United States of America: Introduced Artificial Intelligence Oversight and Liability Bill (HB 711) including AI Authority Governance

On 9 January 2024, the Artificial Intelligence Oversight and Liability Bill (HB 711), including testing requirements for AI systems, was introduced in the Vermont legislature. The Bill proposes to establish testing requirements for developers and deployers of inherently dangerous artificial intelligence (AI) systems. The Bill includes definitions of terms related to AI, responsibilities of developers and deployers, oversight and enforcement mechanisms, requirements for safety and impact assessments of AI systems, standards of care to mitigate risks, prohibitions on unsafe AI products, and avenues for legal action for violations. Specifically, it requires deployers of inherently dangerous AI systems to submit a detailed safety and impact assessment to the Artificial Intelligence Division prior to deployment and every two years after that, including updates for significant system changes. The assessment must cover the system's purpose, use cases, benefits, risks, data handling, transparency, third-party dependencies, proprietary status, post-deployment monitoring, and impact on consequential decisions and biometric data collection. Furthermore, the Bill requires developers and deployers of inherently dangerous AI systems to responsibly avoid foreseeable risks and requires documentation and disclosure of such risks and mitigation strategies to potential users to prevent a wide range of potential harms and abuses. In addition, the Bill prohibits the distribution of inherently dangerous AI systems without strict testing and risk mitigation in accordance with the NIST framework and requires deployers to implement a comprehensive, standards-compliant risk management policy to ensure the safety and accountability of AI deployments.