California Ramps Up Efforts to Regulate AI

California’s policymakers are making efforts to regulate the use of artificial intelligence, especially as a way to combat algorithmic discrimination, including testing algorithms ahead of time and pushing for transparency in the use of tools like ChatGPT.

Introduced by the California State Assemblymember Rebecca Bauer-Kahan in March, the AB 331 bill aims to ensure that AI systems that are used to directly impact “consequential decisions” about people—pertaining to factors such as employment, financial, medical and education—receive additional oversight.

This is to determine whether there’s going to be a potentially adverse impact on these important sectors, said Dominique Shelton Leipzig, a partner at the law firm Mayer Brown.

The bill also covers companies that use AI to determine the target audience for ads across these sectors, since certain decisions made based on these algorithms can have an impact on people’s lives, Leipzig added.

“Marketers should build the algorithms from the get-go with trusting minds,” said Leipzig, “instead of ignoring it and waiting for the worst to come back in two years as it did with privacy.” 

As the ad industry shifts away from third-party cookies and towards algorithms to help match audiences, elements such as accuracy and bias checks need to be baked into systems when companies adopt AI tools. Regulators are concerned that, given the rapid developments around AI, these checks are being overlooked. As such, lawmakers in California and other states, including Washington and Minnesota, are picking up the pace to regulate. Meanwhile, industry leaders, including Elon Musk, have pushed for a moratorium on AI development. 

Under the proposed bill, companies using ChatGPT4, such as Microsoft’s Bing, will be asked to be more transparent about the data that is being used ot train the AI models.

In addition, companies will be required to share their impact assessments for using such technology with the Civil Rights Department or face a fine of $10,000. This includes companies that use ChatGPT for customer service chatbots, such as Snap and Shopify.

The bill also contains a private right of action that allows any California resident to sue a company they think is in violation, after which the company has 45 days to remedy the violation.  

While it could be years before the bill becomes a law—it is due for a committee hearing this week where stakeholders will discuss and suggest possible amendments—companies using generative AI, especially for decision-making purposes, or planning to push their own version of it into the market, would do well to identify these processes ahead of coming regulation.

“Companies are going to have to put in place regular processes to understand when they onboard these types of products, conduct data protection impact assessments and understand its consequences,” said David Stauss, an attorney with experience in data privacy at the law firm Husch Blackwell.

A patchwork of statewide AI laws is likely, said Stauss, a similar scenario to the patchwork nature of privacy laws.

Changes within privacy laws

Meanwhile, state agency, the California Privacy Protection Agency, is focusing on data regulation issues around AI under its privacy law.

The state’s privacy law—the most conservative in the U.S.—lets people opt out of data collection when a business uses automated-decision making technology to collect such information. People also have the right to request information on how the technology is being used.  

Currently, the agency is collecting public comments on what regulations around AI should be issued, but it could take some time before anything is finalized.

“By the end of the session, we could be looking at a whole another regulatory regime,” said Stauss.