International standards organizations can play a crucial role in developing governance frameworks and best practices to help regulate artificial intelligence technologies responsibly on a global level. As AI continues to advance rapidly and become integrated into more applications and workflows worldwide, it is important to establish common standards to address concerns around safety, fairness, transparency, accountability and human rights.
Standards development organizations like the International Organization for Standardization (ISO), the International Electrotechnical Commission (IEC) and the International Telecommunication Union (ITU) bring together experts from industry, government, academia and civil society to work on consensus-driven standards. They have the ability to facilitate discussions between stakeholders from different nations and cultural perspectives. By leveraging this multistakeholder approach, international AI standards can help align regulations and build trust globally in a way that reflects diverse societal values.
Some areas where international AI standards could provide guidance include establishing common frameworks for:
Algorithmic accountability and auditing methods. Standards can outline best practices for documenting design processes, implementing oversight mechanisms, detecting biases and ensuring systems behave as intended over their entire lifecycles. This helps ensure those developing and applying AI are accountable for any social and economic impacts.
Data governance and management. Common standards around data collection methods, personal information protection, documentation of data sources and ongoing monitoring of data distributions can help address privacy, surveillance and social discrimination concerns that might emerge from large datasets.
Transparency into AI system decision-making. Requirements for explaining model inputs/outputs, flagging uncertain predictions and disclosing limitations can help users understand what an AI system can and cannot do. Technical standards specifying explanation formats and human-interpretable justifications facilitate oversight.
Risk assessment and mitigation protocols. Circumscribing when an impact assessment should be conducted, what types of risks to examine (job disruptions, safety, bias etc.) and mitigation strategies can minimize unintended consequences before systems are widely adopted.
Human oversight of high-risk applications. Critical domains like healthcare, education, criminal justice or welfare require human review of significant AI decisions. Standards specifying oversight roles, skills qualifications and intervention procedures can maximize benefits while preventing individual harm.
Validation and certification processes. Common testing methodologies, benchmark datasets and certification schemas give users confidence that systems meet standards of reliability, robustness and fairness before use in real-world, high-stakes scenarios. This encourages responsible innovation.
Transnational data sharing. Agreeing on baseline privacy andconsent standards facilitates international collaboration on medical, scientific and public policy challenges that benefit from large, multinational datasets while preventing exploitation.
ISO and IEC are already working on standards for fairness in machine learning, AI concepts and terminology, data quality assessment and model performance evaluation through Technical Committee ISO/IEC JTC 1/SC 42 on Artificial Intelligence. Other standards under development focus on bias, explainability, auditability and more. The ITU has created focus groups examining ethics, AI applications for good and the environmental impact of technologies.
Developing enforceable international AI regulations will certainly require cooperation between governments. But standards provide a starting point by codifying non-binding best practices. By bringing together diverse views, they can gain broader acceptance than rules unilaterally imposed. And standards encourage continuous improvement, allowing practices to evolve alongside fast-paced technologies.
With participation from AI developers, governments, civil society groups, domain experts and others, international standards offer a framework for addressing cross-border challenges like dis/misinformation, cybersecurity threats, facial recognition abuses and more. By outlining governance procedures, they build institutional capacities and establish mutual obligations between nations. They help foster responsible global development and application of these powerful technologies to benefit humanity.
International standards organizations are well positioned to play a leading role in developing universal guidelines and governance models for using and developing AI responsibly. Their multistakeholder, consensus-driven processes can harmonize regulations worldwide and drive accountability by promoting transparency, oversight, and shared best practices. AI standards established through these venues lay important groundwork to help maximize AI’s benefits and safeguard against unintended social and economic consequences on a global scale.