Geopolitics & Sovereign Positioning

The Department of Defense officially designated Anthropic a supply-chain risk after negotiations failed over how much control the military would have over its AI models, including use in autonomous weapons and mass domestic surveillance, according to Bloomberg and TechCrunch. The classification—previously reserved for adversary-nation companies like Huawei—puts Anthropic at risk of losing access to a wide range of US government business beyond the collapsed $200 million Pentagon contract. The DoD turned to OpenAI as an alternative provider, which accepted terms Anthropic declined, though OpenAI subsequently saw ChatGPT uninstalls surge 295 percent. The Financial Times reports that draft rules now mandate civilian government contracts make AI models available for any lawful use, formalising the access requirements that triggered the Anthropic dispute.

The designation creates a precedent for government leverage over domestic AI companies through procurement requirements that effectively compel model access for security applications. Heidy Khlaaf, chief AI scientist at AI Now Institute, told AI Now that existing safety guardrails for generative AI in high-stakes decisions are deeply lacking and easily compromised, making it highly doubtful systems inadequate for benign cases could be secured for complex military and surveillance operations. The episode reveals a fundamental tension: frontier AI companies positioned as safety-focused face pressure to provide government access that may conflict with stated principles, while competitors willing to accept those terms gain procurement advantage.

The US Commerce Department has drafted regulations that would restrict AI chip shipments to anywhere in the world without American approval, according to Bloomberg. The proposed rules would require permits for Nvidia and AMD AI chip sales globally, extending beyond current country-specific export controls that primarily target China and other adversary nations. This represents a shift from selective restrictions to universal permit requirements that would give the US government veto power over advanced computing sales to any destination, including allied countries.

The move suggests the US is attempting to maintain leverage over global AI infrastructure development by controlling access to advanced chips regardless of destination. Current export controls have already created pressure for alternative chip development, with South Korean startup Rebellions positioning itself to compete with Nvidia and AMD in AI semiconductors, according to Bloomberg. The effectiveness of expanded controls depends on whether they prevent adversaries from acquiring advanced computing or accelerate development of non-US alternatives that eventually erode American chip dominance.

Oracle and OpenAI terminated plans to expand their flagship AI data center in Abilene, Texas after negotiations dragged over financing and OpenAI's changing needs, according to Bloomberg and the Financial Times. The collapse created an opening for Meta to consider leasing the planned expansion site from developer Crusoe, with Nvidia facilitating Meta's discussions with the data center developer. Oracle is simultaneously planning thousands of job cuts as it manages a cash crunch driven by massive AI data center expansion spending, according to Bloomberg. The episode reveals how infrastructure bottlenecks are reshaping competitive positioning among frontier AI developers, with access to power and data center capacity becoming as strategically important as model capabilities.

The conflict in Iran is underscoring risks of building data centers in the Gulf region, with Bloomberg reporting that such facilities are inevitable targets in conflict, according to Sam Winter-Levy of the Carnegie Endowment for International Peace. This adds geopolitical risk considerations to data center location decisions that have primarily focused on power availability and cost. South Korea's HD Hyundai Electric is accelerating US expansion betting on surging demand for transformers and switchgear driven by AI power consumption, according to Bloomberg, illustrating how infrastructure bottlenecks create opportunities throughout the supply chain.

China announced it will harness artificial intelligence to create jobs as 12.7 million university graduates—exceeding Belgium's population—prepare to enter the labour market this year, according to Bloomberg. The framing positions AI as a solution to employment pressure rather than viewing automation as a displacement threat, contrasting sharply with Western discourse focused on AI replacing workers. The announcement suggests Beijing views AI deployment as compatible with maintaining employment levels, potentially through creation of AI-adjacent roles, platform work, or government-supported positions that incorporate AI tools rather than being displaced by them.

This represents a distinct strategic narrative from Western economies where AI is increasingly linked to job cuts, as Oracle's thousands of planned layoffs driven by AI spending illustrate, according to Bloomberg. However, Martha Gimble of Yale Budget Lab notes that data showing AI actually replacing human workers hasn't materialised yet, suggesting current layoffs may be using AI as justification rather than evidence of actual substitution. China's approach may reflect state capacity to direct AI deployment toward employment-compatible applications or acceptance of lower productivity AI uses if they maintain social stability through employment.

The Guardian's editorial board assessed that the intensified use of artificial intelligence in the Iran conflict demonstrates a paradigm shift in warfare has already begun, not remained theoretical, according to The Guardian. The editorial argues that speed of technological development and geopolitical turbulence are collapsing distinctions between theoretical arguments and real-world consequences, echoing UN Secretary-General António Guterres's warning that future development will only accelerate. Tech Policy Press convened experts on technology policy and security to identify key questions as the situation unfolds, though specific details about AI system deployment remain limited.

The significance lies not in confirmed technical details of AI use, which remain sparse in available reporting, but in the assessment by informed observers that operational AI deployment in conflict has crossed a threshold from experimental to routine. This suggests existing international frameworks for weapons control and rules of engagement are being outpaced by deployment reality, creating governance gaps that multilateral institutions have not closed despite years of discussions about autonomous weapons systems.