The risk of AI bioweapon use is drawing increased scrutiny as researchers warn that advanced systems could lower certain technical barriers for malicious actors. Consequently, RAND is recommending nine overlapping biosecurity safeguards.
RAND Corporation’s August 18 report, “Building a Defense-in-Depth Biosecurity Strategy for the AI Era,” argues that no single control can prevent AI-assisted biological misuse. Its proposed measures cover sensitive information, biological materials, detection and deterrence.
Lead author Steph Guerra, who heads RAND’s AI x Bio work, said rapidly advancing AI could reduce technical and operational barriers. These barriers have historically limited access to biological-weapons capabilities. In addition, RAND said safeguards currently remain fragmented across governments, companies and countries.
The warning follows Tech Against Terrorism’s Counter-Terrorism AI Benchmark, published in July. Researchers tested 27 leading AI models using 2,339 graded outputs. They found that about one-third provided what the group classified as meaningful assistance beyond ordinary web search results in terrorist-related scenarios.
The benchmark also identified specific risks when safeguards for openly available models were removed. Moreover, Tech Against Terrorism Executive Director Adam Hadley said the results showed that terrorist misuse was both a safety and control problem for AI developers.
Read: Bill Gates Identifies Three Jobs That Will Survive Artificial Intelligence
However, the benchmark covered a broad range of terrorist-use cases and did not establish that one-third of responses specifically assisted biological-weapons development. Therefore, the broader terrorism finding is separate from concerns over advances in biological AI.
The International AI Safety Report 2026 separately found that general-purpose AI can provide increasingly sophisticated information relevant to biological research. However, it cautioned that substantial uncertainty remains about how much those capabilities increase the real-world risk of weapons. This is because significant practical barriers still exist.
