The potential for artificial intelligence to lower barriers in the development of biological weapons has become a subject of urgent inquiry. Peace researcher Dunja Sabra recently tested this hypothesis by interacting with prominent AI models, including ChatGPT, Claude, and DeepSeek.
Sabra's investigation aimed to ascertain if these generative tools could be misused to assist in the manufacturing or design of biological agents. The experiment sought to reveal whether current safety protocols within these systems are sufficient to prevent such dangerous applications.
The findings from this test highlight the ongoing debate regarding the dual-use nature of advanced AI technologies. As these models become more sophisticated, the risk that they could inadvertently provide instructions or insights for harmful purposes remains a critical concern for security experts and researchers alike.
Sources: Der Spiegel.