Artificial intelligence (AI) is increasingly shaping everyday decisions, from determining who gets a job interview to allocating medical care. However, the United States' first major attempts to regulate AI and address inherent biases are encountering significant obstacles from various fronts.
This year alone, over 400 AI-related bills are under discussion in statehouses across the nation, targeting specific industries or particular aspects of the technology, such as the use of deepfakes in elections or the creation of pornographic images. Amid this flurry of legislative activity, lawmakers in states like Colorado, Connecticut, and Texas are pushing for comprehensive frameworks aimed at curbing AI discrimination.
These landmark bills seek to introduce broad oversight mechanisms to tackle some of AI's most troubling issues. Notably, they address instances where AI systems have failed to accurately assess Black medical patients or have downgraded women's resumes during job application filters. With up to 83 percent of employers utilizing algorithms in their hiring processes, according to the Equal Employment Opportunity Commission, the push for regulation is both timely and critical.
“You have to do something explicit to not be biased in the first place,” explains Suresh Venkatasubramanian, a computer and data science professor at Brown University. His insights highlight the necessity of proactive measures to mitigate bias in AI design and implementation.
The proposed legislation in Colorado and Connecticut mandates companies to conduct \"impact assessments\" for AI systems that significantly influence decisions affecting individuals. These assessments would detail the role of AI in decision-making, the data utilized, potential discrimination risks, and the safeguards in place to prevent bias. While this approach aims to enhance accountability and public safety, companies express concerns over potential lawsuits and the exposure of proprietary information.
David Edmonson of TechNet, a bipartisan network of technology executives, stated that their organization collaborates with lawmakers to ensure that any legislation balances AI’s risks with the need for innovation to thrive. Under the current proposals, companies would not need to regularly submit impact assessments to the government. Instead, they would only be required to report instances of discrimination to the attorney general, leaving the detection of bias largely in the hands of the companies themselves.
This self-reporting model has raised alarms among labor unions and academics, who argue that it could hinder the government’s ability to identify and address AI discrimination proactively. Kjersten Forseth of Colorado's AFL-CIO criticizes the legislation, stating, \"Essentially, you are giving them an extra boot to push down on a worker or consumer.\"
Another point of contention is the scope of legal recourse available under the proposed bills. Most legislation restricts the ability to file lawsuits to state attorney generals and public attorneys, excluding individual citizens. Although Workday, a finance and HR software company, supports this limitation, arguing that non-expert judges handling tech-related cases could lead to inconsistent regulatory outcomes, critics like Sorelle Friedler from Haverford College advocate for citizen involvement as a fundamental right to address grievances.
As the debate unfolds, the outcome of these pioneering AI regulations will set a precedent for how technology intersects with civil rights and corporate responsibility in the United States. The challenge lies in crafting laws that effectively mitigate bias without stifling innovation, ensuring that AI serves as a tool for equitable progress.
Reference(s):
U.S. first major attempts to regulate AI face headwinds from all sides
cgtn.com
