Governments are turning to Claude to automate spying
Add Axios as your preferred source to
see more of our stories on Google.
Illustration: Eniola Odetunde/Axios
State-run surveillance operations around the world are using Claude to streamline their operations, Anthropic said in a threat report released Thursday.
Why it matters: The barriers for governments interested in spying on dissidents, politicians, journalists and human rights activists are getting lower as AI models get more powerful.
Zoom in: Anthropic found people linked to governments in Mali, China and Iran using various Claude models to streamline their surveillance operations.
- In Mali, a single consultant working for national security authorities used Claude to create a system that collects data from the country's mobile operators and builds dossiers about people.
- Anthropic also disclosed another case where Iranian actors used Claude to build and deploy a malicious Firefox browser extension that "harvested users' identities from social networks."
- A religious affairs intelligence office in China has reduced its operations from "many teams of analysts" to just "a single office, using an AI assistant to produce thousands of investigations per month," according to the report.
Threat level: Governments also used Claude to analyze this information and select surveillance targets, per the report.
- For example, the Chinese government used Claude to analyze social media data, determine how politically sensitive the posts were and then pick possible targets for what they called "control," which could include coercive questioning or closer monitoring of someone's movements and communications.
- In June, Anthropic also banned an account that used Claude to build a commercial surveillance platform designed to spy on people in Iran and the Persian Gulf. The AI lab detected this use case before it became operational.
The big picture: Anthropic is seeing AI move from an experimental tool into the everyday bureaucracy of state surveillance.
- One Chinese state security bureau even created an internal manual for using AI in surveillance operations, while other officials used Claude to automate daily intelligence reports and query government surveillance databases.
Between the lines: AI tools are making state-backed surveillance cheaper and more efficient, rather than fundamentally changing who governments target, Jacob Klein, head of threat intelligence at Anthropic, told Axios.
- AI is increasingly allowing individual government employees or contractors to automate work that once required teams of analysts, Klein added.
- "They're effectively automating parts of the job within the intel apparatus," Klein said.
Reality check: Governments didn't need access to Anthropic's most powerful models to significantly expand their surveillance operations.
- All of the surveillance cases in the report involved Claude Haiku, Sonnet or Opus, rather than Anthropic's newer Fable or Mythos-class models.
Yes, but: Anthropic says it banned every account tied to the surveillance operations it uncovered and strengthened its safeguards based on what it found.
- However, cutting off Claude doesn't necessarily end the underlying operation. Klein said Anthropic has seen actors move to open-source models after its safeguards or enforcement created too much friction.
- In Mali, the government ultimately just deployed its surveillance platform locally using other models.
The bottom line: The surveillance use cases aren't hypothetical anymore, and Anthropic expects AI capabilities to keep improving.
- "Authoritarian states are using AI for surveillance, repression and influence operations today," Klein said. "It's no longer hypothetical."
Go deeper: AI's mass surveillance problem
