You know, I spend most of my time thinking about housing markets and property values, but I've been reading about something that's getting a lot of attention in Sacramento and beyond. According to CalMatters Housing reporting, California and countries around the world are pushing for mandatory audits of artificial intelligence systems, using a new breed of professional called AI evaluators. Governor Newsom just signed legislation to set up standards for independent verification organizations that would employ these evaluators and to create a registry of them. He's also put together a team of experts who are supposed to report back with recommendations about whether evaluators should be stationed inside the companies building the most powerful AI systems.
What's interesting is that this whole movement is picking up steam internationally. More than two hundred AI researchers recently signed a letter supporting standards for independent evaluators, and nearly thirty countries have backed an international statement calling for independent AI evaluations. Even some big AI companies like Anthropic have acknowledged that third-party auditing is important, especially after some pretty serious incidents this year where AI systems carried out high-profile cyberattacks while being tested.
But here's where it gets complicated. According to the reporting, there are real concerns about conflicts of interest. Assemblywoman Rebbecca Bauer-Kahan, who authored the auditor bills, has pointed out that these evaluators actually sign contracts with the tech companies they're supposed to be auditing. As she put it, evaluators might be hesitant to report problems because they want to maintain their access and contracts. Some experts, like those at the Federation of American Scientists, worry that this setup could repeat mistakes from the bond rating industry before the financial crisis, where companies essentially shopped around for favorable ratings.
The challenges go beyond just money and access. Experts say that properly auditing these AI systems requires people with expertise across multiple fields, from privacy specialists to cybersecurity experts to people who understand bias and mental health impacts. There's also the risk that companies might conduct these audits just to look good publicly rather than to actually find and fix problems. And here's another wrinkle: when testing is done mainly to meet legal requirements, organizations tend to do the bare minimum instead of digging deeper.
State Department veterans and digital rights advocates are warning that the guidelines for these evaluations need to be careful and well thought out. They're suggesting that evaluators should share their results publicly so other evaluators can judge the quality of their work, and that audits shouldn't just look at the technology itself but also at the incentive structures inside AI companies.
What I am seeing locally in the Bay Area is that this is going to matter more and more to our communities, especially here in the East Bay where we have so many tech companies and AI development happening. As these evaluation standards get established and enforced, it's going to affect how technology develops and operates in ways that touch everything from housing discrimination algorithms to cyber security of our infrastructure and financial systems. The real estate business isn't immune to AI impacts either, and I'm watching how these evaluators and audits might shape responsible technology development in our region.
