Inside the Biden Administration’s AI Policy Approach

Podcast: 80,000 Hours Podcast (Crosspost w/ The Cognitive Revolution)
Guests: Jake Sullivan (Former US National Security Advisor), Nathan Labenz
Release Date: September 26, 2025
Duration: ~66 minutes

Episode Overview

This episode dives deep into US national security, AI risk, regulation, and US-China technological competition, featuring Jake Sullivan, former National Security Advisor to President Biden and inaugural Kissinger Professor at Harvard. Sullivan unpacks both the operational realities and philosophical challenges confronting the US government during a time of accelerated AI advancement, especially under the Biden administration and into Trump’s return. He shares rare candid insights on how AI factors into national security, the pace of technological adoption in government, export controls, international alliances, and the prospects for US-China relations.

Key Discussion Points and Insights

1. How "AGI-Pilled" is US AI Policy?

• Uncertainty as Policy: Sullivan articulates that rapid advances in powerful AI (AGI) are a distinct possibility within years, but far from certain. Policy must be built with this uncertainty in mind.
  • Quote: “I believe that possibility of transformative powerful AI coming in the very near future, that is in the next couple of years, is something we have to take very seriously... But do I believe it's an inevitability? No.” [03:07]
• He sees AI as on par with other urgent national security priorities because “it touches everything: the future of warfare, competitiveness, threats from non-state actors.” [03:48]
• While he takes existential AI concerns seriously, he’s more focused on concrete, immediate security, economic, and societal risks, rather than speculative, “robot apocalypse” scenarios. [11:44]

2. How DC—and Especially the White House—Gets Its Information

• Sullivan describes a wide net: Meeting with every frontier lab, skeptics, VCs, academics, and convening the entire 'national security enterprise.' [05:13]
• Critiques Congress as lagging: “I think there is still a gap between where Congress is and where it needs to be in terms of a sense of urgency.” [06:33]
• Tech demos make an impression, but “Congress really only tends to react when there are real world impacts that could be a catastrophe... it will be a collection of those moments that is really what's going to get Congress into action more than just demos of the tech.” [08:21]

3. Regulatory Approaches—Flexibility and Iteration

• Sullivan doubts that “trying to write a long list of rules or standards is an ill fated undertaking.” [10:54]
• Calls for “preparation and flexibility: being ready, developing a set of rules and standards, of tools that Congress or the executive branch could use as necessary.” [09:58]
• Emphasizes the need for “careful dialogue between Washington and Silicon Valley and the rest of the country.” [10:40]

4. The “Four Buckets” Framework

• Security: Risks from non-state actors (cyber, bio), state actors (wonder weapons, military advantage).
• Economics: Job displacement, concentration of wealth/power, but also massive benefits like productivity, solving public issues.
• Society: Risks (misinformation, alienation) vs. opportunities (AI as teachers, doctors, personalized care).
• Existential: “Too imponderable at the moment” to make strong policy; focus on getting the first three right will build muscles for existential risks.
  • “I try to do is think, all right, let's get very concrete about the risks, identify them...think very concretely about the opportunities. And then are there steps the government can take to increase or expand the opportunities and to manage and minimize the risks. If the answer to that is yes, act. If it's not, then don't act.” [11:44]

5. Current State of Government AI Adoption

• Both the Biden and Trump administrations lag in adopting frontier AI tools at the highest levels.
  • “There’s like still no ChatGPT or anything similar at the White House… not great baseline from which to try to— not dictate, of course, but lead the national charge toward global AI leadership.” [21:14]
• Cultural conservatism and security/legal risks slow adoption, especially on classified systems.
  • “The culture of government in terms of new technology adoption is… historically slow.” [22:12]
• Sullivan: “I would be an advocate for seeing this get incorporated as rapidly as possible.” [23:15]

6. Export Controls, Alliances, and Middle East AI

• Sullivan favors selling American chips and building data centers worldwide (“diffusion”), but insists on “high security standards”—no diversion of chips to China, no excessive dependence on the Middle East or any region for critical compute. [23:55]
• Critiques Trump administration’s “dissonance” between hawkish rhetoric and actual export control policy. [19:56]
• Strongly against frontier training runs being done offshore—especially in the Gulf. “I want them done here… We should do them here in the United States.” [27:28]
• Less concerned about “value alignment” with partners like the UAE than with China—values must be considered but not a red line. [28:12]

7. Why China Is Different: The US-China AI Competition

• Sullivan distinguishes China from other illiberal partners:
  • China alone “has the attributes and capacities to compete with the United States… with an ambition to surpass the US as the world’s leading power in every sphere.” [33:10]
• Lays out four specific concerns:
  1. Industrial Overcapacity/“China Shock 2.0” – flooding markets with subsidized tech/AI (David Autor NYT reference). [29:43]
  2. Speech Control & Influence – Cites NBA/Hong Kong episode; risk if “the entire world is run on Chinese AI.” [32:10]
  3. Cyber Threats/Malware Prepositioning – “Volt Typhoon” and critical infrastructure risk. [33:10]
  4. Military Build-up/Taiwan – China is “engaged in the world’s largest peacetime military buildup.” [34:00]
• “I am determined that competition not turn into conflict…We inaugurated an AI risk dialogue with China to talk to them directly about issues that challenge both our countries.” [35:05]

8. On Export Controls and the “Small Yard, High Fence” Doctrine

• Export controls are not an embargo, and not a regime change tool. They’re to “ensure strategic advantage” at the frontier.
  • “Why are we giving (China) that input to be used against us or our allies? We shouldn’t do that.” [41:32]
• Acknowledges the dual-use dilemma: China’s “civil-military fusion” makes it impossible to send chips for peaceful but not military use.
  • “Are we trying to deny access to AI doctors in China? Absolutely not. ...I do not believe that it will.” [43:52]

9. On Calls for ‘Regime Change’

• Rejects the Dario Amodei/Machines and Loving Grace proposal for leveraging technology to force China into a “grand bargain” or to give up competition.
  • “I am not a regime change guy when it comes to China…Our job should not be to shape China’s government.” [48:06]

10. Global Soft Power & The AI Race

• China’s open-source, “AI for all” narrative is attractive, especially to the Global South, but “most countries are pretty unsentimental…Another element of soft power is innovation and technological prowess. …A lot of countries are betting on the US to stay in the lead.” [53:41]
• The US’s edge: “The American tech stack…is damn good and we think the United States is going to stay at the cutting edge.” [54:27]

11. Stability and the Future of AI Arms Control

• Sullivan foresees “a version of managed competition” between the US and China:
  • “We’re going to live alongside one another as major powers, but where there is going to be intensive competing jostling…with sufficient guardrails, doesn’t spill over into conflict.” [55:47]
• Arms race risk is acknowledged, but ambiguity and gradualism are preferable to hard, upfront commitments on “strategic dominance.” [58:36]
  • “A little bit of ambiguity…in this phase is manageable and is better than explicit declarations …on concepts like strategic dominance.”
• Skeptical of “grand bargain” solutions—prefers a world of “managed competition” as the durable equilibrium. [60:40]

12. AI and the Future of Warfare

• Sullivan sees in Ukraine “glimpses of the future”—autonomous systems, scale, attritability, and potential for “abundance” to redefine conflict.
• Concerned that US defense bureaucracy is less nimble than China’s PLA at adopting AI:
  • “My biggest concern fundamentally is that the United States military, the Pentagon, because of the bureaucracy...is not necessarily poised to adopt, you know, at scale and with speed, AI capabilities. And I think the PLA is better situated to do that. That makes me nervous.” [62:15]

Notable Quotes & Memorable Moments

• On AI Uncertainty:

  • “Super smart people make the case that it’s coming very fast. And other super smart people make the case actually there’s a number of things that have to happen and those things may take quite a long time before we get to something like AGI.” [03:29]

• On US Government Adoption Pace:

  • “You cannot effectively lead and govern on this set of issues if you aren’t an effective user of them for purposes in advancement of America's interests.” [23:10]

• On Diffusion vs Dependence:

  • “We were long dependent on the Middle East for oil. We should not become dependent on the Middle East or anywhere else for compute.” [24:22]

• On Managing US-China Competition:

  • “Compete like hell, but intensively manage that competition so it didn’t tip over into conflict...And if we continue to build the muscles for it, I think it can withstand even the advent of very powerful artificial intelligence. How’s that for optimistic.” [55:47 + 57:10]

• On AI and Economic Impact:

  • “This is the first technology I can think of with such profound national security applications that the government really had very little to do with. That’s essentially a private sector led and driven technology.” [64:30]

• On Export Controls:

  • “Why are we giving [China] that input to be used against us or our allies? We shouldn’t do that. That’s how I look at it.” [41:32]

Timestamps for Important Segments

• Setting the Stage / Sullivan’s AGI Worldview: [02:19–04:34]
• How Washington Gets Its Cues: [05:13–07:46]
• On Congress & Sensational AI Demos: [07:46–09:58]
• Four Buckets Framework—Security, Economics, Society, Existential: [11:44–15:18]
• Current State of AI Adoption in Government: [21:14–23:26]
• Export Controls, Middle East Deals, and Compute Dependence: [23:26–28:52]
• The China Problem—Concrete Risks and Competition: [29:43–35:35]
• Export Controls Debate / Japan Analogy: [37:08–41:32]
• Regime Change? Dario & US Policy: [48:06–49:21]
• Global AI Soft Power Struggle: [53:41–54:51]
• Future AI Arms Control/Stable Equilibrium: [55:47–60:22]
• AI and Modern Warfare: [62:15–63:55]
• Final Reflections on Economic Impact & Policy Challenges: [64:13–65:33]

Tone and Language

Sullivan remains measured, occasionally self-effacing (“boring, I confess…”), fact-oriented, rarely speculative. He pushes for practical, concrete steps over grand visions, preferring preparedness, humility, and managing risk across wide uncertainties.

Final Thought

Sullivan’s blueprint is “managed competition”—compete fiercely, avoid war, build diplomatic muscles to absorb the shocks of rapid, transformative technology. He expresses both humility about the unknowns of AI and urgency about present actionable risks, favoring adaptability over ideology as the best hope for a stable AI future.