Weekly #252: When AI Ethics Collide with National Security
Highlight of the Week: In a significant confrontation that has reshaped the relationship between Silicon Valley and the US military, AI firm Anthropic found itself alienated from the Pentagon after refusing to compromise on its ethical safeguards.
How It Started
The dispute began over a contractual disagreement. The Pentagon demanded assurances that Anthropic’s AI model, Claude, could be utilized for ‘all lawful purposes.’ Anthropic countered this demand, asserting that such vague wording did not adequately limit high-risk applications, particularly concerning mass domestic surveillance and fully autonomous weapons systems. The company sought more explicit restrictions.
How It Escalated
The situation intensified when officials designated Anthropic as a supply chain risk. This designation suggested that Anthropic could lose federal contracts, leading to an order for agencies to cease using its systems and granting a limited six-month period to wind down existing agreements. Traditionally, supply-chain risk labels are reserved for foreign adversaries, not domestic companies negotiating contract terms. Experts argue that this classification has no clear precedent, and Anthropic plans to contest it in court.
Enter the Rival Company
Following Anthropic’s exclusion, OpenAI seized the opportunity to forge its own arrangement with the Pentagon. OpenAI emphasized its commitment to safety, including restrictions on mass surveillance and ensuring human oversight in the use of force. However, CEO Sam Altman later described the deal as rushed and indicated plans to amend it to prevent the AI system from being used for mass domestic surveillance.
The Reactions
The fallout from this episode reverberated throughout the tech industry. Major companies like Amazon, Nvidia, and Apple cautioned the government against broadly applying supply-chain risk designations to US tech firms, fearing negative impacts on innovation and public-private cooperation. Investors in Anthropic expressed concern about the potential harm to the company’s enterprise business and IPO prospects due to lost key contracts.
Interestingly, despite losing its Pentagon contract, Claude topped US app downloads, while uninstalls of the ChatGPT mobile app surged by 295%. This indicates that Claude may maintain its popularity among users and suggests that businesses might still collaborate with Anthropic on non-defense projects.
Why Does It Matter?
This dispute highlights a profound structural tension. Advanced AI systems are increasingly integral to military planning and decision-making, yet leading AI firms have articulated ethical boundaries regarding surveillance, lethal autonomy, and dual-use risks. The confrontation between Anthropic and the Pentagon raises critical questions about who sets these boundaries when national security and corporate governance intersect.
In Other News Last Week
This Week in AI Governance
The UN Secretary General, Antonio Guterres, emphasized the responsibility of AI experts to shape the technology’s use for the benefit of humanity. He noted that the new Independent International Scientific Panel on Artificial Intelligence does not start from scratch but builds on previous efforts by the UN High-Level Advisory Body on AI.
UNESCO has partnered with the National Centre for Artificial Intelligence (CENIA) to promote ethical AI development and enhance AI education in Chile and Latin America, focusing on digital skills and governance.
Child Safety Online
Indonesia has enacted regulations restricting minors under 16 from accessing high-risk digital platforms, set to begin gradual implementation on March 28. The UK is also considering restrictions on social media access for children, inviting public consultation on potential measures.
China’s draft measures regulating online content harmful to minors came into effect on March 1, while the European Parliament faces procedural delays on extending temporary rules for scanning child sexual abuse material.
Cyber Operations Intensify in the USA-Israel-Iran Conflict
The USA and Israel launched coordinated military strikes against Iran, accompanied by US cyber operations that disrupted Iranian communications. Researchers report that Iranian-aligned actors are preparing retaliatory cyberattacks, highlighting the integration of cyber operations into military conflicts.
Google Cuts Play Store Fees After Epic Games Settlement
Following a settlement with Epic Games, Google plans to reduce its Play Store commission on in-app purchases to 20% and introduce options for installing alternative app stores on Android devices.
Looking Ahead
The ICANN85 Community Forum is scheduled for March 7-12 in Mumbai, focusing on global Domain Name System coordination. Additionally, the Data Technology Seminar 2026 will address AI governance and its applications in public service media from March 10-12 in Geneva.
Reading Corner
Featured reading includes “Digital Sovereignty Stack: Infrastructure, Services, Data, and AI Knowledge” and the February 2026 issue of the Digital Watch newsletter, which discusses child safety online and the implications of AI governance.
