Safe and Ethical AI (SEA) Platform Network · Linking Artificial Intelligence Principles (LAIP)

6. Flexibility

When developing regulatory and non regulatory approaches, agencies should pursue performance based and flexible approaches that can adapt to rapid changes and updates to AI applications. Rigid, design based regulations that attempt to prescribe the technical specifications of AI applications will in most cases be impractical and ineffective, given the anticipated pace with which AI will evolve and the resulting need for agencies to react to new information and evidence. Targeted agency conformity assessment schemes, to protect health and safety, privacy, and other values, will be essential to a successful, and flexible, performance based approach. To advance American innovation, agencies should keep in mind international uses of AI, ensuring that American companies are not disadvantaged by the United States’ regulatory regime.

Principle: Principles for the Stewardship of AI Applications, Nov 17, 2020

Published by The White House Office of Science and Technology Policy (OSTP), United States

Related Principles

· Safety Assurance Framework

Frontier AI developers must demonstrate to domestic authorities that the systems they develop or deploy will not cross red lines such as those defined in the IDAIS Beijing consensus statement. To implement this, we need to build further scientific consensus on risks and red lines. Additionally, we should set early warning thresholds: levels of model capabilities indicating that a model may cross or come close to crossing a red line. This approach builds on and harmonizes the existing patchwork of voluntary commitments such as responsible scaling policies. Models whose capabilities fall below early warning thresholds require only limited testing and evaluation, while more rigorous assurance mechanisms are needed for advanced AI systems exceeding these early warning thresholds. Although testing can alert us to risks, it only gives us a coarse grained understanding of a model. This is insufficient to provide safety guarantees for advanced AI systems. Developers should submit a high confidence safety case, i.e., a quantitative analysis that would convince the scientific community that their system design is safe, as is common practice in other safety critical engineering disciplines. Additionally, safety cases for sufficiently advanced systems should discuss organizational processes, including incentives and accountability structures, to favor safety. Pre deployment testing, evaluation and assurance are not sufficient. Advanced AI systems may increasingly engage in complex multi agent interactions with other AI systems and users. This interaction may lead to emergent risks that are difficult to predict. Post deployment monitoring is a critical part of an overall assurance framework, and could include continuous automated assessment of model behavior, centralized AI incident tracking databases, and reporting of the integration of AI in critical systems. Further assurance should be provided by automated run time checks, such as by verifying that the assumptions of a safety case continue to hold and safely shutting down a model if operated in an out of scope environment. States have a key role to play in ensuring safety assurance happens. States should mandate that developers conduct regular testing for concerning capabilities, with transparency provided through independent pre deployment audits by third parties granted sufficient access to developers’ staff, systems and records necessary to verify the developer’s claims. Additionally, for models exceeding early warning thresholds, states could require that independent experts approve a developer’s safety case prior to further training or deployment. Moreover, states can help institute ethical norms for AI engineering, for example by stipulating that engineers have an individual duty to protect the public interest similar to those held by medical or legal professionals. Finally, states will also need to build governance processes to ensure adequate post deployment monitoring. While there may be variations in Safety Assurance Frameworks required nationally, states should collaborate to achieve mutual recognition and commensurability of frameworks.

Published by IDAIS (International Dialogues on AI Safety) in IDAIS-Venice, Sept 5, 2024

3. Scientific Integrity and Information Quality

The government’s regulatory and non regulatory approaches to AI applications should leverage scientific and technical information and processes. Agencies should hold information, whether produced by the government or acquired by the government from third parties, that is likely to have a clear and substantial influence on important public policy or private sector decisions (including those made by consumers) to a high standard of quality, transparency, and compliance. Consistent with the principles of scientific integrity in the rulemaking and guidance processes, agencies should develop regulatory approaches to AI in a manner that both informs policy decisions and fosters public trust in AI. Best practices include transparently articulating the strengths, weaknesses, intended optimizations or outcomes, bias mitigation, and appropriate uses of the AI application’s results. Agencies should also be mindful that, for AI applications to produce predictable, reliable, and optimized outcomes, data used to train the AI system must be of sufficient quality for the intended use.

Published by The White House Office of Science and Technology Policy (OSTP), United States in Principles for the Stewardship of AI Applications, Nov 17, 2020

6. Flexibility

When developing regulatory and non regulatory approaches, agencies should pursue performance based and flexible approaches that can adapt to rapid changes and updates to AI applications. Rigid, design based regulations that attempt to prescribe the technical specifications of AI applications will in most cases be impractical and ineffective, given the anticipated pace with which AI will evolve and the resulting need for agencies to react to new information and evidence. Targeted agency conformity assessment schemes, to protect health and safety, privacy, and other values, will be essential to a successful, and flexible, performance based approach. To advance American innovation, agencies should keep in mind international uses of AI, ensuring that American companies are not disadvantaged by the United States’ regulatory regime.

Published by The White House Office of Science and Technology Policy (OSTP), United States in Principles for the Stewardship of AI Applications, Nov 17, 2020

3. Scientific Integrity and Information Quality

The government’s regulatory and non regulatory approaches to AI applications should leverage scientific and technical information and processes. Agencies should hold information, whether produced by the government or acquired by the government from third parties, that is likely to have a clear and substantial influence on important public policy or private sector decisions (including those made by consumers) to a high standard of quality, transparency, and compliance. Consistent with the principles of scientific integrity in the rulemaking and guidance processes, agencies should develop regulatory approaches to AI in a manner that both informs policy decisions and fosters public trust in AI. Best practices include transparently articulating the strengths, weaknesses, intended optimizations or outcomes, bias mitigation, and appropriate uses of the AI application’s results. Agencies should also be mindful that, for AI applications to produce predictable, reliable, and optimized outcomes, data used to train the AI system must be of sufficient quality for the intended use.

Published by The White House Office of Science and Technology Policy (OSTP), United States in Principles for the Stewardship of AI Applications, Nov 17, 2020

6 Promote artificial intelligence that is responsive and sustainable

Responsiveness requires that designers, developers and users continuously, systematically and transparently examine an AI technology to determine whether it is responding adequately, appropriately and according to communicated expectations and requirements in the context in which it is used. Thus, identification of a health need requires that institutions and governments respond to that need and its context with appropriate technologies with the aim of achieving the public interest in health protection and promotion. When an AI technology is ineffective or engenders dissatisfaction, the duty to be responsive requires an institutional process to resolve the problem, which may include terminating use of the technology. Responsiveness also requires that AI technologies be consistent with wider efforts to promote health systems and environmental and workplace sustainability. AI technologies should be introduced only if they can be fully integrated and sustained in the health care system. Too often, especially in under resourced health systems, new technologies are not used or are not repaired or updated, thereby wasting scare resources that could have been invested in proven interventions. Furthermore, AI systems should be designed to minimize their ecological footprints and increase energy efficiency, so that use of AI is consistent with society’s efforts to reduce the impact of human beings on the earth’s environment, ecosystems and climate. Sustainability also requires governments and companies to address anticipated disruptions to the workplace, including training of health care workers to adapt to use of AI and potential job losses due to the use of automated systems for routine health care functions and administrative tasks.

Published by World Health Organization (WHO) in Key ethical principles for use of artificial intelligence for health, Jun 28, 2021