US security assessments warn AI could aid cyber and biological attacks

In short: U.S. national-security assessments are increasingly treating AI as a tool that could make both cyberattacks and some kinds of biological attacks easier to carry out.

What's going on

U.S. government warnings and outside research are converging on a similar point, AI can lower the “skill barrier” for harmful acts. In plain terms, it can act like a helpful assistant that speeds up work, including work done by criminals or hostile governments.

In cyberattacks, the Department of Homeland Security (DHS) has warned that AI could help attackers build new tools for faster and harder-to-detect hacks. DHS and other analysts say this could raise risks for critical infrastructure, which includes systems like pipelines and railways that people rely on every day. The Center for Strategic and International Studies (CSIS) has also argued that AI could enable larger-scale cyberattacks by nation-states, terrorist groups, or individuals.

In biological threats, DHS has said it is concerned about AI being misused to support chemical, biological, radiological, and nuclear threats, with special concern about biological weapons. A report from the Center for a New American Security (CNAS) suggests AI could help bad actors find information, acquire materials, and over time refine harmful pathogens (disease-causing germs). RAND research is more cautious, finding that today’s large language models (AI systems that generate text like a chatbot) did not measurably change biological attack risk in its assessment.

What to watch

The key shift is that multiple U.S. agencies and research groups are treating AI-enabled cyber and biological misuse as a national-security issue, not just a crime problem. Watch for tighter rules around dangerous biological and chemical research, and for new standards aimed at protecting essential services from AI-assisted hacking.

Source: NYTimes