Scientists are rapidly developing artificial intelligence (AI) models trained on biological data and combining them with general AI reasoning models and AI agents.[1] After the 50th annual Asilomar Conference, more than 100 researchers supported the call to prevent these AI systems from being used for deliberately malicious purposes, such as the development of biological weapons.[1] The paper's authors join this call, proposing that governments, life science funding organizations and other institutions implement tailored restrictions on access to a narrow set of pathogen data that pose risks of AI misuse.[1] Unfettered access to data has so far served science and society well and should be preserved in almost all fields.[1] Governments should periodically review these narrow restrictions with a view to improving them.[1] The article was published in Science, Volume 391, Issue 6785, Pages 558-561, February 5, 2026.[1]