Illinois Lawmakers Just Passed America’s Strongest AI Safety Bill

The Illinois House of Representatives passed a measure connected Wednesday requiring frontier AI labs similar OpenAI, Anthropic, and Google DeepMind to person their information practices audited by a third-party. If signed into law, AI information experts archer WIRED it would beryllium the nation’s starring cheque connected the powerfulness of large AI companies.

The bill, SB 315, present heads to politician JB Pritzker’s desk. In a station connected societal media Wednesday, Pritzker said helium plans to motion the bill, citing a request to clasp Big Tech accountable.

Since Congress has yet to walk immoderate meaningful AI information legislation, authorities lawmakers person happily stepped up successful caller years to beforehand bills that amusement their constituents they’re keeping Silicon Valley successful check. As AI tools go progressively popular, and the companies down them contention towards monolithic IPOs, polls amusement that American voters are looking for much AI regulation.

As a result, information advocates and tech companies person zeroed successful connected authorities legislatures arsenic the superior battleground to hash retired however these laws should look. OpenAI’s main of planetary affairs, Chris Lehane, told WIRED past week that the company’s AI argumentation is present oriented astir passing a bid of akin authorities laws.

California and New York presently person the strongest AI information laws, requiring tech companies to supply accusation astir exemplary guardrails and people reports connected information incidents arsenic they occur. Illinois’ measure goes a measurement further, requiring autarkic auditors to verify that an AI laboratory is adhering to its ain information standards. Previously, nary autarkic assemblage was required to support an AI laboratory accountable to its ain information claims.

“We're successful a concern wherever the AI companies people their ain homework,” says Scott Wisor, argumentation manager astatine Secure AI Project, a nonprofit that supported SB 315. “Should SB 315 go law, Illinois would necessitate an autarkic auditor to cheque whether the AI labs successful information adhere to their information commitments.”

Wisor says it’s broadly expected that, nether SB 315, AI labs could usage the Big Four accounting and auditing firms—Deloitte, EY, KPMG, and PwC—to audit their information practices. He besides says it’s imaginable that AI labs could pat members of the AI Evaluator Forum—a conjugation of smaller probe organizations including METR, Transluce, and AVERI—to measure adherence to information standards.

Illinois authorities typical Daniel Didech, a sponsor of SB 315, tells WIRED that authorities legislatures are playing an important relation by shaping America’s AI argumentation and acting arsenic a investigating crushed for immoderate national laws that mightiness travel successful the future. “Laws similar this make a satellite wherever it’s much apt for the national authorities to walk something,” Didech says.

Corporate Interests

Illinois has emerged arsenic a large arena successful the ongoing combat implicit authorities AI laws. OpenAI antecedently supported a measure successful Illinois that would fto AI labs dodge liability if their models caused catastrophic harm. However, Lehane has since said the company’s broad enactment for the measure was an oversight, and it ne'er supported the liability shield successful the bill. More recently, OpenAI endorsed SB 315.

“The Illinois General Assembly has shown existent bipartisan enactment successful advancing SB 315 and processing a thoughtful model for frontier AI safety. As AI systems go much capable, wide expectations astir safety, transparency, incidental reporting, and accountability matter,” says Lehane successful a connection to WIRED.