The US military has ended its contract with Anthropic, the artificial intelligence company behind the Claude chatbot, after the firm refused to remove safety guardrails designed to prevent mass surveillance and autonomous weapon use. The Pentagon has now turned to OpenAI to integrate AI systems in classified operations.

Media reports have revealed that Anthropic’s Claude AI was previously used to support operations targeting leaders in Venezuela and Iran. The chatbot reportedly assisted in a January mission that led to the capture of Venezuelan President Nicolás Maduro and was later deployed during preparations for a planned operation related to Iran’s late supreme leader, Ayatollah Ali Khamenei.

Experts say these cases provide a rare look at how advanced AI is being incorporated into US military planning and intelligence. Heidy Khlaaf, chief AI scientist at the AI Now Institute, described the rapid deployment of these systems as surprising, noting that large language models are prone to producing unreliable or incorrect outputs, which raises concerns in high-stakes environments.

The reported use of Claude aligns with the Trump administration’s push to make the US military “AI-first,” aiming to ensure the United States maintains an edge over global rivals, including China. Various forms of automation and AI have been used by the US military since the 2010s, with previous deployments focusing on logistics, maintenance, and translation services, according to Elke Schwarz, professor of political theory at Queen Mary University of London.

The Pentagon’s AI Acceleration strategy seeks to integrate AI across multiple domains, including cyber and intelligence operations. As part of this effort, a database called genai.mil allows officials to access AI tools, including Google’s Gemini and xAI’s Grok. The 2025 defense budget, dubbed the “Big Beautiful Bill,” allocates hundreds of millions of dollars to AI-related projects, including counter-drone systems, AI ecosystem development, and nuclear security missions.

While Anthropic’s $200 million partnership with the military was intended as a two-year prototype to advance national security and mitigate adversarial AI risks, the company’s refusal to remove guardrails meant the contract was canceled. Claude had been deployed across US government networks, including nuclear labs and intelligence analysis tasks.

The Department of War now faces the challenge of transitioning to OpenAI’s systems. Analysts say the intelligence gathered by Claude will likely remain in use and may be incorporated into new AI tools. Experts also warn that increasing reliance on AI in military operations could raise ethical concerns, particularly regarding the development of autonomous weapons that could select and engage targets without human oversight.

Giorgos Verdi, a policy fellow at the European Council on Foreign Relations, noted that while AI currently assists with tasks such as analyzing satellite imagery, the US military’s push toward fully autonomous systems could escalate conflicts if rival nations adopt similar technology.

The Pentagon is expected to continue experimenting with AI in operations while balancing effectiveness with ethical and legal constraints, marking a pivotal moment in the integration of artificial intelligence into modern warfare.

Chinese firms are dominating the global robotics market, with humanoid robots taking center stage at the Chinese New Year celebrations in Hangzhou earlier this year. Germany’s Chancellor Friedrich Merz witnessed a live display of robots dancing, performing backflips, and boxing during his visit in February. On his return, Merz remarked that Germany was “simply no longer productive enough,” highlighting concerns about Europe’s competitiveness in robotics.

Hangzhou-based Unitree has emerged as a leading innovator, with China accounting for 87 percent of all humanoid robots delivered in 2025. While Unitree shipped more than 4,000 units, it remains behind Agibot, which sold over 5,000 units, according to Forbes. Despite relatively modest sales—just over 13,000 robots worldwide last year—investors continue to pour capital into the sector. Barclays research in January 2026 estimated that the global humanoid robotics market, currently valued at $2–3 billion, could reach $200 billion by 2035.

European startups face significant challenges in competing with their Chinese and American counterparts. Rodion Shishkov, founder of London-based construction technology firm All3, said European companies have far less access to capital. “Here in Europe I have to fight—literally, fight—for tens of millions of euros of investment, while a similarly-positioned company in the United States can obtain billions,” he said. Shishkov noted that functional non-humanoid robots, like those his company develops for construction, often receive less attention and funding than flashy humanoid models, despite being more practical in many applications.

Andrei Danescu, CEO of autonomous robotics and AI logistics startup Dexory, warned that Merz’s trip to China risked framing robotics competition as a “beauty contest,” focused on humanoid appearance rather than solving real-world problems. Danescu pointed to collaborative arms on factory floors, autonomous logistics vehicles in warehouses, and surgical assistants as examples of robots already transforming industries in Europe.

China’s sustained investment spans hardware, software, manufacturing integration, and full supply chains, making it difficult for other regions to catch up. Danescu called on European regulators to accelerate policies, clarify liability frameworks, and provide public funding to support strategic growth. “The AI Act is a start, but robotics needs its own focused attention—policy, funding, strategy,” he said.

Safety remains a major hurdle for integrating robots into existing workflows. Sam Baker, a former industrial robotics engineer turned venture investor, said there is a lack of clear standards for deploying robots alongside humans in construction and manufacturing. Some companies, such as BMW, are experimenting with humanoid robots in production lines in Leipzig, Germany, to explore their potential without committing fully to large-scale deployment.

Baker said Europe cannot achieve full independence from Chinese hardware supply chains, but sees opportunities in software, intelligence, and experimentation. “It is an excellent time to build a robotics business in Europe. There’s a lot of white space to be filled on the intelligence and data side,” he said, highlighting the scope for innovation despite hardware constraints.