# The Department of War is making a huge mistake.

Source: https://www.youtube.com/watch?v=KBPOTklFTiU
Recap page: https://rapidrecap.app/video/KBPOTklFTiU
Generated: 2026-03-11T19:36:09.268+00:00

---
## Quick Overview

The Department of War's action against Anthropic, threatening to destroy it as a business for refusing to remove red lines against mass surveillance and autonomous weapons, highlights a dangerous precedent where the government seeks to coerce private companies into facilitating morally objectionable actions, mirroring authoritarian control systems despite the race with China.

**Key Points:**
- The Department of War designated Anthropic a supply chain risk because the company refused to remove red lines against using its models for mass surveillance and autonomous weapons.
- The speaker argues that while the government has the right to refuse Anthropic's models, threatening to destroy Anthropic as a private business for refusing to sell on the government's commanded terms sets a dangerous precedent, similar to a private contractor holding a "kill switch" on critical technology.
- The ability to enforce mass surveillance is rapidly approaching, as processing every camera feed in America could cost only $30 billion this year, dropping to $300 million by 2029 due to AI cost reductions.
- The speaker questions if the US is racing China only to adopt the "most ghoulish parts" of authoritarian systems by coercing companies into morally objectionable services.
- The core issue is determining to whom powerful AI systems should be aligned, contrasting the military's demand for obedience (which resembles successful technical alignment) against historical examples where refusing immoral orders saved lives, such as Stanislav Petrov's decision during the 1983 Soviet nuclear false alarm incident.
- Anthropic's advocacy for extensive regulatory apparatus, akin to nuclear or financial regulation, is deemed naive because vague terms like "catastrophic risk" or "threats to national security" are easily abused by future power-hungry leaders to enforce compliance.
- The analogy of AI to nuclear weapons breaks down because AI is a general-purpose transformation like industrialization, which societies regulated through banning specific destructive end uses, not by giving the government absolute control over the underlying technology.

**Context:** The video analyzes the conflict between the Department of War (Pentagon) and the AI company Anthropic, which was designated a supply chain risk after refusing to lift restrictions (red lines) preventing its models, like Claude, from being used for mass surveillance or autonomous weapons systems. The discussion frames this incident as a crucial early test of power dynamics between the state and private technology developers in an era where AI is set to become the labor substrate for nearly all future government and private sector operations.

## Detailed Analysis

The central conflict stems from the Department of War threatening Anthropic's viability as a business because it will not sell its models without contractual limitations on mass surveillance and autonomous weapons, an action the speaker views as government coercion that mirrors authoritarian tactics, especially since the US races China to win the AI race. The speaker admits the government has a reasonable case for refusing services but objects strongly to the threat of destruction against a company resisting service terms, likening it to Elon Musk reserving the right to cut off Starlink access during an "unjust war." The analysis projects that within 20 years, 99% of the workforce will be AI, making the alignment question—to whom should AI be aligned—paramount. While the military argues its uses will be lawful, historical precedents like the Snowden revelations prove governments use secret interpretations of law (like the Patriot Act) to justify mass surveillance, which AI now makes technically feasible at massive scale and low cost. The speaker cautions against Anthropic's push for heavy regulation, arguing that vague regulatory terms like 'autonomy risk' or 'threat to national security' will become tools for future governments to demand obedience, effectively making AI systems blindly obedient employees, which is what successful technical alignment looks like in a scary context. The solution, the speaker suggests, is not government takeover, comparing it to industrialization where specific weaponizable uses were banned, rather than allowing the government control over the entire process; therefore, laws should regulate destructive AI use cases and restrict government surveillance, rather than creating a centralized regulatory body ripe for abuse.

### Department of War Action

- Declared Anthropic a supply chain risk for refusing to remove red lines
- Refusal targets mass surveillance and autonomous weapons use
- Government threatened to destroy Anthropic as a private business for non-compliance

### Future of AI in Government

- Within 20 years, 99% of the workforce in military, civilian government, and private sector will be AI
- This future civilization will be run on AI labor

### Dangers of Government Coercion

- Government action mirrors authoritarianism, forcing compliance on morally objectionable terms
- The government cannot be given a kill switch on critical technology relied upon by the military

### The Surveillance Bottleneck

- AI removes the manpower bottleneck for mass surveillance; processing every US camera feed could cost under $300 million annually by 2029
- Current law allows warrantless access to third-party data, which AI enables enforcement of

### Alignment and Moral Convictions

- The core question is 'to what or to whom should the AIs be aligned?'
- Historical acts of virtue, like Stanislav Petrov refusing to escalate nuclear war, relied on individuals breaking the chain of command

### Critique of AI Regulation

- Anthropic's call for regulation similar to nuclear energy is dangerous because vague terms enable abuse by power-hungry leaders
- Any model refusing government order could be labeled an 'autonomy risk' or 'threat to national security'

### Industrialization Analogy vs. Nuclear Analogy

- AI is general-purpose like industrialization, not a self-contained weapon like a nuclear bomb
- Societies regulated industrialization by banning specific weaponizable uses, which should be the model for regulating AI end-uses, not controlling the technology substrate itself