distinguish AI as a tool from AI inside the product's safety function.
Understand which data leaves your company when you use AI, and what the provider may do with it.
Classify your data by protection need and assign each class a permitted operating model: EU cloud, open-source models hosted in Germany, your own infrastructure, an environment with no internet connection.
Know the confidentiality measures and contract clauses that keep your trade secrets, and your customers', protected when you use AI.
Assess the rights to AI output: copyright in code, open-source licenses, inventions.
Define review and approval steps so that responsibility stays with the engineer.
Evaluate your own use cases and leave with a prioritized plan for getting started.

Agenda

Block 1 – Orientation

AI in the development process and AI in the product: two fields with different requirements
Where AI helps along the V-model: requirements, architecture, FMEDA, implementation, test, traceability, safety case
Vibe coding versus spec-driven development: why traceability decides whether you can certify

Block 2 – Protecting trade secrets and IP

What leaves your company: prompts, context, indexed repositories, logs, telemetry
What the provider may do with it: training, retention periods, subprocessors, server location, access by authorities in third countries (e.g. the US CLOUD Act)
Operating models compared: frontier models in the EU cloud, open-source models hosted in Germany, your own infrastructure, an environment with no internet connection
Legal protection: appropriate confidentiality measures under the GeschGehG, NDAs with customers and suppliers, AI addenda
Who owns the result: copyright in AI-generated code, open-source licenses in the output, naming inventors and handling inventions not yet filed
Your know-how in prompts, context and skills: company knowledge you need to protect like source code
Exercise: classify your data and build the approval matrix

Block 3 – Implementation

Governance: approved tools, access rights, excluding specific project areas, handling private AI accounts
Responsibility and expertise: review gates, roles in the team, keeping technical depth when AI takes over routine work
Your use cases: assessment, prioritization and plan for getting started

Target Audience

Executive management, heads of development, safety and security managers, quality management, IT management, project management. Legal, patent and data protection staff

Prerequisites

Experience with development processes for safety-critical products. No prior AI knowledge required.

← All trainings