Managing compliance, fairness, and transparency while deploying artificial intelligence systems in regulated industries

Artificial intelligence (AI) is being harnessed to complete important tasks across various domains, while concerns about malicious use and unintended side effects are growing. Recent initiatives towards AI governance have also remarked on the trepidation concerning AI technologies. Although such technologies hold immense potential for beneficial use, they can be misused or unintentionally have negative effects, necessitating oversight mechanisms to guarantee compliance with policies and standards on AI. The wide adoption and use of Generative AI (GenAI) systems have led to increasing calls for their regulation amid fears of misuse, harmful outputs, and unfair bias. Regulated industries depend on the decisions made by AI systems that are not only consequential but potentially discriminatory, thus the governance of these hyperparameters is even more consequential. Concerns about impending harm from GenAI outputs have led to calls for regulation, including recent legislative proposals to create “guardrails”. The implementation of measures is of equal interest: indeed, guidelines abound but many implementation details are unknown.

This article combines insights from AI policy experts and practitioners in regulated industries on challenges and solutions around the implementation of compliance, fairness, and transparency measures. Implementation challenges regarding compliance with AI regulations and aims for fair and transparent AI systems are examined, along with opportunities for AI developers and teams to address such challenges. Several implementation challenges regarding compliance with AI regulations and aims for fairness and transparency in AI systems will be discussed. At the same time, possible entry points for AI developers and teams to respond to such challenges will be shared in the hope of facilitating further discussion and collaboration. Many organizations are confronting both similar challenges and possible entry points to address them in comparable manners. Collaboration on methods and practices of innovation, from both the technical and design perspectives, could enhance governance of AI systems that serve the public interest.