Current Affairs Topics Archive
International Relations Economics Polity & Governance Environment & Ecology Science & Technology Internal Security Geography Social Issues Art & Culture Modern History

Anthropic at loggerheads with US military over AI use

February 24, 2026 5 min read Science & Technology Internal Security GS3 (Science & Technology Internal Security) GS2 (International Relations)
On This Page ▼

What Happened

  • A public conflict erupted between AI company Anthropic (maker of the Claude family of AI models) and the US Department of Defense (DoD) over contract terms governing military use of its AI systems.
  • The DoD's position, driven by Defence Secretary Pete Hegseth's January 2026 AI strategy memorandum, demanded that all DoD AI contracts adopt "any lawful use" language — effectively removing all corporate-mandated ethical constraints from AI systems used by the military.
  • Anthropic broadly agreed to military use of Claude but drew firm lines against deployment for autonomous lethal weapons systems and mass domestic surveillance.
  • In late February 2026, Hegseth met Anthropic CEO Dario Amodei at the Pentagon and issued a deadline: accept the new contract language by 27 February 2026 or lose the contract.
  • On 5 March 2026, the Trump administration formally designated Anthropic a national security "supply chain risk," barring it from federal contracting — even as Claude was reportedly being used to assist US military operations against Iran.
  • Anthropic filed suit in federal court in California on 9 March 2026, arguing the designation was unlawful and violated its First and Fifth Amendment rights.

Static Topic Bridges

AI Ethics Frameworks: Corporate Self-Regulation vs. State Mandate

The rapid commercialisation of large language models (LLMs) has produced a tension between corporate AI safety commitments and state demands for unrestricted deployment in sensitive domains. Leading AI companies typically publish "acceptable use policies" or "usage policies" that prohibit uses such as developing weapons of mass destruction, generating child sexual abuse material, or enabling autonomous lethal decision-making. These policies represent a form of voluntary, corporate-level AI governance. The Anthropic-DoD dispute is the most prominent case to date where a state actor formally challenged such constraints as barriers to national security interests.

  • Anthropic's Constitutional AI approach trains Claude to be "broadly safe, broadly ethical, adherent to Anthropic's principles, and genuinely helpful" — in that order of priority.
  • OpenAI, by contrast, reached a separate agreement with the DoD in the same period, accepting "any lawful use" language — illustrating how different firms navigate the same tension differently.
  • The EU AI Act (2024) is the world's first comprehensive binding AI law; it classifies certain AI uses in law enforcement and military contexts under "unacceptable risk" categories.

Connection to this news: The dispute reveals the structural weakness of voluntary corporate AI ethics frameworks when confronted with state power and procurement leverage — a governance gap that international regulators are racing to address.

Lethal Autonomous Weapons Systems (LAWS) and International Law

Lethal Autonomous Weapons Systems (LAWS) — colloquially called "killer robots" — are weapon systems capable of selecting and engaging targets without meaningful human control. The international community has debated LAWS regulation since 2014 at the UN Convention on Certain Conventional Weapons (CCW) Group of Governmental Experts. In December 2023, the UN General Assembly passed its first resolution on LAWS, endorsed by 152 nations, calling for new international norms. Critics including the International Committee of the Red Cross (ICRC) argue LAWS would violate International Humanitarian Law (IHL) principles of distinction (between combatants and civilians), proportionality, and precaution.

  • The Campaign to Stop Killer Robots — a coalition of over 270 NGOs in 70 countries — advocates for a binding international treaty banning fully autonomous weapons.
  • India's position: India voted against the December 2023 UN LAWS resolution, arguing a blanket ban could stigmatise beneficial technologies and that precision autonomy may actually reduce collateral damage. India supports a political declaration rather than a binding treaty.
  • The US, UK, Russia, and China have all resisted a binding ban on LAWS.
  • No binding international treaty specifically governing LAWS exists as of 2026.

Connection to this news: Anthropic's refusal to permit Claude's use in autonomous lethal weapons systems directly engages the core controversy in the LAWS debate — the question of meaningful human control over life-and-death decisions.

India's AI Governance Landscape

India does not yet have a dedicated AI law. Its governance approach is advisory and principle-based. The Ministry of Electronics and Information Technology (MeitY) released India AI governance guidelines under the IndiaAI mission. NASSCOM, the apex body for India's IT industry, has developed a Responsible AI Resource Kit providing self-regulatory guidance for industry actors, emphasising fairness, transparency, privacy, and accountability. The Digital India Act (proposed successor to the Information Technology Act, 2000) is expected to address AI liability and accountability in some form.

  • IndiaAI Mission (launched 2024): ₹10,000 crore outlay over five years for AI compute infrastructure, datasets, and application development.
  • NASSCOM's Developer's Playbook on Responsible AI: enables AI developers to identify and mitigate risks associated with commercial AI deployment in India.
  • India has a stated policy of not developing biological, chemical, or nuclear weapons; its position on autonomous weapons remains one of engagement rather than prohibition.
  • The Personal Data Protection Act (DPDP Act, 2023) governs data that feeds AI systems in India.

Connection to this news: As India becomes a significant consumer and developer of AI systems — including for defence applications — the Anthropic-Pentagon standoff offers a cautionary illustration of what happens when AI governance relies entirely on voluntary commitments.

Key Facts & Data

  • Anthropic founded: 2021 by Dario Amodei, Daniela Amodei, and others (former OpenAI researchers)
  • Claude AI models: Anthropic's flagship LLM family (Claude 3, Claude 3.5 series)
  • US DoD "any lawful use" mandate: January 2026 AI strategy memorandum (Hegseth)
  • Pentagon-Anthropic deadline: 27 February 2026
  • "Supply chain risk" designation of Anthropic by Trump administration: 5 March 2026
  • Anthropic federal lawsuit filed: 9 March 2026
  • UN LAWS resolution (UNGA): December 2023 — 152 in favour, India voted against
  • India voted against the UN LAWS resolution; supports political declaration, not binding treaty
  • EU AI Act: entered into force 2024; world's first binding comprehensive AI law
  • India's IndiaAI Mission budget: ₹10,000 crore over 5 years