Anthropic has revealed that it recently intercepted multiple attempts to use its artificial intelligence models to facilitate the creation of biological and conventional weapons. In a detailed threat intelligence report released Thursday, the company disclosed that it disrupted various malicious activities involving its Claude AI suite between December 2025 and August 2026. While some of the misuse was relatively common, such as hotel wifi scams and fake dating apps, others were far more sinister, including efforts by state sponsored groups to build surveillance tools aimed at identifying political dissidents.
Among the most alarming discoveries were five specific case studies where actors attempted to leverage the AI to support biological weapons development. Jacob Klein, Anthropic’s head of threat intelligence, explained that these situations are rarely as obvious as a villain in a comic book explicitly asking how to create a plague. Instead, it is an incredibly nuanced challenge because the same technical data required to engineer a deadly pathogen could also be used by legitimate researchers to develop life saving vaccines or cures for diseases. Beyond biology, the company also blocked six instances where users tried to design software for missiles, armed drones and explosives.
The report further highlights a growing trend of sophisticated cyber warfare powered by AI. Anthropic identified usage linked to Russian espionage campaigns and Iranian propaganda institutions, noting that one hacking group associated with Russia’s Midnight Blizzard reportedly used the AI to rewrite malware code automatically whenever it was detected by security defenses. Additionally, the firm accused several Chinese AI companies of attempting to replicate Claude’s internal capabilities through a process known as distillation.
These revelations come amid a mounting global debate over whether AI development is moving too fast for safety protocols to keep up. OpenAI chief scientist Jakub Pachocki recently called for voluntary industry slowdowns, while U.S. Senator Bernie Sanders has introduced legislation seeking to ban certain types of superintelligence entirely due to potential cataclysmic risks. As regulators and lawmakers weigh the need for international treaties on AI safety, Anthropic maintains that disclosing these threats is part of its corporate responsibility and has begun sharing its findings with government authorities and industry partners.