June 15 Tech and Democracy Roundup
The Preemption Fight, Anthropic vs. Washington (Again), and AI Leaders Push Their Policy Agendas.
Welcome back to your Tech and Democracy Roundup.
I had a great time at the Axios AI+ Summit in New York on June 3, where policymakers and tech executives discussed how AI will impact society over the next few years.
I was particularly interested in a conversation with Alex Bores, a Democratic State Assembly member who is running for a U.S. House seat in New York’s 12th District. Bores, who authored one of the strongest state AI safety bills in the country, described a group text with state legislators across the country coordinating on AI bills:
“Congress wasn’t acting, and so we, as individual state legislators, were sharing texts back and forth to create that national standard,” Bores said. “SB 53 (California) borrowed from earlier versions of RAISE (New York). The final version of SB 315 (Illinois) borrowed from both.”
It’s as if Congress caught wind of Bores’ remarks. The next day, Reps. Jay Obernolte (R-CA) and Lori Trahan (D-MA) unveiled a draft of The Great American AI Act, a bill that would preempt state AI laws for three years – specifically calling out the safety laws in California, New York and Illinois, saying they would be “federalized”.
The bill is unlikely to advance. But federal preemption of state AI laws remains a real possibility. The White House is currently negotiating with Senator Marsha Blackburn (R-TN) to include federal preemption as part of a broader kids’ online safety package.
Advocacy groups and state officials were critical of these latest federal preemption efforts. Brad Carson, president of Americans for Responsible Innovation and a former Democratic representative from Oklahoma, called preemption of state laws a “generational mistake”. Gov. Ron DeSantis wrote on X: “Preempting states re: AI without enacting a sensible federal framework is just an amnesty for Big Tech.”
The battle over who should regulate AI – states, Congress, or the White House – rages on. Now, onto three other big stories, plus the headlines.
3 Big Stories
Trump signed an AI oversight Executive Order. The order, signed on June 2, asks tech companies to share frontier models with the government at least 30 days prior to their release, and establishes a dedicated “AI cybersecurity clearinghouse” to coordinate reviews with the private sector. It is a watered-down version of an order Trump backed out of signing in May. To many observers, the order is better than nothing, but leaves a lot to be desired:
Mariana Olaizola Rosenblat (NYU Stern Center for Business and Human Rights): “The administration has drawn the right map but stopped at the trailhead.”
Paulo Carvao (Harvard Kennedy School): “The order gets the policy problem right. Frontier AI models with advanced cyber capabilities should not be released into the world without serious testing. It leaves the legitimacy problem unresolved. Secrecy, voluntary participation and industry proximity are a fragile combination.”
Council on Foreign Relations: “It reflects an administration trying to sustain its deregulatory, innovation-first posture while confronting the novel cyber risks posed by powerful new models.”
U.S. Department of Commerce ordered Anthropic to suspend foreign access to its latest models. On June 9, Anthropic released Fable 5 and Mythos 5. Three days later, Commerce Secretary Howard Lutnick informed Anthropic CEO Dario Amodei that both models would be subject to export controls, effectively prohibiting their use outside the U.S. and by foreigners within it. Anthropic responded by shutting down the models entirely, claiming it was the only way to comply. Many tech policy experts took issue with the directive – but some reveled in Anthropic’s latest dispute with the government:
Adam Thierer (R Street Institute): “We are talking about a significant escalation in the politicization of AI and centralization of control over advanced computation in this country.”
Dean Ball (Foundation for American Innovation): “An administration whose posture is that we should export advanced AI chips to China, which also wants to ban every non-American on Earth from using our best models? I have no words.”
Yann LeCun (NYU Tandon School of Engineering): “Dario Amodei’s ridiculous fear mongering about Mythos/Fable (and AI in general) finally pays off. One reaps what one sows.”
OpenAI and Anthropic released latest policy frameworks. On June 3, OpenAI released a blueprint for democratic governance of AI, a 9-page document calling for transparency requirements and stronger international cooperation to mitigate AI risks. Coming from OpenAI – a company currently facing 18 lawsuits by families of people who committed or attempted suicide – this line struck me: “Companies should face enforceable consequences for failing to comply with safety obligations. Liability frameworks should preserve accountability for severe harms and should not provide blanket safe harbors.”
Meanwhile, Anthropic released two policy frameworks, and CEO Dario Amodei published his third long-form essay arguing that AI regulation should be modeled after the Federal Aviation Administration: “Frontier AI models, like airplanes, should be required to go through technical testing and auditing, and their release should be blocked or reversed as a threat to public safety if they do not meet high standards of safety.” Amodei published his essay just days before the Commerce Department issued its export control directive.
Taken together, these three episodes illustrate an important point about the state of AI governance in the U.S.: While Congress stalls on passing national legislation, the White House, federal agencies, and AI companies are all vying to shape AI policy to their advantage.
More Tech & Democracy Headlines
White House
The administration released a National Security Memorandum on AI. It outlines an ambitious timetable for federal agencies to establish new AI policies and integrate the most advanced AI models into national security systems.
President Trump organized a meeting with AI company leaders to discuss the government acquiring equity stakes in their firms. Senate Republicans are not fans of the idea.
Congress
Senators Chris Coons (D-Del.) and Jack Reed (D-R.I.) introduced the Responsible Artificial Intelligence in Defense Act (RAIDA). Senator Adam Schiff (D-CA) introduced the Human Authority in Lethal Operations (HALO) Act. The bills are part of a coordinated effort to push the Pentagon to keep humans in control of autonomous weapons. The HALO Act focuses on legal accountability, while RAIDA provides a technical framework for human control.
Senators Mark Kelly (D-AZ), Jim Banks (R-IN), and Todd Young (R-IN) introduced the Artificial Intelligence Data Authorization and Transparency Act. The bill would modernize the labor market surveys used by the Bureau of Labor Statistics (BLS) to ensure the data collected can effectively capture AI’s impact on the workforce.
Senator Bernie Sanders (I-VT) introduced the American AI Sovereign Wealth Fund Act. The bill would establish a national wealth fund through a one-time 50% tax on the stock of OpenAI, Anthropic, and xAI. The Senator defended the proposal in the New York Times: “I recognize that for the government to have a major stake in a company is complicated. But the principle is simple: When a public resource generates wealth, the public should share in that wealth.”
States
Florida became the first state to sue OpenAI. The lawsuit, filed by Florida Attorney General James Uthmeier, claims the company failed to warn users and parents about severe risks, including chatbot addiction and the facilitation of violence.
Seattle enacted a year-long ban on new AI data centers. The city council voted unanimously in favor of a one-year moratorium.
Industry
Meta released new safety features to limit harmful content shown to teenagers on Instagram and Facebook, its first major policy change since being found liable in March for intentionally designing addictive features. The moves underscore the power of state attorneys general to hold Big Tech accountable.
Anthropic launched Claude Corps to teach nonprofits to use AI more effectively. The year-long fellowship program will place 1,000 fellows at US-based nonprofits, where they’ll build AI tools with their host organization.
Civil Society
The Center for Countering Digital Hate published new research showing that threats against politicians spiked after Meta loosened its speech rules last year.
See you in two weeks.