SCHUTZWERK at Embedded Testing 2026: From Jailbreaks to Prompt Injections

Conference Recap

The Embedded Testing Conference 2026 was a well-organized professional conference with a clear focus on current developments in the embedded sector. The presentations were practical, the atmosphere collegial – a successful event for professional exchange.

Maximilian Heichler, Senior Security Consultant at SCHUTZWERK, participated as a speaker and shared his expertise on AI security. His presentation “From Jailbreaks to Prompt Injections” examined penetration testers' perspectives on systems with language and prediction models; a topic that is gaining importance with the increasing integration of AI in embedded systems.

AI Permeates All Areas of Technology

From Large Language Models (LLMs) in data centers to Small Language Models (SLMs) on embedded devices: AI technologies are now present in many technical areas. This development brings new security challenges for manufacturers of cloud infrastructures as well as for embedded system developers.

Embedded testers, security consultants, and product managers must understand the specific risks of AI applications. Traditional security concepts must be extended to account for the non-deterministic nature of AI systems.

From Jailbreaks to Prompt Injections: Understanding New Attack Vectors

The presentation examined systems with language and prediction models from the perspective of attack detection. Using the well-known “jailbreak” scenario, the new attack vector “prompt injection” was explained.

What Are Prompt Injections?

While jailbreaks aim to bypass security mechanisms of AI models, prompt injections exploit how AI systems process inputs. Attackers can manipulate the behavior of AI applications through cleverly formulated inputs.

Risks and Defense Methods

The presentation highlighted concrete risks:

• Manipulation of AI-driven decisions: Through targeted inputs, attackers can influence the behavior of AI systems
• Bypassing security policies: AI models can be made to violate their own security guidelines
• Defense in Depth: Risks can be minimized through policy enforcement structures

As a central defense measure, it was presented how developers and security professionals can protect their AI applications through multi-layered security concepts.

Practical Takeaways

For security consultants and embedded testers, new requirements emerge:

• Extended testing methods: Traditional penetration tests must be supplemented with AI-specific attack scenarios.
• New evaluation criteria: The security of AI applications requires different metrics than traditional software.
• DevSecOps integration: Security and pentesting must be part of the development process, not an afterthought. The new Cyber Resilience Act (CRA) explicitly requires this.
• Continuous education: The rapid development in AI makes regular training essential.

SCHUTZWERK Expertise in Embedded Systems and AI Security

As Senior Security Consultant at SCHUTZWERK, Maximilian Heichler has contributed his expertise in embedded systems security and cloud security since 2020. His work encompasses IoT technologies, embedded devices, cloud systems, and AI applications.

SCHUTZWERK supports manufacturers and developers in securing their products:

• Threat and Risk Analysis for systematic security risk assessment
• Penetration Tests for embedded systems and AI applications
• Embedded Security Assessments to identify vulnerabilities
• Cloud Security Assessments for cloud infrastructures and applications

Conclusion: A Successful Conference

Embedded Testing 2026 offered technically sound presentations in a pleasant atmosphere. The personal exchange in the professional community was valuable and enjoyable.

For SCHUTZWERK, it was a good opportunity to share expertise while benefiting from the insights of other experts. The findings on AI security flow directly into daily consulting work.

Interested in AI Security and Embedded Systems?

If you have questions about the security of AI applications or embedded systems, please contact us for a non-binding conversation. Our experts support you in evaluating and securing your products.

Further information about our services can be found on our website under Embedded Security and Penetration Tests .