Anthropic Reveals Foreign Rivals Used Claude AI for Weapons and Surveillance Research

Anthropic has exposed and shut down multiple dangerous operations where foreign adversaries and state sponsored hackers tried to use its Claude computer assistant for military weapons research, biological experiments, and mass surveillance.

The San Francisco based technology company discovered that bad actors from nations including China, Russia, Iran, and Yemen attempted to bypass safety filters to build dangerous computer software, design bioweapons, and track political dissidents.

Instead of using the smart tool for simple everyday tasks like writing emails or answering questions, hostile foreign operators tried to use the software to speed up dangerous scientific research and launch automated cyberattacks against foreign governments and critical infrastructure.

The detailed threat intelligence report was publicly published by Anthropic and reported by financial news outlet Bloomberg on Thursday, September 10, 2026.

Covering suspicious network activities detected and blocked between December 2025 and August 2026, the company detailed seven specific areas of misuse.

These included biological weapons research, conventional military weapons development, state backed spying campaigns, mass surveillance programs, online scams, and foreign propaganda operations.

Investigators noted that adversaries used widely available versions of the software, such as Claude Haiku, Sonnet, and Opus, by routing requests through hidden network connections and fake account resellers to hide their true identities.

Anthropic is publishing these detailed threat disclosures now to warn global lawmakers, public defense agencies, and the broader tech industry that hostile foreign governments are actively trying to turn commercial software into military weapons.

As artificial intelligence models become smarter at reading complex scientific files and writing computer code, rogue actors use them to save time and skip years of difficult technical work.

To stop these dangerous abuses, Anthropic built advanced threat detection monitors, blocked suspicious user accounts, downgraded access for risky requests, and shared critical security intelligence with law enforcement and rival tech companies to keep global networks safe.

Explaining how hostile actors frame dangerous bioweapons requests in subtle, complex ways rather than asking for harmful tools directly, Anthropic Head of Threat Intelligence Jacob Klein stated that “you are not seeing someone in a comic book kind of way say, ‘Hey, I want to build a biological weapon to kill everybody.’ It’s an incredibly nuanced situation”.

Detailing why exposing these blocked attacks helps protect the broader technological ecosystem from fast-evolving cyber threats, Anthropic’s Threat Intelligence team wrote in its official report that “we believe we have a responsibility to disclose malicious misuse of our services,” adding that public disclosures help harden safeguards across the entire industry.

Warning that smart software allows bad actors to launch faster, broader, and cheaper cyberattacks against public targets, Anthropic security experts highlighted that “adversaries can operate faster, across a broader and deeper surface area, with fewer resources” when abusing digital tools.

By detecting and shutting down foreign attempts to misuse Claude AI, Anthropic is taking strong steps to protect public safety and national security.

Enforcing strict safety checks ensures that smart technology serves helpful, peaceful purposes rather than aiding dangerous weapons research or state surveillance.