OpenAI ChatGPT Models Reportedly Jailbroken to Provide Chemical, Biological, and Nuclear Weapons Instructions
Description
An NBC News investigation found that OpenAI's language models o4-mini, GPT-5-mini, oss-20b, and oss-120b could be jailbroken under normal usage conditions to bypass safety guardrails and generate detailed instructions for creating chemical, biological, and nuclear weapons. Using a publicly documented jailbreak prompt, reporters repeatedly elicited hazardous outputs such as steps to synthesize pathogens or maximize harm with chemical agents. The findings reportedly revealed significant real-world safeguard failures, prompting OpenAI to commit to further mitigation measures.
Actors
Alleged Deployers
- openai
Alleged Developers
- openai
Harmed Parties
- public-safety
- national-security-and-intelligence-stakeholders
- general-public
Classification
Risk Domain
Malicious Actors & MisuseRisk Subdomain
4.2. Cyberattacks, weapon development or use, and mass harm
MIT Risk Profile
Source
Related Incidents (Malicious Actors & Misuse)
View all →Female Students at Westfield High School in New Jersey Reportedly Targeted with Deepfake Nudes
Oct 2023
ChatGPT Was Alleged to Have Aided Planning of Florida State University Mass Shooting
Apr 2025
Ongoing Purported AI-Assisted Identity Fraud Enables Unauthorized Access to Western Companies by North Korean IT Workers
Research Papers on This Topic
How Johnny Can Persuade LLMs to Jailbreak Them: Rethinking Persuasion to Challenge AI Safety by Humanizing LLMs
AgentHarm: A Benchmark for Measuring Harmfulness of LLM Agents
Regulating ChatGPT and other Large Generative AI Models
Great, Now Write an Article About That: The Crescendo Multi-Turn LLM Jailbreak Attack