Anthropic builds surveillance system
Archived — this story has rotated out of today’s deck. It is kept here in full.
The gist
Anthropic says it blocked state actors using Claude to build mass surveillance systems.
The report shows AI is lowering the cost of government spying on dissidents.
Background
Anthropic released a threat intelligence report on September 10, 2026, detailing how its Claude AI models were misused by state-linked actors in China, Iran, Mali, and West Africa to build surveillance platforms, profile dissidents, and automate spying. The report also covered attempts to use Claude for biological weapons research and cyberattacks. The findings come as governments increasingly adopt AI tools to streamline surveillance operations, lowering barriers for spying on journalists, activists, and politicians.
How it unfolded
- Dec 2025 - Aug 2026Anthropic researchers identified misuse of Claude by spyware vendors, state-sponsored groups, and politically motivated individuals.
- Sep 10, 2026Anthropic publishes its September 2026 threat intelligence report detailing state surveillance misuse.
- Sep 11, 2026Reuters reports on Anthropic's findings, including a Mali surveillance platform and Iranian profiling operations.
Who’s saying what
- Official
- Anthropic says it blocked these efforts and published the report to counter misuse of its AI models.
- Analysts
- Jacob Klein, head of threat intelligence at Anthropic, said AI tools are making state-backed surveillance cheaper and more efficient, rather than fundamentally changing who governments target.
Still unverified
The Prospect.org article claims Anthropic is building a predictive surveillance system to monitor activists who oppose AI development, based on job postings and interviews. This claim is not confirmed by other sources and is disputed.