AI Model Used In Bioweapon Development Attempt, Anthropic Warns

Anthropic, a leading AI safety company, has revealed in a new report that malicious actors sought to exploit its Claude language model for developing bioweapons, raising fresh alarms about potential dual-use risks in advanced artificial intelligence systems.

The most severe incident documented by the firm occurred in May, when a scientist attempted to leverage Claude to draft a grant application for gain-of-function research on the chikungunya virus—a mosquito-borne pathogen responsible for periodic outbreaks across South and Southeast Asia. The researcher aimed to enhance the virus's transmissibility and virulence through the AI system's assistance.

The disclosure underscores mounting concerns about how powerful AI tools could be repurposed for harmful applications, even as these technologies gain widespread adoption across research institutions and industries. Anthropic stated it has implemented safeguards to prevent such misuse, though the incident demonstrates the vulnerability of open-access AI systems to malicious intent.

Bioweapon development remains a critical global security concern, with international treaties like the Biological Weapons Convention prohibiting such research. The incident highlights the need for robust oversight mechanisms and responsible deployment practices as artificial intelligence becomes increasingly integrated into scientific research workflows across the region and globally.