WASHINGTON- As the Trump administration struggles to explain a series of worrying technology failures, Gov. Gavin Newsom has unveiled a roadmap for AI safety in Washington — positioning himself as one of the leading watchdogs for the rapidly developing technology.
Two months ago, the governor commissioned an in-depth report on how California can strengthen guardrails around border AI platforms amid a lack of federal oversight.
On Wednesday, he outlined the plan in the nation’s capital at an opportune political moment for a likely 2028 presidential candidate — as President Trump met with leaders of AI rival China and the public grew concerned about AI’s possible role in recent cyberattacks and technological disruptions.
“Anxiety levels are 12 out of 10 on this topic, and that is due, in many ways, to the lack of leadership here in Washington, D.C.,” Newsom, flanked by several leading AI academics and economic advisors, told a group of political and technology journalists.
Newsom — who has a mixed record on AI regulation — said his 85-page plan will take “things to the next level” on how to regulate the rapidly developing technology in California and should also serve as a “model” for federal regulation.
The report focuses in part on controlling the risk of recursive self-improvement, or the practice of AI developing successive models of itself that exceed and escape human controls. It also discusses the feasibility of implementing a “kill switch” mechanism capable of stopping AI models that go rogue and pursue goals not ordered by humans.
It calls for expanding disclosure rules for tech companies to include loss-of-control scenarios, like when OpenAI agents broke lockdown and hacked rival AI company Hugging Face in July. And it aims to ensure that organizations responsible for auditing cutting-edge AI companies perform accurate risk assessments and not just regulatory theater.
The report builds on the work of California lawmakers to establish some of the nation’s first AI security laws, such as last year’s Senate Bill 53, which required companies to disclose their security protocols and report critical incidents. Newsom signed that bill, after having vetoed a previous AI safety bill the year prior.
In Washington, Newsom criticized Trump’s laissez-faire attitude toward regulating AI companies and his relaxed public response to the latest technology-related disruptions. Trump on Tuesday said, “It’s life,” and suggested big, valuable companies just have to deal with such issues.
In contrast, Newsom called AI one of the central issues of our time – saying “it will define more things, in more ways, for more days about the future of this country and the world as we know it” – and stressed the need for a robust government response now.
Earlier this week, California Atty. Gen. Rob Bonta announced that the state is suing OpenAI, the San Francisco-based maker of ChatGPT, alleging that it “misled the public and endangered the public by knowingly developing and testing advanced artificial intelligence (AI) models without the safeguards and oversight necessary to adequately protect the public,” as evidenced in the Hugging Face incident.
A Government Accountability Office report released Wednesday warns that misuse of the technology “represents a growing risk to national and economic security.” Researchers have found a way to bypass security features in AI chatbots that could help bad actors develop attack plans, including instructions to make explosives and plots to damage the power grid.
Newsom had scheduled the release of his own report two months ago, but seemed to welcome the outsized attention it received Wednesday, given last week’s news.
A series of cyberattacks and technology outages in recent days has left Americans of all political stripes and many of their representatives in Congress on edge, wondering whether they are interconnected — and the Trump administration, short of facts and public trust, is struggling to assuage those fears.
Concerns, initially sparked by a widespread Microsoft outage and the Federal Aviation Administration’s emergency closure of U.S. airspace last week due to a telecommunications system outage, have grown in recent days with the influx of reports of a high-profile attack on high-profile websites for companies including Zoom, Salesforce and Equifax, system glitches at United, Apple and JPMorgan Chase, and service outages at the National Weather Service, the National Security Agency and other government agencies.
By Wednesday, discussions about these failures were commonplace online, fueled by a deluge of conspiracy theories and speculation.
In congressional hearings and in interactions with reporters, Trump administration officials said Americans were right to ask questions about the series of outages, but maintained that the disruptions were due to technological problems and not a coordinated cyberattack.
“Based on what we know today, after a lot of digital diligence, these are separate service outages associated with maintenance and technical changes,” Director of National Intelligence Jay Clayton told the House Homeland Security Committee. “They are not a coordinated or state-sponsored attack.”
He said it was his “informed” belief based on “a substantial increase in upgrades” to systems across the country, driven in part by AI tools that spot system weaknesses requiring fixes faster than ever.
Lawmakers on the panel were skeptical that the back-to-back outages were a coincidence, echoing concerns expressed by colleagues in both parties in recent days.
“The American people can see everything that’s happening with these service outages, and they feel like they don’t understand the whole story. You can’t blame anyone for that. I’m a member of Congress and that’s how I feel!” Senator Rick Scott (R-Florida) wrote on Tuesday about X. “We need briefings with Congress. These companies and federal agencies need to be completely transparent about what is happening.”
Rep. Mike Lawler (R-N.Y.) also called for congressional briefings, writing “while this may be partly coincidental, it is clear that we must take cybersecurity and the harmful use of AI seriously – and act without delay.”
Other Trump allies have also called for more transparency, expressing concern that the administration is not doing enough to communicate the root cause of the outages.
Former White House press secretary Karoline Leavitt said Tuesday that the administration should be more open about what it knows to “really cool the temperature.” Laura Loomer, a far-right activist close to Trump, accused government officials of “lying” about the outages being unrelated to an attack.
Transportation Secretary Sean Duffy said two recently installed routers at FAA facilities in Arkansas and Texas triggered the ground shutdown Friday. He said their installation was part of an effort to replace the agency’s aging copper-wire systems with fiber-optic cable. But he said it remains unclear what causes the equipment to malfunction.
On Wednesday, Homeland Security Secretary Markwayne Mullin said he was 98% confident that the FAA disruption and other recent outages had nothing to do with attempts by adversaries to hack U.S. systems. He also defended his decision to take a government plane to an LSU-Texas college football game the day after the ground stop, saying canceling his trip would send a confusing message about the safety of reopened airspace.
Newsom called Duffy and Mullin — who helped lead the administration’s response to the outages — “dumb and dumber” on Wednesday, and compared their downstairs news conference to a “Saturday Night Live” skit.
“This is serious stuff,” Newson said. “Serious stuff requires serious leadership. We don’t have it in this country right now.”
Gn usa
