Anthropic Disrupts State Sponsored AI Misuse and Industrial Espionage
10 sources across 10 countries
Who reported this
- The Daily Star
- UOL
- France 24
- Der Spiegel
- South China Morning Post
- The Indian Express
- The Japan Times
- Rappler
- Reuters
- Al-Monitor
What the colours mean
- Left
- Centre-left
- Centre
- Centre-right
- Right
- A hatched block means the outlet is affiliated with, or controlled by, a state.
Political lean describes where an outlet sits within the politics of its own country. It is never a position on a single global scale.
Political lean is comparable inside one country and not across them, which is why the bar groups by country first. Publicly funded broadcasters are not marked as state-linked.
The owner of each outlet is listed as a matter of record, not as a judgement about the outlet.
AI company Anthropic released a report on Thursday detailing the disruption of several malicious operations using its Claude models between January and July. The company reported blocking state sponsored surveillance campaigns originating in China, Iran, and West Africa that targeted dissidents and ethnic minorities, including pro democracy figures in Hong Kong and Tibetan and Falun Gong communities. In one instance, a contractor for Malian national security authorities used the AI to design intelligence gathering software; in another, Iranian actors used it to identify individuals via social media accounts.
Anthropic also accused several Chinese AI firms, including Alibaba, Moonshot, DeepSeek, and Xiaomi, of attempting to replicate Claude's capabilities. The company alleged that Alibaba conducted a large scale illicit distillation attack to improve its Qwen models. Furthermore, Anthropic claimed that Moonshot and DeepSeek secretly routed live customer requests to Claude, passing the responses off as their own. One example cited involved Moonshot relaying nearly 300,000 requests over ten days through thousands of fraudulent accounts.
Beyond espionage, the report highlighted attempts to use AI for developing conventional weapons, including missiles and drones, with cases linked to China, Russia, and Yemen. Anthropic also blocked queries related to biological research involving highly pathogenic bird flu and the chikungunya virus. While the company noted that some of this work involved scientists and could have been for legitimate medical study, it blocked the activity due to potential military links.
Additional threats included a suspected Russia linked cyber espionage campaign by the group Midnight Blizzard, which allegedly used AI to target Ukrainian government and military sectors. The company also identified the use of Claude by Russian state media to create fake independent news regarding an election in Moldova.
How each side framed it
- Centre-left
- These outlets emphasized the disruption of state sponsored surveillance and the risks to minorities and dissidents.
- Centre
- These outlets focused on the broad range of threats, including surveillance, biological weapons, and industrial theft.
- Centre-right
- This outlet focused primarily on the industrial competition and the specific accusations of model cloning and data routing.
Sources
- Centre Al-Monitor: China, Iran among countries that have used AI to aid spying, Anthropic says
- Centre-left Der Spiegel: Artificial Intelligence: Anthropic accuses Chinese AI rival of cheating
- Centre France 24: Anthropic thwarts surveillance operations conducted using its AI in China and Iran
- Centre-left Rappler: Anthropic disrupts Russian, Chinese AI campaigns targeting its Claude models
- Centre Reuters: Anthropic disrupts Russian, Chinese AI campaigns targeting its Claude models - Reuters
- Centre-right South China Morning Post: Moonshot, DeepSeek secretly routed user requests to Claude, Anthropic claims
- Centre The Daily Star: China, Iran among countries that have used AI to aid spying, Anthropic says
- Centre The Indian Express: Anthropic blocks weapons work queries on Claude AI from China, Russia users
- Centre The Indian Express: Russian espionage, dissident monitoring and AI cyberattacks: What Anthropic’s report says
- Centre The Japan Times: Anthropic disrupts Russian, Chinese AI campaigns targeting its Claude models
- Centre-left UOL: China and Iran are among countries that used AI for espionage, says Anthropic
100% of the statements in this article were traced back to the source articles listed above.