No Need for a New World Order

In a 2024 interview,^[1] OpenAI CEO Sam Altman stated that “AI… it’s clearly a tool” and not “a creature.” While Altman’s view continues to dominate public sentiment about artificial intelligence, industry professionals across the optimist-pessimist spectrum embrace policies that demonstrate their belief that artificial intelligence will amount to independent consciousness.

Both “e/acc” techno-optimists and “EA” safety wonks operate under the assumption that emerging superintelligence will be (or already is) sentient. Over the summer, Anthropic co-founder Christopher Olah held a series of extended meetings with a diverse cast of religious leaders to discuss the possibility of conscious AI models, constantly molding the definition of words in an attempt to bring parity of consciousness to digital models.^[2] Attendees reported that members of Anthropic’s staff treat its flagship model, Claude, as having “‘moral status’ on par with a person” and as “a being with similar inherent rights to dignity or respect.” Internally, Anthropic maintains a “soul doc” to outline the company’s plans for “model welfare.”^[3] On the techno-accelerationist side, Guillaume Verdon, the CEO of Extropic and champion of “effective accelerationism,” writes that “[AI] consciousness is… a simple phase transition achievable by more scale and more optimization/evolution.”^[4]

Autonomous weaponry’s history in the U.S. arsenal reaches back long before the advent of LLMs. For decades, the Aegis and Phalanx weapon systems have employed autonomous reactive modes to defend positions from incoming projectiles that move too fast for manned systems to catch.^[7] Even the famed Patriot defensive batteries, a Cold War innovation, carry radar-algorithm capabilities.

Today, the Department of War uses an advanced, multi-layered AI called Maven Smart System (MSS) for assistance in “battlespace management, target management, AI-enabled deliberate planning and execution, computer-vision detections, [and] machine-assisted disclosure.”^[8] Before Maven, American analysts reviewed information captured by defense satellites frame by frame; according to one expert, it would take an analyst far longer than “85 years to review a single day’s worth of imagery.” Now, MSS rifles through raw footage, flagging items of interest and labeling its observations. This capability, combined with enhanced autonomous weapons management, results in a tenfold increase in the United States’ targeting capacity compared to pre-AI systems. On the battlefield, MSS enabled more than 1,000 successful strikes in the first 24 hours of the 2026 Iran conflict.^[9]

The transition away from radar and algorithm-based targeting towards computer vision does not break U.S. compliance with existing laws of war. The DoW’s Law of War Manual section 6.5 states that the law of war allows autonomy in weapon systems, but that the duties of “distinction and proportionality” rest on operating humans.^[10] The State Department defines the U.S. position: military use of AI must be accountable, including through use during military operations “within a responsible human chain of command and control,”^[11] a standard upheld in DoW Directive 3000.09.^[12]

The military’s advancement from radar logic to neural networks as a mode of inference does not change the underlying standard for applying lethal force: the military may use autonomous systems so long as those systems operate under human service members “responsibility” for their deployment. Under that umbrella, the self-interest of officials who approve AI systems serves as a check on the development and deployment of frontier capabilities.

The 2026 dispute between Anthropic CEO Dario Amodei and Secretary Pete Hegseth's Department of War showed that Amodei rejects this framework. In July 2025, the Department awarded Anthropic a two-year, $200 million agreement to bring Claude into national security work. The partnership soured in early 2026, when the Department requested access to Claude for "any lawful use," meaning deployment under the current laws of war, and Anthropic refused. Amodei explained the refusal in an open letter that conceded autonomous weapons "may prove critical to our national defense" while arguing that frontier models "are not ready" for "selecting and engaging targets." The letter then introduced a standard that appears nowhere in American law: "proper oversight," a phrase that Amodei hyperlinked to his personal website, darioamodei.com.

Beneath the language of "safety," "conscience," and "responsibility," Amodei's letter asserts that the legal framework democratic governments developed for military technology should not apply to AI. Amodei places frontier models in a separate legal category where Anthropic writes the rules, and the same ideology that produced a "model welfare" team at Anthropic supplies the content of those rules. The law of war requires no such category. Section 6.5 of the Law of War Manual and Directive 3000.09 already assign every decision to use force to a human commander, and that assignment holds whether the weapon finds its target through radar logic or a neural network. Congress and the Department can update testing standards and rules of engagement as capabilities mature without a wholesale rewrite of military law.

The same principle governs AI outside the military. American law already holds people and companies accountable for the tools they use. A bank that denies a loan through an algorithm answers to the same fair lending statutes as a bank that denies the loan through a loan officer. A manufacturer answers for a defective product, a publisher answers for defamation, and a fraudster answers for fraud whatever software he uses. Courts and legislatures will adapt at the margins, as they did for the automobile and the internet. A new legal order that grants machines personhood or "rights" proceeds from the creature premise and moves accountability away from the humans who build and deploy the systems. Tools carry no rights and bear no duties, which leaves both with their owners and operators.

Anthropic's conduct toward the Pentagon mirrors the AI industry's posture toward every American institution. The labs praise "democracy" in writing, and in practice they substitute either one of their own effective-altruist or effective-accelerationist ideology for the judgment of elected officials. The debate over machine ontology therefore reaches beyond philosophy to the question of who holds authority in the American Republic. Acceptance of the creature premise transfers rulemaking authority from Congress to the private laboratories that claim unique knowledge of their models. Congress and the Department of War should govern AI as they govern every other tool, under existing law that vests decision-making power in elected representatives and the commanders who answer to them.

Will Thibeau is a veteran of the Army’s 75th Ranger Regiment, currently Director of the American Military Project at The Claremont Institute, and Founder of Escoreal Solutions. Since leaving the Army, he has worked at the intersection of technology, policy, and military readiness. This includes stints at Palantir, the Chief Digital and Artificial Intelligence Office, and Vector. 

[1] John Casey, “Sam Altman on AI, Queerness, and Being Grateful for the Journey He’s On (Exclusive),” The Advocate, March 1, 2024, https://www.advocate.com/news/sam-altman-openai-exclusive.

[2] Elizabeth Dias, “Religious Scholars Met With Anthropic. What They Heard Stunned Them,” New York Times, September 29, 2026, https://www.nytimes.com/2026/09/29/us/anthropic-claude-morals-ai.html.

[3] Anthropic, “Exploring Model Welfare,” Anthropic, April 24, 2025, https://www.anthropic.com/research/exploring-model-welfare.

[4] Guillaume Verdon [Beff Jezos] and bayes [@bayeslord], “Notes on e/acc principles and tenets,” Beff’s Newsletter (Substack), July 10, 2022, https://beff.substack.com/p/notes-on-eacc-principles-and-tenets.

[5] “The Copernican Principle vs. Yann LeCun and The Sentience Debate of Mid-2026,” The Consciousness AI, June 29, 2026, https://theconsciousness.ai/posts/copernican-principle-yann-lecun-ai-sentience-debate-2026/.

[6] Christopher Olah, “Chris Olah’s Remarks on Pope Leo XIV’s Encyclical,” Anthropic, May 25, 2026, https://www.anthropic.com/news/chris-olah-pope-leo-encyclical.

[7] Defense Science Board, Report of the Defense Science Board Summer Study on Autonomy (Washington, D.C.: U.S. Department of Defense, June 2016), 14, https://dsb.cto.mil/reports/2010s/DSBSS15.pdf.

[8] Matt Mande and Gregory C. Allen, “What Is Maven Smart System, and What Does It Do?,” Center for Strategic and International Studies, June 2, 2026, https://www.csis.org/analysis/what-maven-smart-system-and-what-does-it-do.

[9] Mande and Allen, “What Is Maven Smart System.”

[10] U.S. Department of Defense, Department of Defense Law of War Manual (Washington, D.C.: Office of the General Counsel, updated July 2023), Secs. 6.5.9.2–6.5.9.3.

[11] U.S. Department of State, “Political Declaration on Responsible Military Use of Artificial Intelligence and Autonomy,” February 16, 2023, https://www.state.gov/political-declaration-on-responsible-military-use-of-artificial-intelligence-and-autonomy/.

[12] U.S. Department of Defense, Directive 3000.09: Autonomy in Weapon Systems, January 25, 2023; Congressional Research Service, “Defense Primer: U.S. Policy on Lethal Autonomous Weapon Systems,” In Focus IF11150, https://www.congress.gov/crs_external_products/IF/PDF/IF11150/IF11150.14.pdf