New York City's Local Law 144 mandates the use of independent impartial bias audits for automated employment decision tools (AEDTs) used for employment or promotions. The enforcement date has been pushed back to 2023 due to concerns about who qualifies as an independent auditor and the suitability of the impact ratio metrics. The updated rules clarify that bias audits must be conducted by a third party and include metrics for calculated impact ratios based on selection rate or average score. The audit can be based on test data when historical data is not available. Additionally, employers must provide AEDT data retention policies, making them available on their website. Holistic AI offers auditing services for businesses seeking compliance.

The New York Department of Financial Services published a circular letter in January 2019 to insurers authorized to write life insurance in the state. The letter warns insurers not to use external data sources, algorithms, or predictive models in underwriting or rating unless it has been determined that the system does not collect or use prohibited criteria. The burden and liability lie with the insurer, and the NYDFS reserves the right to audit and examine an insurer’s underwriting criteria, programs, algorithms, and models and can take disciplinary action if necessary. The letter also highlights the obligation to comply with existing anti-discrimination and civil rights laws and regulations. Insurers should provide transparency to consumers regarding the reason or reasons for any adverse underwriting decisions made using external data sources or predictive models. Failure to comply may result in an NYDFS audit and breach of existing anti-discrimination laws.

The article discusses the current regulatory environment surrounding artificial intelligence (AI) and the need for AI auditing to ensure the safety, legality and ethics of AI systems. The process of AI auditing involves four stages, including triage, assessment, mitigation, and assurance. The assessment phase evaluates the efficacy, robustness and safety, bias, explainability, and algorithm privacy of the system. The audit outcomes are used to inform the residual risk of the system, and mitigation actions are suggested to address the identified risks. The importance of conducting an audit of an AI system is highlighted, including improving stakeholder confidence and trust and future-proofing systems against regulatory changes.

The Artificial Intelligence Training for the Acquisition Workforce Act (AI Training Act) has been signed into law by President Biden, with the aim of educating federal agency personnel on the procurement and adoption of AI. The Act requires the Office of Management and Budget (OMB) to create or provide an AI training program to aid informed acquisition of AI by federal executive agencies, covering topics such as the science of AI, its benefits and risks, and future trends. The AI Training Act is part of a wider national commitment to trustworthy AI, including Executive Order 13960 and the Blueprint for an AI Bill of Rights.

The EU ministers have greenlit the adoption of a general approach to the EU AI Act, which aims to balance fundamental rights and the promotion of AI innovation by defining AI, expanding the scope of the act, clarifying governance, extending the prohibition of social scoring to private actors, designating high-risk systems, and clarifying compliance feasibility for high-risk systems. The final text includes several changes to increase transparency and simplify required conformity assessments. The European Council will now negotiate with the European Parliament, with an agreement expected to be reached by early 2024. Businesses are advised to take steps to manage the risks of AI systems to embrace AI with greater confidence.