Business Software Alliance, a global trade association representing large technology companies such as Microsoft, Oracle, and Salesforce, is calling for California Congressional Bill 1018, which regulates, limits the scope of, and includes a more accurate definition of, high-risk use cases for artificial intelligence.
Also known as the Automated Decision Safety Act, the bill sets new rules for how artificial intelligence and other automated decision systems are used in situations that have a significant impact on people's lives, such as housing, work, healthcare, credit, education, and law.
If passed since 2027, AI developers, companies that create or significantly modify such AI systems, and deployers, companies that use them to use them must test these tools before using these tools, provide users with clear notice and explanations, and provide the right to decisions made using these tools, or to appeal decisions.
At the Senate Judiciary Session in July, Rep. Rebecca Bauer Kahan, author of the bill, told lawmakers that the law sets common-sense guardrails for AI systems to reduce critical local bias.
“The reason this is so important is the way these AI tools are built, so we enter the data, historical data, and use it to determine how the world works, and then output the decisions.
However, BSA's senior vice president, Craig Albright, said the bill uses ambiguous language that could affect low-risk AI systems, provides conflicting enforcement by various government agencies, and misunderstands the way AI systems are actually developed and used.
“This bill is really misguided in several ways and will have serious consequences,” Albright said.
Albright argued that the bill must clearly define terms such as “tools used to aid human decision-making” and “quality and accessibility of critical opportunity benefits.” He said that both can be widely applied. He also urged lawmakers to specify which AI tools are defined as “a system intended to make decisions about their eligibility.”
“By clearly reading the current text, there are scenarios where the doctor's office is using scheduling software to make appointments with patients.
Albright also said the bill misconstrued what he calls it, the “AI value chain,” the stage involved in creating and deploying AI tools by various companies. He argued that the bill expects businesses along the chain to test their systems for potentially high-risk use, but that is “not feasible.”
The BSA began lobbying lawmakers in February to address these concerns after the bill was introduced. The group released a letter from the opposition in July, saying, “[e]Effective AI regulations require that responsibilities be assigned based on their actual roles and risks. Both the developer and the deployer need to play a role, but AB 1018 is wrong. ”
The next destination on the bill is the Senate Approval Committee, where state lawmakers returned to Capitol on August 18th from summer vacation.
