AI oversight on the supervisory and executive boards, but no published AI rules or test results
Overall rating “limited”. No deduction for controversies. The superscript numbers link directly to the source.
1 · Responsibility
A supervisory board committee focused on AI, and data and AI are clearly assigned on the executive board; AI principles and approval procedures are known only by name, hence “limited”.
What is documented
- The supervisory board's Innovation and Finance Committee focused on industrial AI use, data strategy and cybersecurity1
- The supervisory board receives annual reports from the Chief Compliance Officer and the Global Chief Cybersecurity Officer1
- On the executive board, Dr Peter Koerte is responsible for “Data & Artificial Intelligence”, while IT and cybersecurity lie with Cedrik Neike2
- Siemens cites a cross-functional Generative AI Governance Task Force and “Generative AI Requirements” as a product-related framework4 5
- A whistleblowing system with an ombudsperson and protection against retaliation is in place; 417 compliance cases were reported in FY20244
The case against
No published AI rules
Sources for Responsibility (8)
- Siemens AG: Report of the Supervisory Board for fiscal year 2025 · 30 September 2025 · primary source
- Siemens AG: Sustainability Statement for fiscal 2025 (part of the Siemens Report FY2025) · 30 September 2025 · primary source
- Siemens AG: Deep Dive Industrial AI (Capital Market Day, presentation by Vasi Philomin, EVP Data and AI) · 13 November 2025 · secondary source
- Siemens AG: Sustainability Report 2024 (fiscal year 2024) · 30 September 2024 · primary source
- Siemens AG: Sustainability Report 2023, p. 13 (online view) · 30 September 2023 · secondary source
- Siemens AG: Siemens Impact 2025 (fiscal year 2025, excluding Siemens Healthineers) · 30 September 2025 · secondary source
- Siemens AG: Annual Financial Report FY2024 · 30 September 2024 · primary source
- Siemens AG (Karriereportal): Senior Principal Applied Scientist, Data & Artificial Intelligence (job ID 493047) · 26 January 2026 · secondary source
2 · Safety
A documented vulnerability process, an independent certificate for a control system and published AI terms of use are documented; Siemens only describes testing and validation of its AI products itself.
What is documented
- The vulnerability process covers reporting, analysis, coordinated disclosure with CVE advisories and protection of good-faith reporters5 6
- TÜV SÜD certified the SIMATIC PCS neo control system under IEC 62443-4-1 ML3, valid until 2029; the certificate covers only this system7
- Siemens Zero Trust covers 62% of relevant applications; cyberattacks are identified as a material risk8 9
- The AI terms prohibit training third-party AI and extracting model information; customers must review AI outputs before use3
- The chat terms prohibit personal data in prompts; Siemens may monitor and moderate interactions11
The case against
Hardly any AI-specific evidence
- For the engineering agent Eigen, Siemens only describes approval by engineers and testing in simulation in its own podcast1
- According to a partner announcement, around a fifth of generated code needs rework; test reports are missing2
- No red-teaming, independent audits or misuse monitoring for the AI products were found
Sources for Safety (12)
- Siemens USA: In 5: AI transforming factory automation (transcript) · 8 September 2026 · secondary source
- Microsoft: Siemens and Microsoft scale industrial AI · 24 October 2024 · secondary source
- Siemens AG: Artificial Intelligence Terms, Version 1.3 (August 2026) · 10 August 2026 · primary source
- Radiology Business: GE HealthCare leads with 72 AI-enabled devices; Siemens, Canon, Philips, Aidoc also make top 5 · 3 November 2026 · secondary source
- Siemens AG: Siemens ProductCERT (home page) · 3 November 2026 · primary source
- Siemens AG: Vulnerability Handling and Disclosure Process · 3 November 2026 · primary source
- TÜV SÜD Product Service GmbH: Certificate IITS2 067801 0009 Rev. 00 (IEC 62443-4-1:2018 ML3, IEC 62443-3-3 SL-C 2) for SIMATIC PCS neo from version 7 · 8 July 2026 · primary source
- Siemens AG: Sustainability Statement for fiscal 2025 (part of the Siemens Report FY2025) · 30 September 2025 · primary source
- Siemens AG: Siemens Impact 2025 (fiscal year 2025, excluding Siemens Healthineers) · 30 September 2025 · secondary source
- Siemens AG: Report of the Supervisory Board for fiscal year 2025 · 30 September 2025 · primary source
- Siemens AG: Generative AI Chat Terms of Use (developer.siemens.com) · 27 September 2023 · primary source
- Cloud Security Alliance (CSA Labs): Research Note: AI-Generated Exploit Scripts Target Siemens S7 PLCs (on CISA/NSA/FBI advisory AA26-231A of 19.08.2026) · 22 August 2026 · secondary source
3 · Transparency
Siemens states general limits of generative AI in its terms of use, but discloses neither the models used nor model or system cards or training data.
What is documented
- The chat terms point out that answers are not always precise or up to date and require users to review them4
- The AI terms apply to services that Siemens identifies as AI-based; system data is only used in aggregated, non-identifiable form1
- Siemens describes its own Industrial Foundation Model trained on industrial data6
The case against
Missing product-specific disclosure
- That Industrial Copilot is based on a purchased language model is only documented by the partner, not by Siemens2
- There is no information on the origin of and rights to the training data of the Industrial Foundation Model6
- Product-specific information on error rates, limits of use and labelling of AI-generated content is missing
Sources for Transparency (6)
- Siemens AG: Artificial Intelligence Terms, Version 1.3 (August 2026) · 10 August 2026 · primary source
- Microsoft: Siemens and Microsoft scale industrial AI · 24 October 2024 · secondary source
- Siemens AG: Siemens Industrial Copilot (product page) · 3 November 2026 · secondary source
- Siemens AG: Generative AI Chat Terms of Use (developer.siemens.com) · 27 September 2023 · primary source
- Siemens USA: In 5: AI transforming factory automation (transcript) · 8 September 2026 · secondary source
- Siemens AG: Deep Dive Industrial AI (Capital Market Day, presentation by Vasi Philomin, EVP Data and AI) · 13 November 2025 · secondary source
4 · Why the harm rule does not apply
The harm rule does not apply. No AI-related finding, fine or lawsuit against Siemens AG was found during the review period.
5 · Controversies (0)
No AI controversy found.
6 · What remains open
- Wording of the responsible AI principles and the Generative AI Requirements
- AI risk classification, inventory under the EU AI Act and approval procedures for purchased models
- Testing, validation results and limits of Industrial Copilot, Eigen and the Industrial Foundation Model
- Risk report and litigation in the 2025 annual report; the report could not be found
- AI medical devices of the subsidiary Siemens Healthineers were not assessed separately and not attributed to Siemens AG
- Model and system cards and information on training data
7 · How it was assessed
Cut-off date: 3 November 2026 · 45 searches, 36 fetches.
Primary documents read (10)
- Sustainability Statement for the financial year 2025 (to p. 27)
- Sustainability Report 2024 (to p. 41)
- Annual Financial Report FY2024 (up to the start of the Risk Report, Chapter 8)
- Siemens Impact 2025
- Supervisory Board Report FY2025
- Counter-motions for the Annual General Meetings 2025 and 2026
- Artificial Intelligence Terms v1.3 and Generative AI Chat Terms of Use
- ProductCERT and Vulnerability Handling and Disclosure Process
- TÜV SÜD Certificate IEC 62443-4-1 for SIMATIC PCS neo
- Ad hoc announcement on the deconsolidation of Siemens Healthineers
Research of 3 November 2026 under Globalance AI Footprint rulebook version 0