The burgeoning power of artificial intelligence has ignited a fierce debate among technologists, policymakers, and the public regarding its future governance. At the heart of this discussion lies a fundamental question: who should wield control over the development and deployment of advanced AI systems? This complex issue recently resurfaced with a stark warning from Erik Voorhees, founder of the AI service Venice, who drew a parallel to the fictional struggle against a dark lord to caution against granting politicians unchecked authority over artificial intelligence. His pointed commentary, shared on November 12th, underscored a growing apprehension about the concentration of political power over machine intelligence, particularly in the lead-up to potential legislative action following the upcoming U.S. election.

Voorhees’s pronouncement followed a widely circulated clip featuring David Sacks, co-host of the popular podcast "All-In," who predicted a significant Democratic initiative to establish a federal department tasked with approving frontier AI models before their public release. This prediction aligns with existing legislative proposals, including a September bill introduced by Senators Bernie Sanders and Representative Greg Casar, which seeks to create such a department with broad approval powers over advanced AI systems. Concurrently, a campaign platform from Abdul El-Sayed has outlined a similar agency capable of blocking AI releases, albeit with a distinct set of proposed safeguards. The juxtaposition of these proposals and Sacks’s commentary has amplified concerns about the potential for governmental overreach and the implications for AI innovation.

The clip, disseminated by @firesidealpha on November 7th, captured Sacks forecasting a substantial push for AI legislation in the forthcoming Congress. He articulated a vision of a federal AI department meticulously reviewing the most sophisticated AI models. This department, according to Sacks’s prediction, would communicate its concerns to developers and possess the authority to withhold release until these issues are satisfactorily addressed. Sacks further elaborated on his concerns, alleging that elements within the Democratic party harbor aspirations for AI-powered surveillance and censorship. While the legislative proposals in question grant release-blocking powers, the specific political motivations Sacks attributes to these ambitions remain a point of contention, with proponents of the legislation emphasizing safety and risk mitigation over suppression.

These remarks originated from the November 7th episode of the "All-In" podcast, announced by @theallinpod. In the full episode, available on YouTube, Sacks expanded on his predictions, foreseeing a potential "bargain" where the government accrues enhanced power, while established, closed-model AI companies might benefit from regulatory protection. This perspective suggests that the implications extend beyond mere censorship, raising the specter of compliance rules inadvertently shielding incumbent developers from competitive pressures.

Voorhees, in his critique, champions the principles of open-source transparency and decentralization as superior mechanisms for ensuring AI safety. He contends that a system of licensing, as proposed by some legislative efforts, inherently undermines equality by creating artificial barriers to entry and innovation. His core objection, therefore, is to any entity, be it corporate or governmental, that assumes a gatekeeping role over the development and dissemination of advanced AI.

The Blueprint for AI Oversight: Proposed Departments and Approval Mandates

The concept of a governmental body with the authority to approve or deny the release of advanced AI systems is not merely hypothetical; it is already being actively pursued through legislative channels. On September 23rd, prior to Sacks’s November remarks, Senators Bernie Sanders and Representative Greg Casar unveiled the "Ban Artificial Superintelligence Act." The Senate version of this legislation, S.5493, proposes the establishment of a dedicated Department of Artificial Intelligence. This department would be empowered to pause advanced AI development until a comprehensive regulatory framework is in place and would explicitly prohibit the development of artificial superintelligence.

The proposed approval mechanism within S.5493 is notably detailed. Section 9(d) of the bill stipulates that no individual may deploy, release, import, or transfer an advanced AI system without first obtaining the department’s explicit approval. The Secretary of this new department would be mandated to inspect AI systems for characteristics deemed to pose a danger to the public, including potential superintelligence and precursor indicators. Crucially, approval would be withheld if such dangerous characteristics are identified.

Furthermore, the bill mandates that any entity seeking to develop or distribute advanced AI must obtain a charter. Maintaining this charter would require providing regulators with access to the systems, processes, personnel, and physical infrastructure involved in AI development. The legislation also includes separate provisions that would empower the department to enforce pauses and implement shutdown measures for specified dangerous AI systems.

This legislative framework would effectively insert a federal decision-making authority between AI developers and the release of their models. The institutional structure it outlines bears a striking resemblance to the federal AI department described by Sacks, far more so than a voluntary testing arrangement. However, it is important to note that this remains proposed legislation and does not yet represent an enacted policy or a consensus agreement for the next Congress.

Despite the significant powers vested in the proposed department, the bill does incorporate certain constraints on the regulator. The Secretary of the Department of Artificial Intelligence would be subject to Senate confirmation. The department would also be required to consult with independent scientific advisors and adhere to restrictions on the financial interests of its employees. The legislation further provides for review by the U.S. Court of Appeals for the Federal Circuit of findings underlying any charter revocations. This specific remedy, however, is not framed as a general appeal process for every individual model-release decision.

Building upon the legislative momentum, Abdul El-Sayed’s "AI Under Control" platform, launched on October 13th, advocates for an independent, Cabinet-level agency. This agency would be endowed with the authority to inspect AI laboratories, block releases, order development pauses, and command system shutdowns. The campaign document explicitly acknowledges its foundation in the Sanders-Casar legislation, indicating a convergence of ideas around the need for robust AI oversight.

El-Sayed’s proposal calls for a freeze on the training and release of the most powerful AI systems until established safety rules and enforcement mechanisms are in place. Independent scientists would be tasked with certifying the safety of these systems before any pause could be lifted. The platform also emphasizes the need for independent pre-release audits and mandates that breaches be reported to the government within 24 hours, accompanied by public summaries.

The safeguards embedded within El-Sayed’s proposal are designed to mitigate the risks associated with concentrated regulatory power. A scientific board, selected independently of the agency itself, would oversee its governance. Exemptions granted by the agency would be made public, and industry fees would serve to fund the regulator’s operations. Officials would be prohibited from subsequently accepting employment with the companies they regulate, aiming to prevent regulatory capture. The platform also champions whistleblower protections and the preservation of existing state-level AI regulations.

Crucially, in addressing Sacks’s concerns about surveillance, El-Sayed’s document explicitly reiterates opposition to warrantless surveillance. It also outlines prohibitions against autonomous weapons firing and the replacement of human oversight in life-or-death decisions. These stated prohibitions suggest that the release-blocking powers proposed are intended to address safety risks rather than facilitate the kind of surveillance Sacks alleges.

However, these proposed safeguards do not entirely eliminate the concentration of authority that Voorhees finds objectionable. An agency empowered to halt an AI release, even with external scientific oversight, retains a significant veto power over covered releases. While the safeguards aim to constrain the use of this power, the fundamental ability to block a release remains a potent tool.

The El-Sayed platform does not provide a highly detailed procedure for release approval or appeal. The mechanism for its independent scientific board, selected from outside the agency, also differs from S.5493’s advisory board, whose members would be appointed by the Secretary. While both proposals share the goal of granting coercive powers to a regulatory body, they diverge on the specifics of who would advise or govern that regulator.

Who Would Constrain the Regulator?

The proposed legislative frameworks offer differing approaches to constraining the power of a potential AI regulatory body. The introduced Senate legislation, S.5493, suggests a multi-pronged approach to oversight. This includes the requirement for Senate confirmation of the Secretary, the provision of independent scientific advice to guide decision-making, the implementation of financial restrictions on agency employees to prevent conflicts of interest, and a review process for findings that lead to the revocation of a developer’s charter. These measures are designed to inject accountability and expertise into the regulatory process.

The campaign platform, in contrast, emphasizes the role of an independently selected scientific board in managing the agency’s governance. This approach aims to insulate the regulatory body from undue political influence and ensure that decisions are informed by a broad spectrum of scientific expertise. While both proposals seek to mitigate the risks of unchecked power, they employ distinct mechanisms for achieving this objective.

It is important to reiterate that both the Senate bill and the campaign platform are currently proposals, not an established common system of AI regulation. S.5493 specifically targets defined "advanced AI" systems, and the Federal Circuit review mentioned pertains to findings underlying charter revocations, not a general appeal process for every model release decision. The campaign platform, while outlining broad powers, lacks a detailed procedural framework for approval or appeal.

The Post-Election Landscape: Shifting Political Winds and AI Policy

David Sacks’s reference to the "next Congress" gains particular significance in light of recent electoral shifts. CryptoSlate, citing Associated Press reporting updated on November 8th, indicated that Democrats had secured control of the House and unified control of Congress in the election. This outcome could provide a more fertile ground for the legislative initiatives discussed, potentially accelerating their progress through committees and onto the floor for debate and potential passage.

It is crucial to distinguish existing federal policy from the proposed approval regimes. The executive order issued by President Trump on June 2nd, "Promoting Advanced Artificial Intelligence Innovation and Security," directs the development of a voluntary framework for government access to frontier AI models. However, Section 3(c) of this order explicitly states that it does not authorize mandatory government licensing, preclearance, or permitting for new AI models. This limitation pertains to the specific framework established by that order, not to all potential governmental powers over AI that could be enacted through legislation.

For Erik Voorhees and his venture, Venice, the issue of AI governance is not a new concern but is deeply intertwined with his business model. His July 2024 introduction to Venice, where he is identified as founder and CEO, describes an application leveraging leading open-source AI models. The company’s ethos, as articulated in this introduction, argues against the concentration of intelligence under either corporate or governmental monopoly. This stance aligns with broader discussions within the cryptocurrency and decentralized technology communities, which have been exploring the potential for AI that operates beyond the reach of traditional regulatory structures. CryptoSlate itself has previously covered the connection between restrictions on model access and the rise of decentralized AI solutions in June.

The immediate legislative test will be whether these proposed AI governance frameworks garner sufficient committee action and leadership support in the upcoming Congress. The content of any new legislative text, particularly concerning the specific criteria for release, the allocation of decision-making authority, and the mechanisms for reviewing adverse decisions, will be critical in shaping the future of AI development and deployment in the United States. The debate, amplified by figures like Voorhees and Sacks, underscores the profound implications of these policy choices for innovation, safety, and the very nature of technological progress. The specter of "Sauron" – a concentrated, unchecked power – looms large in the minds of many, driving the urgent need for thoughtful and balanced governance.