The UNESCO Recommendation on the Ethics of AI Implementation offers a comprehensive, globally applicable ethical framework, while the EU AI Act Compliance Program provides legally enforceable regulations within the EU. The choice between them depends on the specific context and requirements, with the EU AI Act being more stringent and enforceable within its jurisdiction.
| Attribute | UNESCO Recommendation on the Ethics of AI Implementation | EU AI Act Compliance Program |
|---|---|---|
| Scope of ethical guidelines covered | Addresses ethical issues related to AI within UNESCO's mandate, encompassing education, culture, communication, and information. Emphasizes human rights, human dignity, and environmental sustainability. Covers all stages of the AI lifecycle, from research and design to deployment and use, including data governance, gender equality, and AI applications in various sectors like education, culture, labor markets, and health. | Transparency, accountability, and human rights. Obligations on 'high-risk' AI systems and enforces transparency for certain AI systems. Prohibits specific AI practices. |
| Geographical applicability | Applicable to all 194 UNESCO member states. | Applies to AI systems used within the EU, regardless of where they are developed or deployed. Also applies if the AI system's output is used in the EU. Foreign suppliers must appoint an authorized representative in the Union. |
| Legal enforceability | A voluntary framework and not legally binding. However, it signifies a consensus among member states and can influence national regulations. | Legally enforceable. Non-compliance can result in significant penalties, including administrative fines. Member states are responsible for laying down rules on penalties. |
| Industry-specific focus | Offers policy action areas that can be translated into action across various sectors, including health, education, culture, and environment. Addresses the unique ethical challenges within specific sectors like healthcare, emphasizing the need to minimize bias, ensure privacy, and maintain human oversight. | Allows existing regulators to tailor guidelines to specific sectors. Includes use cases in areas such as education, employment, law enforcement, and migration. |
| Compliance monitoring mechanisms | Suggests that member states monitor all phases of the AI system lifecycle, including algorithms, data, and AI actors. Encourages the development of mechanisms for ethical impact assessment, auditing, and continuous monitoring. UNESCO supports these efforts by developing methodologies for ethical impact assessment and readiness assessment. | Market surveillance authorities will conduct regular audits and facilitate post-market monitoring. Providers and deployers must report serious incidents and malfunctions. The AI Office will supervise the implementation and enforcement of the AI Act. |
| Risk assessment methodologies | Emphasizes the need for ethical impact assessments to identify and assess the benefits, concerns, and risks of AI systems. Suggests due diligence and oversight mechanisms to address the impact of AI systems on human rights, the rule of law, and inclusive societies. | Employs a risk-based approach, categorizing AI systems based on their potential societal impact. Risk assessments help identify and mitigate potential risks. Regular evaluations should identify vulnerabilities. |
| Transparency and explainability requirements | Highlights transparency and explainability as essential for ensuring respect for human rights, fundamental freedoms, and ethical principles. Emphasizes the need for transparency in AI systems, especially in high-risk decisions. | Promotes transparency and explainability in AI systems. Transparency involves openly sharing details about data sources, algorithms, and decision-making processes. Explainable AI (XAI) techniques provide insights that humans can understand. Providers of general-purpose AI models have technical documentation obligations. |
| Accountability frameworks | States that AI actors must be accountable for the design and implementation of AI systems. Emphasizes that it should always be possible to attribute ethical and legal responsibility to physical persons or legal entities. | Establishes accountability frameworks to ensure responsible AI development and deployment. Ensures organizations are responsible for the outcomes of their AI systems. |
| Human oversight and control provisions | Stresses that AI systems should not displace ultimate human responsibility and accountability. Emphasizes the importance of human oversight in AI systems, ensuring that humans retain control and can intervene when necessary. | Includes provisions for human oversight and control of AI systems. Mechanisms are required to ensure human operators can intervene or override decisions when necessary. Aims to minimize risks to health, safety, or fundamental rights. |
| Data governance and privacy standards | Calls for the development of data governance strategies that ensure the continual evaluation of the quality of training data, proper data security and protection measures, and feedback mechanisms. Advocates for the protection of privacy throughout the AI lifecycle and the establishment of adequate data protection frameworks. | Addresses data governance and privacy concerns. Enforces standards to protect sensitive information. Aligns closely with the General Data Protection Regulation (GDPR). |
| Bias detection and mitigation techniques | Urges AI actors to minimize and avoid applications and outcomes that strengthen discriminatory or biased outcomes. Promotes diversity and inclusiveness in AI development teams and training datasets. | Includes bias detection and mitigation techniques. Training datasets must be carefully curated to ensure they represent the diversity of the population the AI system is intended to serve. |
| Resource availability for implementation | UNESCO is developing tools and methodologies to support member states in implementing the Recommendation, including a Readiness Assessment Methodology (RAM) and an Ethical Impact Assessment (EIA). It also supports networks of experts and knowledge centers to assist in capacity-building and research on AI ethics. | Includes measures to support innovation, particularly for small and medium enterprises, such as AI sandboxes. The Commission and Member States will encourage voluntary codes of conduct. |
| Price | Not available | Not available |
| Ratings | Not available | Not available |