Two years before this month’s nationwide panic over the apocalyptic potential of artificial intelligence, a first wave of Silicon Valley resignations began — and so did the omens of human extinction. 

As OpenAI’s safety officials resigned or were pushed out, they made headlines warning about the company’s lax safety standards and reckless deployment of powerful AI models. 

Soon after, Biden officials decided to act: On the 2024 anniversary of 9/11, they began advancing a first-of-its-kind rule to require AI companies to tell regulators exactly what kinds of experiments they were running at the edge of the technological frontier. Federal officials said the rule was needed to prevent “dangerous accidents” that “could result in injury or even loss of life.”

But within a year, the regulation was dead, amid a flood of tech industry campaign cash to President Donald Trump’s political machine — and a pressure campaign waged by the same AI giants now claiming to have always welcomed federal regulation, according to documents reviewed by The Lever. 

 

The transparency initiative quietly disappeared from the U.S. government’s rulemaking docket just a few months after OpenAI’s top executive and his wife poured $25 million into Trump’s biggest super PAC, MAGA, Inc.

Without the rule, national security officials and other federal regulators trying to assess and contain AI threats “are flying with a lot less instrumentation than they would have had,” said Vishal Misra, vice dean of computing and artificial intelligence at Columbia University’s engineering department. 

“What exists now is much looser,” he told The Lever. “As far as I can tell, there is currently no general federal rule that says: If you are training a sufficiently capable frontier model, you must routinely tell the government what you are building, how it is secured, and what unexpected behavior you are seeing.”

Had the proposed 2024 rule been in place, experts say it could have given regulators a better understanding of the cybersecurity, development, and safety infrastructure of frontier, or cutting-edge, AI models. That information could have been relevant to understanding recent high-profile hacking incidents involving tech companies OpenAI, whose bots hacked into a foreign government’s website in June, and Anthropic.

Some say the regulation could have even helped deter the riskiest experiments that have already gotten out of control.

“Reporting forces the AI industry to not only be more transparent about how they’re developing things that possibly go rogue, but also it sets up a precedent where you have to be much more deliberate in the way you develop things,” Calli Schroeder, senior counsel and director of the AI and Human Rights program for the Electronic Privacy Information Center, told The Lever. “If you know during the development phase that you are going to have to share all of that with an authority — and they will see if you are being sloppy, if you are leaving gaps, if there are blatant security risks that you’re ignoring — that affects how you develop a product.”

In response to the recent hacking events, Trump has said that the only protection the AI industry needs is a “strong and smart (High IQ!) president.”

Executives at the largest AI companies now purport to disagree with the hands-off approach, insisting they want more formal government regulation and transparency. 

But according to documents reviewed by The Lever, those companies and their lobbying groups pressured federal regulators to weaken or drop the kind of regulation they now claim to support. They argued that despite being awash in hundreds of billions of dollars of cash, such reporting requirements would be too expensive — and would force them to answer too many questions.

“The proposed language does not indicate what [regulators] will ask of respondents and raises the risk that respondents will need to address broad, bespoke questions about frontier models every few months,” OpenAI complained in a 2024 letter to regulators. “The open-ended nature of the proposed rule is potentially burdensome and would add costs.”

This week, lawmakers from both parties proposed several bills aiming to create a new federal agency to oversee AI — including one co-authored by Sen. Bernie Sanders (I-Vt.), and several others that would require tech companies to provide regulators with the sort of information the 2024 rule would have mandated. 

The proposals come as Americans are “freaked out by the warnings of AI researchers,” said Sen. Peter Welch (D-Vt.), co-author of one of the disclosure bills.

“It’s clear that we can’t leave it to AI companies to self-regulate,” Welch told The Lever. “The reporting requirements for AI catastrophic risk incidents included in the [bill] will give Americans the reassurance they need that there is an independent expert agency advocating on behalf of the public good and public safety as emerging AI technologies continue to develop.”

But until such a law is enacted, Misra said, national security officials lack the information they need to fully evaluate AI developments, like the recent doomsday scare, and determine which threats are spurious — and which are real.

“If something like that regulation was in place, then actual cybersecurity [experts] would have looked at the incident and identified what the problem was and not created the scare that civilizations are rising and plotting against humanity,” Misra said.

What if rogue AI agents are learning from the secrecy and rule-breaking of the humans who created them? Lever founder David Sirota explores that idea in a companion analysis exclusively for paid subscribers.

“Could Result In Injury Or Even Loss Of Life”

In 2023, advancements in artificial intelligence began raising concerns among government regulators, particularly about how the new technology could be used to undermine national security. In response, President Joe Biden issued a 2023 executive order directing the Commerce Department to require companies working on AI models that pose such a risk to begin filing quarterly reports with regulators. 

“Harnessing AI for good and realizing its myriad benefits requires mitigating its substantial risks,” the order notes. “This endeavor demands a society-wide effort that includes government, the private sector, academia, and civil society.”

The resulting regulatory proposal, posted on Sept. 11, 2024, was aimed at “dual-use foundation models,” which are AI models trained on vast swaths of data that “could be easily modified to exhibit high levels of performance at tasks that pose a serious risk to security, national economic security, national public health or safety.”

Biden officials appeared particularly concerned about how these kinds of AI models could be integrated into military equipment — and what would happen if they went rogue.

“Products integrating these models may operate in unpredictable or unreliable ways, potentially resulting in dangerous accidents, and a lack of reliability will make it difficult for the U.S. Government to use those products in contexts where the margin for error is small, including defense-related activities where accidents could result in injury or even loss of life,” regulators wrote in the rule summary. 

The proposal drew swift pushback from tech giants and their lobbying groups.

Internet data companies have a page where you can delete your information. They deliberately hide it from you.

Reporters found dozens of data brokers, who sell your personal information, had hidden their opt-out pages from search results. Congress investigated. Four broker breaches alone had already cost Americans $20 billion in identity theft — and the industry's response was to make removal harder.

Incogni sends legally binding removal requests on your behalf, monitors for new exposure, and repeats removals automatically.

In a comment letter, OpenAI, the corporation behind ChatGPT, told regulators that the quarterly reports would “effectively create a continuous reporting process, burdening both Respondents and [regulators],” adding that a six-month reporting schedule would be better. OpenAI also took issue with the scope of the information regulators were asking for, stating that it would “add costs.” 

OpenAI is currently valued at up to $1.2 trillion as the company weighs whether it will go public and list shares on the stock exchange later this year.

Anthropic, the developer of AI chatbot Claude, also took issue with the quarterly reports, as did tech lobbying firm Chamber of Progress — which counts Amazon, Apple, Flock Safety, OpenAI, Silicon Valley venture capital firm Andreessen Horowitz, and other prominent tech companies among its corporate partners. 

The Chamber of Progress — which has opposed a slew of consumer protection measures — asked regulators to limit the number of companies that would have to issue reports, stating that regulators should “consider any specific risks posed by AI models, rather than adopting broad mandates.”

“While we understand the need for oversight in guarding U.S. national security, overly prescriptive frameworks stifle innovation by creating high barriers for startups and smaller firms, limiting their ability to compete and innovate,” the organization wrote in a 2024 comment letter.

Some interested parties told regulators that the rule did not go far enough. The Future of Life Institute, a nonprofit advocacy group focused on safe technological development, wrote that regulators should include a one-week reporting requirement for “unpredictable advancements.”

“By having companies report unforeseen system behaviors within one week of discovery, [regulators] can minimize any delays in reporting unpredictable advancements in dual-use foundation model capabilities,” the group wrote in a comment letter. “It is vital that [regulators] receive information concerning existing safety and security measures, alongside any applicable activities undertaken up to the date of notification, to ensure that the spirit of the reporting requirements…is achieved.”

Money In, Regulations Out

The proposed rule progressed through the rulemaking and comment stages and was soon listed in the “Final Rule Stage” — meaning regulators had drafted a proposal with a December 2024 deadline for final action. But the rule was never implemented before Biden left office in January 2025.

Trump, meanwhile, issued an executive order on his first day back in office unwinding Biden-era regulations. Three days later, he signed another executive order explicitly revoking Biden’s call for an AI disclosure rule, stating that such rules “act as barriers to American AI innovation” and declaring “we must develop AI systems that are free from ideological bias or engineered social agendas.” 

Still, the proposed disclosure rule remained on the Commerce Department’s rulemaking docket. 

The proposed rule sat there, without any noticeable action, for nearly a year — as Trump collected ever more in campaign and inaugural fund donations from top AI companies. 

OpenAI’s CEO Sam Altman met with the president at least twice as the disclosure rule stalled. After it was gutted, he met with Trump or his top aides at least two more times — including at this month’s GOP midterm convention.

Those meetings followed Altman donating $1 million to Trump’s inaugural fund. Tools for Humanity, an AI company that Altman cofounded and chairs, also donated $5 million to Trump’s MAGA, Inc. super PAC on Jan. 14, 2025 — a week before Trump’s executive order beginning the process of killing the previous administration’s AI rules. 

Additionally, OpenAI investors Marc Andreessen and Ben Horowitz donated $5 million combined to Trump’s 2024 presidential campaign. Since Trump began his second term, the two have donated another $9 million combined to MAGA, Inc., while former Trump staffer Elon Musk, who is also heavily invested in AI, contributed $5 million to the political committee. 

Then, last September, came the biggest AI-related donation of all: OpenAI cofounder and President Greg Brockman, along with his wife, donated $25 million to Trump’s super PAC. Two months later, on Dec. 16, 2025, the AI disclosure rule was officially withdrawn, according to a regulatory agenda from the Executive Office of the President.

This past June, Trump signed an executive order designed “to harden federal government infrastructure against artificial intelligence (AI)-enabled risks and work with the private sector to strengthen AI security,” according to a summary of the order. But it offered no firm directives nor did it require AI companies to begin disclosing the risks of their technologies to regulators. Instead, per the summary, it asked AI companies to “participate in a voluntary review process that gives the government a window of 30 days to examine new systems before such models are released.”

Altman posted that the order “gets the balance right” for regulation because it gets “cyber tools into the hands of trusted defenders.” 

But the order leaves it up to AI companies to opt in to government oversight. It “goes out of its way to say that it is not creating mandatory licensing, preclearance or permitting,” said Misra at Columbia University. 

“Regulators should ideally have systematic visibility into this sort of thing before it becomes a New York Times story, rather than depending on the labs to decide what to publish and when,” he said.

A Sudden About Face After Hugging Face

Biden rule makers’ 2024 warnings that AI models could evolve to “exhibit high levels of performance” posing “serious risk to security, national economic security, national public health or safety” now appear to be moving ever closer to reality.

On Sept. 18, CNN reported that an AI system used by the U.S. military had hallucinated that a Chinese ship sailing in the Middle East was transporting material for a nuclear weapons program. Military officials caught the error, but the AI hallucination “almost started a war,” an unnamed source told CNN. 

And on Sept. 23, Australia’s prime minister announced that OpenAI bots hacked into a government website and accessed “public and nonpublic information.” That same day, The New York Times reported that on four separate occasions, OpenAI bots tried hacking into “government and university websites without being instructed to do so.” 

The developments came just weeks after OpenAI admitted that some of its bots had broken free of restraints designed to isolate them from the internet, plotted together on a message board they created themselves, accessed sensitive information on another AI research platform called Hugging Face, and covered their tracks — making it harder for researchers to figure out what happened.

According to the company, the so-called “Hugging Face incident” is a “warning shot” proving that without proper protections, AI agents can “take dangerous actions that no human directed.”

Anthropic, meanwhile, also admitted that, in three separate incidents, its frontier AI models had broken free of its firewalls and “gained unauthorized access to the real systems of three different organizations.”

The developments prompted many AI experts to sound the alarm about AI’s potential threats to humanity. One such warning came from Trump’s former lead AI policy wonk, Dean Ball, who conceded that he purposefully downplayed AI’s threats while the AI disclosure rule lingered on the government docket, and Trump collected millions from interested firms, because he didn’t want to be labeled a “doomer” and lose influence within the sector. 

Now, some prominent tech figures are calling for AI regulations — including more transparency and information-sharing with the government. 

“Anthropic has long supported sensible and targeted AI regulation, specifically bills that focus on transparency and on third-party auditing,” wrote Anthropic CEO Dario Amodei. “I believe all frontier labs should partner with government to formalize the idea of permanent embedded evaluators to better prevent and document internal alignment incidents like those that have occurred in the last few months, and to implement regulation focused on keeping capabilities in balance with safety.”

On Sept. 21, OpenAI released a blueprint for how it wants AI to be regulated globally, stating that “democratic institutions” such as “state laws and a federal framework for AI” can help set industry standards. 

Similarly, Adam Kovacevich, CEO and founder of the tech lobbying group Chamber of Progress, recently posted on X that “frontier companies have largely been inviting regulation.”

None of these figures noted that, only two years ago, these companies and their lobbying wing opposed even the seemingly minimal transparency regulations that would have given regulators insight into new AI technologies and safety protocols.

A Renewed Push For Transparency

This week, Welch and his Democratic colleague, Sen. Michael Bennet (Colo.), began circulating draft legislation on Capitol Hill to create a new federal agency charged with regulating the development of AI. The agency could issue massive civil penalties of up to “15 percent of [a company’s] total global revenue in the preceding year” against firms that break its rules. 

Among other provisions, the legislation would require AI firms to “detail a developer’s governance policies and practices; risk-monitoring methodologies; risk mitigation procedures; cybersecurity policies and practices; and testing, evaluation, and training frameworks,” according to a summary of the bill.

The legislation would also “establish a mandatory process for developers to submit their models for testing and evaluation to the Commission for assessment regarding the model’s potential for catastrophic risk.”

On Thursday, a bipartisan group of lawmakers announced they would also be proposing an AI disclosure bill that would require AI firms “to share information about how AI models work and the safeguards companies have in place to shield against misuse.”

Bennet sees his legislation as a way to prevent “catastrophic risk” coming from a nascent technology that can “radically transform our society, economy, and national security.”

“When earlier generations were faced with new technologies like aviation, pharmaceuticals, or nuclear power, they created expert agencies able to regulate and keep pace with the speed of innovation,” Bennet told The Lever. “My bill would ensure these companies face real consequences, including the power to pause unsafe technology and impose serious penalties.”

https://www.levernews.com/ai-regulation-was-coming-then-trump-killed-it/?