Seven States Push for Stricter Artificial Intelligence Regulations

**Regulating Bias in AI: A Critical Frontier**

In recent years, the integration of artificial intelligence (AI) into various sectors of daily life has surged, from preening stacks of resumes to sifting through rental applications, and shaping the contours of medical care. Yet, along with the revolutionary benefits of AI tools such as AI images generators and AI video generators, concerns about discrimination have emerged. A burgeoning industry largely unchecked by central governance, AI bears a potential bias that could skew decisions based on race, gender, or income levels. Recognizing the dangers of unchecked AI, legislators in at least seven US states are embarking on pioneering attempts to frame laws that mitigate bias in AI, as Congress remains largely inactive on this pressing issue.

The proposed state regulations represent an early foray into the broader discourse balancing AI’s vast benefits against its demonstrable risks. “AI does in fact affect every part of your life whether you know it or not,” stated Suresh Venkatasubramanian, an AI expert and co-author of the White House’s Blueprint for an AI Bill of Rights. The matter is especially pressing considering that AI doesn’t always ‘work fine.’

With such rapid growth in the AI industry, legislative success is contingent upon navigating complex issues while negotiating with an industry worth untold billions. Despite the introduction of nearly 200 AI-related bills in state legislatures last year—with only a dozen passing—there is sustained enthusiasm as over 400 AI-focused proposals are currently under discussion. The specter of AI discrimination lingers, compelling states to craft robust measures.

The proposed bills address so-called “automated decision tools”, which although pervasive, largely operate out of public sight. Around 83% of all employers, including 99% of Fortune 500 companies, implement AI algorithms to assist in hiring processes—often unbeknownst to applicants. The crux of the problem lies in the training data for AI systems, which may embed historical biases, effectively camouflaging discrimination. The tale of Amazon’s scrapped hiring algorithm project, which disproportionately favored male candidates, emphasizes how easily AI can perpetuate long-standing inequalities.

AI tools have become a staple in various decision-making processes, these systems learn from historical data rife with past discrimination, leading to biased outcomes—a problem recognized by pioneers in the field, like Christine Webber, who represented plaintiffs in a class-action lawsuit highlighting AI-driven discrimination in tenant screening. The enigmatic nature of such rejections, as detailed in the case of Mary Louis, underscores the need for transparency and accountability.

To counteract AI bias, several bills are making their way through state legislatures from California to Connecticut, drawing lessons from the first comprehensive—and subsequently unsuccessful—attempt at regulating AI bias spearheaded by California last year. The new legislative efforts would require companies using automated decision tools to conduct thorough “impact assessments.” These assessments would detail the involvement of AI in decision-making, data collected, potential risks of discrimination, and the safeguarding measures taken by the company. Some would go further, obligating companies to disclose their use of AI in decisions to customers, granting them an opt-out clause.

While industry voices, represented by BSA’s Craig Albright, show tentative support for such measures, the practical efficacy of these bills is yet to be demonstrated. Initial efforts, like the Washington state bill, have faltered, and the renewed push in California, backed by modifications and tech company support, faces uncertain prospects.

The array of legislative attempts in states including Colorado, Rhode Island, Illinois, Connecticut, Virginia, and Vermont represents progressive strides towards regulating AI’s more pernicious aspects. However, experts like Venkatasubramanian caution that the vaguely defined impact assessments and their secretive nature hamper their effectiveness in identifying and combating bias.

Despite the challenges and complexities involved, the development of regulatory frameworks for AI is a positive movement towards accountability and transparency in technology that has become an indelible component of contemporary society. The latest AI news & AI tools could greatly benefit from an established set of guidelines that ensure ethical practice. As AI’s role in daily life continues to expand, so too must our efforts to safeguard against its potential for bias, ensuring technology serves the interests of fairness and justice for all.