mlsanityData to 7 Oct 2026

Artificial intelligence and biological misuse: Differentiating risks of language models and biological design tools

arXiv 2306.13952 ↗First posted 24 Jun 2023cs.CYcited by 43

Not on the map. Why?

Abstract (arXiv)

As advancements in artificial intelligence (AI) propel progress in the life sciences, they may also enable the weaponisation and misuse of biological agents. This article differentiates two classes of AI tools that could pose such biosecurity risks: large language models (LLMs) and biological design tools (BDTs). LLMs, such as GPT-4 and its successors, might provide dual-use information and thus remove some barriers encountered by historical biological weapons efforts. As LLMs are turned into multi-modal lab assistants and autonomous science tools, this will increase their ability to support non-experts in performing laboratory work. Thus, LLMs may in particular lower barriers to biological misuse. In contrast, BDTs will expand the capabilities of sophisticated actors. Concretely, BDTs may enable the creation of pandemic pathogens substantially worse than anything seen to date and could enable forms of more predictable and targeted biological weapons. In combination, the convergence of LLMs and BDTs could raise the ceiling of harm from biological agents and could make them broadly accessible. A range of interventions would help to manage risks. Independent pre-release evaluations could help understand the capabilities of models and the effectiveness of safeguards. Options for differentiated access to such tools should be carefully weighed with the benefits of openly releasing systems. Lastly, essential for mitigating risks will be universal and enhanced screening of gene synthesis products.

Key facts. Cited by 43 papers in the map (data through 7 Oct 2026). First posted on arXiv 24 Jun 2023.

Builds on

No reference list yet.

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2023: 5; 2024: 16; 2025: 15; 2026: 7.

43 papers.

Sort:

By area:

  1. 2024Negative Preference Optimization: From Catastrophic Collapse to Effective Unlearningcited by 484
  2. 2023Large Language Model Alignment: A Surveycited by 241
  3. 2024Black-Box Access is Insufficient for Rigorous AI Auditscited by 123
  4. 2024Large Language Model Unlearning via Embedding-Corrupted Promptscited by 112
  5. 2024Visibility into AI Agentscited by 77
  6. 2024Language Models Resist Alignment: Evidence From Data Compressioncited by 50
  7. 2024Revisiting Who's Harry Potter: Towards Targeted Unlearning from a Causal Intervention Perspectivecited by 33
  8. 2023Open-Sourcing Highly Capable Foundation Models: An evaluation of risks, benefits, and alternative methods for pursuing open-source objectivescited by 31
  9. 2024Probing the limitations of multimodal language models for chemistry and materials researchcited by 31
  10. 2025Virology Capabilities Test (VCT): A Multimodal Virology Q&A Benchmarkcited by 27
  11. 2024Mapping the Ethics of Generative AI: A Comprehensive Scoping Reviewcited by 19
  12. 2023Controlling risks of AI in chemical science with agentscited by 17
31 more
  1. 2024SciSafeEval: A Comprehensive Benchmark for Safety Alignment of Large Language Models in Scientific Taskscited by 17
  2. 2024UnStar: Unlearning with Self-Taught Anti-Sample Reasoning for LLMscited by 14
  3. 2024Training Compute Thresholds: Features and Functions in AI Regulationcited by 13
  4. 2023Towards Publicly Accountable Frontier LLMs: Building an External Scrutiny Ecosystem under the ASPIRE Frameworkcited by 11
  5. 2025Distillation Robustifies Unlearningcited by 11
  6. 2024Prioritizing High-Consequence Biological Capabilities in Evaluations of Artificial Intelligence Modelscited by 10
  7. 2024The Necessity of AI Audit Standards Boardscited by 9
  8. 2025Your LLM Agents are Temporally Blind: The Misalignment Between Tool Use Decisions and Human Time Perceptioncited by 7
  9. 2025DRAGON: Guard LLM Unlearning in Context via Negative Detection and Reasoningcited by 7
  10. 2025Improving LLM Unlearning Robustness via Random Perturbationscited by 6
  11. 2025Jailbreak-Tuning: Models Efficiently Learn Jailbreak Susceptibilitycited by 6
  12. 2023Towards Responsible Governance of Biological Design Toolscited by 5
  13. 2025LLM Unlearning Without an Expert Curated Datasetcited by 5
  14. 2026Measuring Mid-2025 LLM-Assistance on Novice Performance in Biologycited by 5
  15. 2025Mapping AI Benchmark Data to Quantitative Risk Estimates Through Expert Elicitationcited by 4
  16. 2025Efficient Safety Alignment of Large Language Models via Preference Re-ranking and Representation-based Reward Modelingcited by 4
  17. 2025Distinguishing Task-Specific and General-Purpose AI in Regulationcited by 3
  18. 2024CRiskEval: A Chinese Multi-Level Risk Evaluation Benchmark Dataset for Large Language Modelscited by 2
  19. 2025Ready2Unlearn: A Learning-Time Approach for Preparing Models with Future Unlearning Readinesscited by 2
  20. 2025How Catastrophic is Your LLM? Certifying Risk in Conversationcited by 2
  21. 2026A Blind Spot in Alignment: Quantifying Biosecurity Risks in Large Language Modelscited by 2
  22. 2024The whack-a-mole governance challenge for AI-enabled synthetic biology: literature review and emerging frameworkscited by 1
  23. 2025Foundation models may exhibit staged progression in novel CBRN threat disclosurecited by 1
  24. 2025Can Large Language Models Design Biological Weapons? Evaluating Moremi Biocited by 1
  25. 2026Break Me If You Can: Self-Jailbreaking of Aligned LLMs via Lexical Insertion Promptingcited by 1
  26. 2024Targeted Angular Reversal of Weights (TARS) for Knowledge Removal in Large Language Modelscited by 0
  27. 2025Dimensional Characterization and Pathway Modeling for Catastrophic AI Riskscited by 0
  28. 2026Measuring Safety Alignment Effects in Autonomous Security Agentscited by 0
  29. 2026Harmonizing AI Safety Thresholdscited by 0
  30. 2026FUSE: An Evaluating Framework for Dangerous Capabilities of LLMscited by 0
  31. 2026Agentic BAIM-LLM Evaluation (ABLE): Benchmarking LLM Use of Protein Design Toolscited by 0