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

Trump tells U.S. government to ‘immediately’ stop using Anthropic AI tech

February 28, 2026 5 min read Science & Technology International Relations GS3 GS2
On This Page ▼

What Happened

  • US President Donald Trump directed all federal government agencies to "immediately cease" using artificial intelligence tools developed by Anthropic (maker of the Claude AI assistant), effective February 27, 2026.
  • The directive followed a dispute between Anthropic and the Pentagon (US Department of Defense): Anthropic had sought assurances that its AI would not be used for mass domestic surveillance of Americans or in fully autonomous (lethal) weapons systems.
  • The Pentagon rejected Anthropic's conditions, insisting on "access without limitations." When Anthropic declined to remove its usage restrictions, Defense Secretary Pete Hegseth declared Anthropic a "Supply-Chain Risk to National Security."
  • Trump's order provides a six-month phase-out for agencies — such as the Department of Defense — that were actively using Anthropic products.
  • Anthropic responded that "no amount of intimidation or punishment from the Department of War will change our position on mass domestic surveillance or fully autonomous weapons" and announced it would legally challenge the supply chain risk designation.
  • Simultaneously, OpenAI (Anthropic's main competitor) announced a new contract with the Pentagon — filling the government AI vacuum left by Anthropic's exclusion.

Static Topic Bridges

Artificial Intelligence Governance: Safety vs. Capability Trade-offs

The central tension in AI governance is between maximising AI capability (for commercial, military, and national security advantages) and implementing safety guardrails (to prevent misuse, discrimination, and catastrophic risks). This tension has produced two broad camps: "AI accelerationists" who argue restrictions hamper competitiveness, and "AI safety advocates" who argue unfettered AI deployment risks irreversible harms.

  • Anthropic, founded by former OpenAI researchers including Dario Amodei and Daniela Amodei, explicitly positions itself as an "AI safety company," publishing safety research and designing its AI systems (the "Constitutional AI" method) with ethical constraints built in.
  • The debate over AI in autonomous weapons specifically concerns "lethal autonomous weapons systems" (LAWS) — weapons that can select and engage targets without human authorisation. The UN has been discussing a potential international treaty on LAWS since 2014 without reaching consensus.
  • The US withdrew from the JCPOA on AI safety norms, signalling a shift in the Trump administration away from multilateral AI governance frameworks toward competitive national deployment.
  • India's National AI Strategy (NITI Aayog, 2018) emphasises responsible AI and has established a framework for AI governance — but India's regulatory approach remains voluntary rather than legally binding.

Connection to this news: The Anthropic-Pentagon dispute crystallises the core AI governance dilemma: when a government seeks to deploy powerful AI for sensitive applications without safety constraints, what obligation does the AI developer have to refuse? The standoff has direct implications for how AI developers globally — including Indian companies — structure their government contracts.

US Federal AI Procurement Policy

The US federal government is one of the world's largest single buyers of technology, spending over $100 billion annually. The Trump administration's executive order framework for AI (December 2025) established that federal AI procurement must prioritise "national and economic security" while requiring compliance with "Unbiased AI Principles" — a set of content neutrality requirements emphasising that AI systems must not alter outputs based on political viewpoints.

  • By March 2026, US federal agencies were required to update procurement contracts for large language models (LLMs) to include compliance requirements, Acceptable Use Policies, and transparency documentation (model cards, training data summaries).
  • The General Services Administration (GSA) issued draft AI contract terms in early 2026 setting standards for government AI deployments.
  • OpenAI's rapid announcement of a Pentagon contract after Anthropic's exclusion illustrates the competitive dynamics: AI companies face pressure to be "government-friendly" to access major public sector contracts.
  • The US Congressional Research Service has flagged "supply chain risks" from foreign AI systems (particularly Chinese-developed models) as a national security concern — a framework now being applied to domestic AI companies as well.

Connection to this news: The Trump administration's use of the "supply chain risk" designation against an American AI company is unprecedented — it demonstrates that the federal AI procurement framework can be weaponised against domestic companies that resist government demands, creating a chilling effect on AI safety research.

India's AI Policy and Regulatory Framework

India is the world's third-largest AI ecosystem by number of companies and talent pool. The government has pursued AI development through the National AI Mission (2023), IndiaAI initiative, and the Digital India programme. India has thus far avoided imposing hard regulatory constraints on AI, positioning itself as a hub for AI development and deployment.

  • India's IndiaAI Mission (2024) includes a ₹10,372 crore investment in AI compute infrastructure (10,000 GPU cluster), application development for priority sectors, and an AI safety institute.
  • India has not enacted binding AI legislation; the Digital Personal Data Protection (DPDP) Act, 2023 provides data governance norms but does not specifically address AI-generated risks.
  • The Ministry of Electronics and Information Technology (MeitY) issued advisory guidelines in 2023 requiring AI platforms to obtain government permission before deploying models that could cause harm — a requirement subsequently diluted.
  • India's approach contrasts with the EU's AI Act (2024) — the world's first binding AI law — which categorises AI systems by risk level and imposes strict requirements on "high-risk" applications including military, policing, and biometric surveillance.

Connection to this news: India's governance gap — the absence of a statutory AI framework — means there is no clear answer to the question Anthropic raised: can an AI company refuse a government's demand for unlimited military use? The Anthropic-Pentagon dispute will inform how India designs its forthcoming AI governance architecture.

Key Facts & Data

  • Anthropic founded: 2021, by Dario Amodei, Daniela Amodei, and other former OpenAI researchers; headquarters in San Francisco.
  • Claude AI: Anthropic's AI assistant, used by US government agencies including the Department of Defense before the ban.
  • Six-month phase-out period ordered for agencies currently using Anthropic tools.
  • OpenAI (competitor) announced a new Pentagon contract immediately following Anthropic's exclusion.
  • India's IndiaAI Mission (2024): ₹10,372 crore investment; 10,000 GPU compute cluster; AI safety institute.
  • EU AI Act (2024): first binding AI law globally; prohibits "unacceptable risk" AI applications and imposes conformity assessments on "high-risk" systems.
  • UN discussions on Lethal Autonomous Weapons Systems (LAWS): ongoing since 2014 under the Convention on Certain Conventional Weapons (CCW); no binding treaty yet agreed.
  • Anthropic's "Constitutional AI" method: a training approach where the AI system is given a list of ethical principles and trained to self-evaluate its outputs against them.