← All articles

· Rimon Soliman

MathWorks and TÜV SÜD: ISO 26262 Guidance for AI-Assisted Workflows

On 15 October 2026, MathWorks announced an extension of the TÜV SÜD certification of its tools, with guidance on classifying AI tools under ISO 26262. It is a concrete example of how AI can enter a regulated development process.

AIISO 26262Simulinkfunctional safetyTÜV SÜD

On 15 October 2026, MathWorks announced an expansion of the TÜV SÜD certification for MATLAB, Simulink, Stateflow and System Composer. These tools are delivered with the IEC Certification Kit. What is new is the addition of generic guidance on classifying artificial intelligence tools under ISO 26262, the functional safety standard for the automotive sector.

What was announced

According to MathWorks, it is the first tool vendor to offer guidance for certain AI-assisted engineering workflows in functional safety applications. The guidance is based on specific use cases assessed by TÜV SÜD and covers AI-assisted Model-Based Design through the Simulink Agentic Toolkit.

The three use cases covered are:

  • agentic model development;
  • model review;
  • test case generation.

The source describes a step beyond simple prompting: AI agents work directly on Simulink models and engineering workflows, rather than merely suggesting text that the engineer must then shape.

The stated purpose

MathWorks states that the guidance shows how these use cases can fit into safety-critical development in a controlled, verifiable way. The declared goal is to accelerate development while maintaining engineering rigor, traceability and compliance with standards.

Neither a ban nor silence: a certification body shows how to use AI inside a regulated process.

Why it matters for automation professionals

The most interesting aspect is the approach. Until now, in many regulated environments, generative AI has either been prohibited or left in a gray area. Here, a third-party body instead outlines a path: AI is treated as a tool, with emphasis on verifiability and traceability.

For PLC programmers and safety system designers, this is a useful pattern. Anyone who wants to use AI to produce logic, reviews or tests in safety-related contexts will still need to demonstrate what the tool did, with which inputs, and how the engineer checked the result.

Caveats and limits

ISO 26262 applies to the automotive sector. Any applicability of this reasoning to standards such as IEC 61508 or IEC 61511, typical of industrial and process safety, is the author's inference and not a claim made in the announcement. Moreover, this analysis is based on a partial version of the source: we do not know the exact scope of the certification, its limitations, or details on availability or additional statements.

Practical takeaways

Even without working in automotive, a few lessons can be drawn:

  • explicitly classify the AI tools used in the development process and document their role;
  • always keep human review of generated results, particularly for safety-related logic;
  • retain verifiable records of prompts, changes and tests produced with the help of AI;
  • before adopting a similar workflow in IEC 61508 or 61511 projects, check with the certification body what is acceptable.

The 15 October 2026 announcement does not settle every question about AI in functional safety, but it marks a concrete step: the discussion shifts from "can it be used?" to "how do you demonstrate it is under control?".