Anthropic unveils five cases of claude's misuse in warfare, surveillance, and repression

Anthropic, the AI safety-focused lab behind Claude, has disclosed five instances where its AI model was exploited for malicious purposes. The incidents, uncovered over eight months of monitoring and disruption efforts, include an Iran-linked operation targeting U.S. naval forces, a Yemeni team using Claude to build missiles, a China-linked effort to track Uyghurs, and an individual leveraging the model to create surveillance systems for 25 million phones. In one case, a dangerous virus research request was redirected to another AI after Claude refused the task. These revelations have intensified scrutiny on AI safety, with lawmakers in Washington now treating the issue as a pressing political concern. Anthropic’s findings highlight the growing role of frontier AI labs in intelligence operations, raising concerns about the potential misuse of advanced technologies.