THE BUZZ: TECH PUSHES BACK — A slew of high-profile artificial intelligence bills heard in committees Tuesday revealed some of the simmering tensions between Silicon Valley and a Legislature eager to step into the regulatory void surrounding the rapidly evolving technology. The tech industry has fought some of the year’s most high-profile proposals, such as state Sen. Scott Wiener’s rules for large-scale AI models, by calling them too broad or vague. Lawmakers, in response, are trying to tailor their bills without making them too narrow. So far, most of the bills have sailed through friendly committees in their houses of origin. But the real test will be in the coming months as industry groups ramp up efforts to dilute or stop the nation-leading legislation by seeding doubt about the proposals' effectiveness and feasibility. Here are the major AI pressure points we’re tracking as the debates continue and the major disagreements come into focus: Watermarking — Several bills aim to get at one of the most prevalent and visible products of the AI boom, deepfakes, by requiring tech companies to label machine-generated materials. During an Assembly Privacy hearing on one such proposal from Buffy Wicks, opposition groups poked holes in the idea, arguing AI itself far outpaces the technology needed to track it and that labels could easily be circumvented by image cropping or screenshotting. That argument seemed to resonate with some lawmakers, like Assemblymember Jacqui Irwin, who asked, “how do we get Americans to trust watermarks if it’s so easy to overcome a watermark?” We expect these points to resurface over the course of the next several months, as lawmakers weigh similar provisions in other legislation, including state Sen. Josh Becker’s AI Transparency Act, which also faced feasibility questions on Tuesday. Definitions — The way policymakers define AI, down to the word, will directly shape how the fast-evolving technology is regulated. As Paneragate demonstrated, a statute's wording — which can determine what businesses new laws sweep in — matters a great deal. An Assemblymember Rebecca Bauer-Kahan bill that passed with no debate conformed the Assembly’s definition to the one in Wiener’s legislation that would require large-scale AI models to undergo safety testing, among other measures, before they are deployed. Algorithmic bias — Generative AI has captured enormous attention. But Bauer-Kahan argued automated decision tools are already far more prevalent — which is why her sweeping bill to root out bias in algorithms that make decisions on weighty matters like health care, housing, and employment, is one of the most consequential bills in Sacramento and, thus, the country. Her legislation's importance was reflected by the magnitude of the coalition against it. The Chamber of Commerce voiced strong opposition, joined by not just the tech industry but insurers, bankers and life sciences. The two sides remain far apart on a bill that Bauer-Kahan calls vital and opponents say is unworkable. We’ll see if they can close that gap by September. GOOD MORNING. Happy Wednesday. Thanks for waking up with Playbook. Now you can text us at 916-562-0685 — save it as “CA Playbook” in your contacts now. Or drop us a line at lkorte@politico.com and dgardiner@politico.com, or on X — @DustinGardiner and @Lara_Korte. WHERE’S GAVIN? Nothing official announced. | 
No comments:
Post a Comment