The IBM AI Ethics Board Initiatives and the UNESCO Recommendation on the Ethics of AI both aim to promote ethical AI practices, but they operate at different levels. IBM's initiatives are geared towards internal governance and practical tools, while UNESCO's recommendation offers a global framework for policy development and international cooperation. The choice between them depends on whether the focus is on internal company practices or broader policy influence.
| Attribute | IBM AI Ethics Board Initiatives | UNESCO Recommendation on the Ethics of AI |
|---|---|---|
| Scope of ethical guidelines | Built upon principles of trust and transparency, including explainability, fairness, robustness, transparency, and privacy. | Provides a universal framework of values, principles, and actions to guide states in formulating legislation and policies on AI, consistent with international law. Addresses ethical issues related to AI within UNESCO's mandate, encompassing the impact of AI on human beings, societies, and the environment. Promotes AI systems that work for the good of humanity and prevent harm. |
| Transparency and accountability mechanisms | The IBM AI Ethics Board provides governance and decision-making. AI FactSheets capture transparency information. | Emphasizes transparency and explainability of AI systems. Promotes the implementation of accountability measures for AI-related decisions. AI systems should be auditable and traceable, with oversight, impact assessment, and due diligence mechanisms in place. |
| Stakeholder involvement in development | IBM engages with research organizations and multistakeholder organizations. Founding member of the Partnership on AI and the Vatican's 'Rome Call for AI Ethics'. | Developed through broad global consultation involving experts, developers, and stakeholders from around the world. Promotes multi-stakeholder, multidisciplinary, and pluralistic dialogue and consensus-building about ethical issues relating to AI systems. Emphasizes the participation of diverse stakeholders for inclusive AI governance. |
| Enforcement and compliance mechanisms | The AI Ethics Board reviews AI use cases. Regular training is mandated for employees on AI ethics. | A voluntary framework with no direct penalties, but it may influence national regulations. Encourages member states to implement strong enforcement mechanisms and remedial actions to address harms caused by AI systems. Promotes the auditability and traceability of AI systems to this end. |
| Coverage of AI lifecycle stages | Designed to be operationalized throughout the AI lifecycle, from initial concept through deployment and monitoring. Embeds ethical risk management at each stage. | Covers all stages of the AI lifecycle, from research, design, and development to deployment and use, including maintenance, operation, trade, financing, monitoring and evaluation, end-of-use, disassembly, and termination. |
| Consideration of bias and fairness | IBM provides tools and practices to identify and minimize bias in AI systems. Developed open-source toolkits like AI Fairness 360. | Emphasizes the need to minimize, avoid strengthening, or perpetuating applications and discriminatory or biased outcomes throughout the AI system lifecycle to ensure fairness. Calls for AI actors to promote social justice, fairness, and non-discrimination. |
| Focus on human rights and dignity | Acknowledges that the use of facial recognition for mass surveillance or racial profiling is a clear violation of basic human rights and freedom. | The protection of human rights and dignity is the cornerstone of the Recommendation. It establishes values aligned with the promotion and protection of human rights, human dignity, and environmental sustainability. |
| Emphasis on privacy and data protection | Data and insights belong to their creator, and clients retain complete ownership of their data. Has a global Privacy and AI Management System (PIMS) to centrally manage GDPR and other privacy compliance. | Emphasizes the importance of privacy and data protection throughout the AI lifecycle. It calls for adequate data protection frameworks and mechanisms for effective data governance strategies to ensure individual privacy. |
| Promotion of AI safety and security | Has robust safety guardrails such as the Granite Guardian models. Developed the Adversarial Robustness Toolbox. | States that unwanted harms (safety risks) and vulnerabilities to attack (security risks) should be avoided and addressed by AI actors. It promotes the development of international standards for measurable and testable levels of safety and transparency. |
| Support for education and awareness | The AI Ethics Board sponsors workstreams that deliver thought leadership, policy advocacy, and education and training about AI ethics. Has a company-wide educational curriculum on the ethical development of AI. | Promotes public understanding of AI and data through open and accessible education, civic engagement, digital skills, and AI ethics training. It encourages Member States to provide adequate AI literacy education to the public to empower people and reduce the digital divide. |
| Adaptability to evolving AI technologies | The AI Ethics Board regularly reviews new laws to update IBM's policies accordingly. Emphasizes continuous monitoring. | Interprets AI broadly as systems with the ability to process data in a way that resembles intelligent behavior. This is intended to ensure that the Recommendation remains relevant despite the rapid pace of technological change. |
| Measurable impact and effectiveness | Initiatives deliver measurable outcomes, such as performance gains and cost-effective deployment. Uses AI governance solutions to improve fairness in decision-making and reduce AI bias in credit approvals. | Includes monitoring and evaluation chapters and means for implementation in the form of a Readiness Assessment and the Ethical Impact Assessment to ensure real change on the ground. It directs Member States to monitor and evaluate policies, programs, and mechanisms related to AI ethics. |