The September report identifies espionage, fraud and other abuses that Anthropic interrupted between December 2025 and August 2026.
In its September report published on the 10th, Anthropic describes malicious uses of Claude in seven areas, including cybersecurity, surveillance, influence and fraud. The company says it has interrupted the identified activities and strengthened its protections based on the observations collected.
The document highlights an increase in the execution and orchestration of tasks by AI, while human operators continue to choose targets. Anthropic specifies that these case studies concern remarkable threats, and not a representative sample of uses. The attributions and evaluations of the contribution of AI remain those of the company.
Read the full article on the original source
Go to the source →