The Texas Responsible AI Governance Act represents a welcome attempt to prevent algorithmic discrimination but is significantly flawed, according to leading consumer advocates.
Texas state Rep. Giovanni Capriglione (R) recently shared a draft of the legislation he’s proposing with stakeholders, his office told Inside AI Policy. Among other things, it would require developers to share descriptions of their training data with their third party deployers in an effort to avoid the risk of harmfully biased outputs.
“It’s fantastic that Rep. Capriglione is tackling the important issue of bias in AI decision systems. When automated decision systems are used to help determine whether you are offered a job, or are selected for an apartment, companies should be accountable,” Grace Gedye, a policy analyst for Consumer Reports told Inside AI Policy.
But she added, “We are still analyzing the bill, and … there are some ways we’d like to see it tightened up that are already apparent.”
The bill is similar to one that became law May 17 in Colorado which Consumer Reports cautiously supported despite concerns about loopholes that create an exception for “narrow procedural tasks.”
The Texas bill -- like the Colorado law -- would require developers and deployers of artificial intelligence systems to conduct impact assessments for determining and addressing harms from algorithmic discrimination, unless there is “sufficient human review,” in which case the system would not meet the definition of “high-risk AI.”
The bills both define algorithmic discrimination as “any condition in which an artificial intelligence system when deployed creates an unlawful differential treatment or impact that disfavors an individual or group of individuals on the basis of their actual or perceived age, color, disability, ethnicity, genetic information national origin, race, religion, sex, veteran status, or other protected classification in violation of the laws of this state or federal law.”
But in the wake of an “anti-woke” provision Sen. Ted Cruz (R-TX) was able to pass through the Senate Commerce Committee, and Republicans promising to undo President Biden’s executive order on AI if elected, opponents have latched onto the Texas bill as one they say would “mandate [Diversity Equity and Inclusion] for AI.”
“To be clear, discrimination is already flatly illegal under Texas and federal statutes, and if any AI innovator were to engage in activities that violate civil rights or other consumer protections, a plethora of laws, regulations, and court-based remedies exist to address those harms,” Adam Thierer of the R Street Institute noted in an Oct. 31 post.
The Texas bill would be enforced by the state’s attorney general, who would be empowered to use public complaints as an impetus for investigating and potentially fining the company or companies involved.
That’s among the chief concerns Consumer Reports’ Gedye has about the bill which she says would form an unhelpful barrier to accountability.
“Typically in this country, if you are discriminated against you can take action on your own behalf and file a lawsuit,” Gedye said. “This bill puts enforcement of the antidiscrimination provisions in the hands of the attorney general -- an office already charged with enforcing many other laws -- and then takes the unusual step of providing companies with a right to cure. This will significantly hamper enforcement and creates totally perverse incentives.”