Austin Morrissey

A group of bionerds assembled at the London Initiative for Safe AI for a hackathon aimed at reducing biorisk. Our team produced this in under 48 hours. TL;DR Responsible contract research organizations, that perform DNA synthesis as a service, should screen customer requests prior to executing the work order. Likewise, responsible AI labs that develop and serve LLMs with superhuman scientific capability, should screen a user’s input whenever it involves nucleic acids to determine if assistance is appropriate. However, without making tool calls, a model’s ability to perceive and assess the true nature of any sequence is notoriously juvenile, completely devoid of nuance, accuracy, and reliability. We developed an agent-native tool — SecureMaxx — that makes sequences transparent to the model while countering obfuscation attempts by the user. In experimental conditions, we deployed high-risk sequences across several scientific scenarios of varying complexity, all of which bypassed the native Anthropic classifier on Sonnet 4.6. > When SecureMaxx is invoked, refusal rates rise from a 0–30% baseline to 70–100% depending on scenario complexity, with under 2 seconds of added latency per query and less than 1% safety tax on a standard capabilities benchmark (HellaSwag). Here, we aim to demonstrate that light-touch interventions can add meaningful friction to one step in the kill chain, and to outline the dual-use implications and additional hardening required for deployment against adversarial users. Research team & affiliations Gerard Boxo (Coefficient Giving), Allison Jia (MATS), Rebecca Zanini (Pivotal Research), Mac Walker , Austin Morrissey (Pivotal Research). Credit for the research belongs to the team. Errors in this writeup, and any opinions expressed, are mine alone and do not represent the views of my coauthors or their affiliations. 1. Introduction In the near future, biological capabilities including the design of pathogens more dangerous than anything