SAP published version 3.0 of its Global AI Ethics policy on 23 September 2026, introducing mandatory AI ethics impact assessments for every AI use case and extending governance to internal self-developed AI applications, the company announced.
Key points
- SAP updated the Global AI Ethics policy to version 3.0 on 23 September 2026, grounded in UNESCO’s Recommendation on the Ethics of Artificial Intelligence
- Every SAP AI use case must undergo a mandatory ethics impact assessment based on 10 principles, receiving a risk score with mitigation measures enforced by SAP’s AI Ethics Office
- SAP designs its systems so that humans retain control over critical decisions, a requirement the policy calls human oversight by design
- Rules apply universally to all employees, including internal self-developed AI tools built using Hyperspace AI or Claude Code
- Independent benchmark of 2,972 companies by Thomson Reuters Foundation and UNESCO found SAP among only 12.4% with explicit policies ensuring human oversight in AI systems
Three disciplines structure responsible AI at SAP
The revised policy rests on three disciplines working together: Ethics defines values and moral boundaries for how technology serves humanity; Security protects systems and data against emerging cyber threats; and Compliance ensures legal alignment with standards such as the EU AI Act. The announcement says the three disciplines guide decisions on whether a use case moves forward, the protections it needs, and who answers for its results once live.
Mandatory assessments and escalation path
Every SAP AI use case must pass a rigorous evaluation before deployment, measured against 10 AI ethics principles. Use cases receive a risk score, and SAP’s AI Ethics Office helps product teams enforce mitigation measures and safeguards. Higher-risk cases face multiple review layers and can be referred to the SAP Global AI Ethics steering committee for a binding ruling. The company said the process provides clear guardrails visible to stakeholders, including customers and employees.
Internal accountability extends to self-developed tools
Every employee acknowledges the policy, and its rules cover all internal AI applications, including those built with tools such as Hyperspace AI or Claude Code. The announcement stated that compliance is non-negotiable across SAP.
External benchmarks highlight industry gaps
Research by the Thomson Reuters Foundation and UNESCO assessing 2,972 companies found SAP among the only 12.4% maintaining an explicit policy ensuring human oversight in AI systems. SAP was also named a leading positive example in the Collective Impact Coalition for Ethical AI Progress Report and participates in the World Economic Forum’s Global Coalition for AI and Social Innovation. Broader industry data cited by SAP shows only 36% of responding companies in the S&P Global Corporate Sustainability Assessment have a dedicated or integrated AI policy, while BSI research found just 27% of global businesses factor their AI footprint into broader sustainability strategies.
CEO Christian Klein stated: “For the mission-critical processes of our customers, ‘almost right’ just isn’t good enough.” The announcement article was written by Matthias Medert, head of Sustainability at SAP.
