Can Large Language Models Design Biological Weapons? Evaluating Moremi Bio

Advances in AI, particularly LLMs, have dramatically shortened drug discovery cycles by up to 40% and improved molecular target identification. However, these innovations also raise dual-use concerns by enabling the design of toxic compounds. Prompting Moremi Bio Agent without the safety guardrails to specifically design novel toxic substances, our study generated 1020 novel toxic proteins and 5,000 toxic small molecules. In-depth computational toxicity assessments revealed that all the proteins scored high in toxicity, with several closely matching known toxins such as ricin, diphtheria toxin, and disintegrin-based snake venom proteins. Some of these novel agents showed similarities with other several known toxic agents including disintegrin eristostatin, metalloproteinase, disintegrin triflavin, snake venom metalloproteinase, corynebacterium ulcerans toxin. Through quantitative risk assessments and scenario analyses, we identify dual-use capabilities in current LLM-enabled biodesign pipelines and propose multi-layered mitigation strategies. The findings from this toxicity assessment challenge claims that large language models (LLMs) are incapable of designing bioweapons. This reinforces concerns about the potential misuse of LLMs in biodesign, posing a significant threat to research and development (R&D). The accessibility of such technology to individuals with limited technical expertise raises serious biosecurity risks. Our findings underscore the critical need for robust governance and technical safeguards to balance rapid biotechnological innovation with biosecurity imperatives.

Localized Data Work as a Precondition for Data-Centric ML: A Case Study of Full Lifecycle Crop Disease Identification in Ghana

The Ghana Cashew Disease Identification with Artificial Intelligence (CADI AI) project demonstrates the importance of sound data work as a precondition for the delivery of useful, localized datacentric solutions for public good tasks such as agricultural productivity and food security. Drone collected data and machine learning are utilized to determine crop stressors. Data, model and the final app are developed jointly and made available to local farmers via a desktop application.

minoHealth.ai: A Clinical Evaluation Of Deep Learning Systems For the Diagnosis of Pleural Effusion and Cardiomegaly In Ghana, Vietnam and the United States of America

A rapid and accurate diagnosis of cardiomegaly and pleural effusion is of the utmost importance to reduce mortality and medical costs. Artificial Intelligence has shown promise in diagnosing medical conditions. With this study, we seek to evaluate how well Artificial Intelligence (AI) systems, developed my minoHealth AI Labs, will perform at diagnosing cardiomegaly and pleural effusion, using chest x-rays from Ghana, Vietnam and the USA, and how well AI systems will perform when compared with radiologists working in Ghana. The evaluation dataset used in this study contained 100 images randomly selected from three datasets. The Deep Learning models were further tested on a larger Ghanaian dataset containing five hundred and sixty one (561) samples. Two AI systems were then evaluated on the evaluation dataset, whilst we also gave the same chest x-ray images within the evaluation dataset to 4 radiologists, with 5 - 20 years experience, to diagnose independently. For cardiomegaly, minoHealth-ai systems scored Area under the Receiver operating characteristic Curve (AUC-ROC) of 0.9 and 0.97 while the AUC-ROC of individual radiologists ranged from 0.77 to 0.87. For pleural effusion, the minoHealth-ai systems scored 0.97 and 0.91 whereas individual radiologists scored between 0.75 and 0.86. On both conditions, the best performing AI model outperforms the best performing radiologist by about 10%. We also evaluate the specificity, sensitivity, negative predictive value (NPV), and positive predictive value (PPV) between the minoHealth-ai systems and radiologists.