Practical AI Episode: AI Policy and the Battle for Computing Power Date: March 9, 2026 Guest: Ben Buchanan (Johns Hopkins SAIS, former White House Special Advisor on AI)

Episode Overview

This episode dives deep into the interplay of artificial intelligence (AI) policy, global competition, and the vital importance of computing power. Host Chris Benson (Lockheed Martin) welcomes Ben Buchanan, a renowned AI policy expert, professor, and former White House Advisor, to discuss how the public and private sectors shape AI’s future, the geopolitical stakes around computing hardware (specifically semiconductors), and the emerging guardrails around AI’s responsible use. The conversation blends historical context, technical detail, and actionable policy insights—grounded in both US national security and global impacts.

Key Discussion Points and Insights

1. Ben Buchanan’s Background and Unique Perspective (01:30–03:05)

• Buchanan shares how he “accidentally” entered the AI field:

  “The cool thing, at least in my view, about a lot of jobs I had is that they didn’t exist before I had them.”
  — Ben Buchanan (01:57)

• His journey from studying cyber operations to AI policy, emphasizing the shift from government-led to private-sector-led innovation, setting up the episode’s central theme.

2. How AI Differs from Past Revolutionary Technologies (03:31–07:21)

• Buchanan highlights AI as the first major tech revolution in a century driven primarily by the private sector, not the government or military.

• Comparison with nuclear age, space age—AI’s private-sector origins are a “vexing challenge for US government,” making policy guidance and oversight more complex.

• Current policymakers often lack direct technical experience, requiring effective translation from experts:

  “My job [in the White House] was to explain things and to say, how can we put this in terms that make sense to policymakers?... Also, how can we view this not just as a scientific or technological question, but as a geopolitical question?”
  — Ben Buchanan (05:49)

3. Computation (“Compute”) vs. Data: The True AI Bottleneck (09:50–12:39)

• Buchanan breaks the myth that “data is the new oil,” arguing that access to immense computation is the primary accelerator for modern AI.
• References OpenAI’s “Scaling Laws” paper—“The more computing power you use to train an AI system, the more powerful the resulting AI system.” (10:25)
• Control of the advanced chip (semiconductor) supply chain—mostly held by democratic countries (Taiwan/TSMC, Netherlands/ASML, USA, Japan)—is a decisive strategic advantage.

4. Taiwan’s Central Role and Geopolitical Stakes (13:46–16:27)

• Buchanan stresses that Taiwan’s semiconductor industry is essential—not just for AI, but for the global economy and security:

  “Something like 97% of the advanced computer chips in the world are made in Taiwan...It is very fortunate for democracies that, maybe as a historical accident...they own the computing supply chain.”
  — Ben Buchanan (10:25)

• Loss of Taiwanese chip output would cost trillions, triggering massive geopolitical consequences.

• Recent US policies (e.g., CHIPS and Science Act) aim to bring chip manufacturing stateside, though Taiwan remains the indisputable leader.

5. Communicating AI Policy: Students vs. Policymakers (16:27–17:36)

• Contrast in engagement: policymakers want actionable briefings (“what’s happening now?”), while students can dive deep into theoretical explorations.

• Both audiences fundamentally ask: Where is this technology going? What does it mean for humanity and democracy? What should we do now?

  “If you’re asking, do my graduate students know more than Congress does? The answer is yes.”
  — Ben Buchanan (16:27)

6. Bipartisanship (and Partisanship) in US AI Policy (18:20–22:06)

• Historically, AI policy has not been highly polarized; there’s been significant bipartisan cooperation, especially around export controls for advanced chips.
• Trump and Biden administrations shared more similarities than differences on some AI-export and national security issues, but recent policy shifts (especially around export controls and “letting the private sector cook”) are emerging.

7. Government–Private Sector Dynamics: Chips and Models (22:06–24:51)

• US government sees advanced chips as both powerful and scarce; policy has been to restrict exports to adversary nations (notably China) to maintain a technological edge.
• Collaboration is encouraged between government (especially DoD and intelligence) and AI companies, but with guardrails for safety and values.

8. Guardrails and Values in AI Deployment (24:51–27:35)

• Application-specific judgment: Autonomy is acceptable in some domains (e.g., missile defense, cyber operations), but regulation/policy must vary accordingly.

• US-led effort to build international consensus: “Political Declaration on the Use of Autonomy in Military Systems” with 58 countries agreeing on principles.

  “Wherever we decide as a nation to draw the lines, it’s vitally important that we go...set that aside [internationally], the norms and standards with the rest of the world.”
  — Ben Buchanan (27:18)

9. Speed vs. Caution: The Regulatory Balance (29:30–32:42)

• Buchanan uses the historical analogy of railroads—safety innovations ultimately allowed trains to go both faster and safer.

• Argues that opportunity and safety are not conflicting:

  “My view is that we get AI opportunity through AI safety...through developing technology that is safe, secure, and trustworthy and people can trust.”
  — Ben Buchanan (30:24)

• Warns against a “race to the bottom” between nations on safety, advocating a strong democratic lead to better coordinate safety frameworks.

10. International Coordination and Evolving Global Order (34:22–36:44)

• International standards benefit US business and safety, but are getting harder to achieve as geopolitical ties weaken.
• US has driven several multi-lateral AI initiatives (G7, UN) but fraying alliances could “hurt us and hurt our businesses.”
• Positive diplomatic engagement, even with rivals, is necessary; democratic preeminence is ideal, but inclusion is critical for global tech affecting all.

11. Chinese Innovation, Open Models, and Compute Limits (36:44–40:59)

• Recent surge of open-source AI models from Chinese teams is noted—these are often trained on “smuggled or stockpiled American chips” and constrained by compute access.
• Emphasizes US and allies’ continued strategic advantage in compute.
• Buchanan’s stance: “Chinese developers are very talented...But it is the case that computing power remains incredibly important, in fact probably the most important US advantage.” (38:50–40:59)

12. AI and Cyber Operations: The Next Battlefield (41:51–43:58)

• AI’s influence on cyber operations is immediate and dual-use (defense/offense), with implications for both intelligence and daily life.
• AI vastly improves vulnerability discovery—Anthropic’s latest models found “500 high severity vulnerabilities in open source software” (42:36).
• The US (DARPA, etc.) is actively pushing to maintain the edge here.

13. Measuring Democratic Success in the Age of AI (45:35–47:48)

• Buchanan gives a three-pronged framework for “winning” as a democracy in AI:

  1. Invention: “Are we inventing the technology and pushing it forward, bending it...towards safety and justice?”
  2. Adoption: “Are we adopting this technology...in national security, economy, and prosperity?”
  3. Values: “Are we using this technology in accordance with our values...domestically, internationally, and in terms of democratic governance?”

  “We have all the winning cards on invention...The only way America loses is if it folds and does things like sell chips to China...But...how do we make sure AI is advancing rather than undermining democratic values? ... Those are the three things I would use.”
  — Ben Buchanan (45:35)

Notable Quotes & Memorable Moments

• On Analogies in Explaining AI:

  “To resist metaphor is to endure the thing itself. And I always just say we have to endure AI itself.”
  — Ben Buchanan (08:23)

• On Global Chip Power:

  “Making a computer chip, in my view, is the hardest thing we do as a species.”
  — Ben Buchanan (10:25)

• On Bipartisanship:

  “I don’t see [AI policy] as a partisan thing...the delta between Trump 2 and Trump 1 is much bigger than the delta between Trump 1 and Joe Biden's administration…”
  — Ben Buchanan (20:09)

• On Speed vs. Safety:

  “We get AI opportunity through AI safety...not incredibly cumbersome regulations...but through developing technology that is safe, secure, and trustworthy.”
  — Ben Buchanan (30:24)

• On Democratic Success:

  “Are we inventing [AI]? Are we adopting [AI]? Are we using [AI] in accordance with our values?”
  — Ben Buchanan (45:35)

Key Timestamps

• [01:57] – Ben Buchanan’s non-traditional journey into AI policy
• [03:31] – Private sector’s central role in modern AI
• [09:50] – Why compute, not data, is the bottleneck
• [13:46] – The importance of Taiwan and the chip supply chain
• [16:27] – Teaching students vs. briefing policymakers
• [18:20] – Bipartisan nature of AI policy, past and present
• [22:06] – Export controls and government–tech relationships
• [24:51] – Guardrails for AI, international collaboration
• [29:30] – Speed vs. caution: lessons from the railroad era
• [34:22] – Global order, alliances, and the challenge of AI governance
• [36:44] – Chinese models, compute constraints, US advantages
• [41:51] – AI’s impact on cyber operations
• [45:35] – Framework for measuring democratic “victory” in AI

Summary: Why This Episode Matters

This episode gives listeners an unvarnished look into how AI policy, power, and national security intersect at the highest levels—and why global leadership in compute and democratic values is critical for the next decade. Whether you’re a technologist, policymaker, or simply an engaged citizen, Buchanan and Benson provide a nuanced, actionable roadmap for understanding AI’s real-world impact—far beyond the buzzwords.

For more insights and future episodes, visit PracticalAI.fm or connect on LinkedIn, X, or Bluesky.