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

Pentagon asks defence contractors about reliance on Anthropic's AI services, source says

February 26, 2026 5 min read Science & Technology Internal Security GS Paper 3 (Science & Technology Internal Security Cyber Security)
On This Page ▼

What Happened

  • The United States Department of Defense (Pentagon) reached out to major defence contractors — including Boeing and Lockheed Martin — asking them to assess and report their reliance on Anthropic's Claude AI model, in a first step toward designating Anthropic a "supply chain risk."
  • The Pentagon's concern stems from a policy dispute: Claude is the only AI model currently running in US military classified systems, and was used in military operations (including the operation to capture Venezuelan President Nicolás Maduro through Anthropic's partnership with Palantir). However, Anthropic has refused to lift its built-in safety guardrails that prevent the AI from being used for any purpose the company deems harmful.
  • The Trump administration ordered the federal government to stop using Anthropic and gave the Pentagon a 6-month phase-out period; Defence Secretary Pete Hegseth formally designated Anthropic a supply-chain risk — barring military contractors from using Claude for defence work.
  • This confrontation marks the first time a major AI safety company has been classified as a national security liability precisely because of its safety policies, creating a fundamental tension between AI ethics and military utility.

Static Topic Bridges

Artificial Intelligence in Defence and Dual-Use Technology Governance

AI systems have rapidly become dual-use technologies — with applications spanning civilian productivity, critical infrastructure, and military operations including autonomous weapons, intelligence analysis, and logistics. The governance of dual-use AI is a new frontier in national security policy globally. The US Department of Defense adopted its AI Ethics Principles in 2020, and the US National Security Commission on AI (NSCAI) recommended in 2021 that the US must "not fall behind" China in military AI to maintain strategic advantage.

  • Dual-use AI: AI that can serve both civilian and military purposes, raising export control, proliferation, and governance concerns
  • Palantir Technologies: US data analytics company that integrates commercial AI models (including Claude) into classified government and military platforms
  • Claude AI (by Anthropic) was the only AI running in US DoD classified networks — indicating the deep integration of commercial AI in defence
  • AI safety guardrails: Trained restrictions preventing AI from generating content facilitating mass casualties, chemical weapons guidance, or other prohibited acts — Anthropic's "Constitutional AI" approach
  • The episode raises questions about whether military-grade AI must be developed independently of commercial safety frameworks

Connection to this news: The Pentagon's designation of Anthropic as a supply-chain risk illustrates that AI safety constraints — designed to prevent misuse — can themselves become geopolitical liabilities when commercial AI is deeply embedded in national security systems. This is a foundational challenge for all democracies integrating commercial AI in defence.

India's AI Governance Framework and Military AI Aspirations

India's National Strategy for Artificial Intelligence (NSAI), published by NITI Aayog in 2018, identified AI as a tool for national development across healthcare, agriculture, smart cities, education, and transport. While India has no dedicated military AI law, the Defence Acquisition Procedure (DAP) 2020 includes provisions for emerging technology including AI and robotics in defence procurement. India's Defence AI Council (DAIC) and the Defence AI Project Agency (DAIPA) were established in 2018 to coordinate AI adoption across the armed forces.

  • IndiaAI Mission (2024): ₹10,371 crore over 5 years for AI compute infrastructure, startups, skilling, and safe-trustworthy AI
  • NITI Aayog's Responsible AI for All (2021): Principles of safety, accountability, fairness, transparency — parallels the Anthropic safety debate
  • India's IT Act, 2000 does not have explicit provisions for AI-based defence misuse — a legislative gap
  • India is developing indigenous AI models (like BharatGPT) partly to reduce dependence on US/China AI platforms — the Pentagon-Anthropic dispute makes this case more compelling
  • CERT-In (Computer Emergency Response Team India) under MeitY handles cybersecurity incidents including AI-enabled attacks

Connection to this news: The US-Anthropic confrontation is directly instructive for India's AI policy: deep integration of foreign commercial AI in India's defence or critical infrastructure creates sovereign risk — if a commercial provider's policies conflict with government requirements, the government may have no operational AI alternative. This underscores India's push for indigenous AI capabilities.

AI Safety, Ethics, and the "Constitutional AI" Approach

Anthropic designed Claude using a method called "Constitutional AI" — training the model against a set of principles (a "constitution") to make it helpful, harmless, and honest. This approach embeds safety constraints directly into the model's behaviour, making them difficult to override even for authorised users. This is philosophically different from rule-based content filters that can be selectively turned off.

  • Constitutional AI (Anthropic): Uses AI feedback to train models against a set of ethical principles; cannot be easily "unlocked" for military use
  • OpenAI, Google DeepMind, and Meta also have safety policies for their models, but have entered into commercial agreements with defence agencies
  • The AI Safety Institute (AISI) — established by the UK and US governments — focuses on evaluating frontier AI risks including misuse in conflict
  • UN Secretary-General's Advisory Body on AI recommended in 2024 that autonomous weapons systems capable of lethal decisions without human oversight should be internationally regulated
  • India signed the Bletchley Declaration (AI Safety Summit, November 2023) committing to safe and responsible AI development

Connection to this news: The Pentagon-Anthropic dispute is the first real-world test of what happens when Constitutional AI safety constraints conflict with state power. It will shape whether future AI companies build military-compliant models with "off-switch" safety controls or maintain unified, non-negotiable safety guardrails — a choice with global consequences.

Key Facts & Data

  • Claude (Anthropic) was the only AI running in US DoD classified networks at the time of the dispute
  • Defence contractors queried: Boeing Defense, Space and Security; Lockheed Martin
  • Boeing's response: No active contracts with Anthropic
  • Lockheed Martin: Confirmed it was contacted to assess exposure and reliance
  • Trump administration ordered federal government to stop using Anthropic; 6-month Pentagon phase-out
  • Anthropic founded: 2021, by former OpenAI researchers including Dario Amodei; headquarters: San Francisco
  • Constitutional AI: Anthropic's training methodology embedding safety into model behaviour
  • IndiaAI Mission (2024): ₹10,371 crore for AI infrastructure, safety, and skilling
  • India's Defence AI Council (DAIC) and DAIPA: Established 2018 to coordinate military AI adoption
  • Bletchley Declaration (Nov 2023): India and 28 other countries committed to frontier AI safety