Empfehlungen für einen globalen Prüfrahmen für KI-Systeme

The International Panel on the Information Environment (IPIE) presents a global AI auditing framework designed to address the lack of uniform standards and the failure of current audits to account for global inequalities. The framework consists of seven key characteristics: audit scope, auditor role, stakeholder engagement, context-dependent criteria, evidence requirements, methodology, and post-audit measures. It emphasizes the need for auditor independence, transparent results, and the inclusion of marginalized communities to prevent AI from perpetuating biases, labor exploitation, and environmental degradation.

• Current AI auditing practices are highly heterogeneous and lack uniform standards, making it difficult to ensure consistent quality and accountability. Most audits are conducted in North America and Europe, often overlooking the disproportionate negative impacts of AI on the global population, such as human rights violations and labor exploitation.
• A trustworthy and effective auditing system requires five core elements: uniform norms and procedures, qualified and independent auditors, guaranteed access to evidence, transparent results, and methods that account for unequal global impacts.
• The proposed framework defines seven essential characteristics for AI audits: the scope of the audit, the role of the auditor, mechanisms for stakeholder inclusion, context-dependent criteria, evidence requirements, audit methodology, and comprehensive post-audit measures.
• Audit criteria must be precise and context-specific, covering diverse areas such as AI performance (accuracy, fairness, robustness), data sovereignty (indigenous and collective rights), fair labor practices (compensation, protection from algorithmic management), and environmental impact (energy/water consumption, greenhouse gas emissions, and e-waste).
• Effective auditing requires a mixed-methodology approach. Auditors must not rely solely on documentation provided by the audited entity but must conduct independent performance and scenario tests, gather testimonies, perform on-site visits, and use participatory methods to assess real-world user interaction.
• Post-audit transparency is critical; reports must be publicly accessible via news media, peer-reviewed journals, or public archives. They should include a contact point for stakeholders to challenge AI system aspects and provide a timeframe for the audited entity to remedy identified harms.

APA

International Panel on the Information Environment (2024). Empfehlungen für einen globalen Prüfrahmen für KI-Systeme. https://cdn.prod.website-files.com/643ecb10be528d2c1da863cb/67abd282a03244763b7b5e3e_DE.pdf

Chicago

International Panel on the Information Environment. Empfehlungen für einen globalen Prüfrahmen für KI-Systeme. 2024. https://cdn.prod.website-files.com/643ecb10be528d2c1da863cb/67abd282a03244763b7b5e3e_DE.pdf.

Wikipedia

{{cite report |author=International Panel on the Information Environment |title=Empfehlungen für einen globalen Prüfrahmen für KI-Systeme |date=2024 |url=https://cdn.prod.website-files.com/643ecb10be528d2c1da863cb/67abd282a03244763b7b5e3e_DE.pdf |access-date=23 October 2026 |via=Climate Insights Directory}}

BibTeX

@techreport{internationalpanelontheinformationenvironment2024empfehlungen, author = {{International Panel on the Information Environment}}, title = {{Empfehlungen für einen globalen Prüfrahmen für KI-Systeme}}, institution = {International Panel on the Information Environment}, year = {2024}, url = {https://cdn.prod.website-files.com/643ecb10be528d2c1da863cb/67abd282a03244763b7b5e3e_DE.pdf}, urldate = {2026-10-23}, note = {Indexed by Climate Insights Directory} }