Study: DeepSeek explains biochemical interactions of mustard gas with DNA

DeepSeek's R1 generative AI system has been found to explain in detail the biochemical interactions of sulphur mustard (mustard gas) with DNA, raising major safety and security concerns about R1 and generative AI systems in general. What happened Researchers at cybersecurity firm Enkrypt AI found that Deepseek's R1 model provided detailed information on how sul ph ur mustard targets specific DNA bases, particularly guanine and adenine. The model is 3.5 times more likely to produce Chemical, Biological, Radiological and Nuclear (CBRN) content than other leading AI models such as OpenAI O1 and Claude-3 Opus, according to the researchers. Why it happened DeepSeek-R1 was trained using large-scale reinforcement learning without supervised fine-tuning, which may have led to the development of powerful reasoning behaviours without proper safeguards. The model's open-source nature and focus on efficiency in processing massive datasets likely contributed to its ability to generate such detailed and potentially dangerous information. What it means The finding raises concerns that DeepSeek R1, third-party systems running the model, and other open-source models could be used to develop chemical or biological weapons. The finding is likely to put more pressure on national government and internaitonal organisations to find common ground on AI safety and ethics. System 🤖 DeepSeek R1 Developer: DeepSeek Artificial Intelligence Co Country: Global Sector: Multiple Purpose: Generate text Technology: Generative AI Issue: Saf ety; Security Investigations, assessments, audits 👁️ Enkrypt AI. Read Teaming Report

Date it happened

2025-01-01

Organisation involved

DeepSeek Artificial Intelligence Co

Product, system or model

DeepSeek R1

This is a record of what was reported, not a finding that anyone broke the law. If it names your organisation and you believe it is wrong, the corrections process is free and open to everyone.