OpenAI develops early warning system for GPT-4 biothreat risk

Thumbnail Image

The information displayed in the AIM should not be reported as representing the official views of the OECD or of its member countries.

OpenAI assessed GPT-4’s risk in aiding biological threat creation, running a study with experts and 100 participants. Results showed only a slight increase in capability compared to online resources. To monitor future misuse, OpenAI is building an early warning “tripwire” system and urges ongoing research on LLM biosecurity hazards.[AI generated]

Why's our monitor labelling this an incident or hazard?

The event involves an AI system (GPT-4) and its potential use in a harmful context (bioweapons development). However, the article states that the AI's usefulness in this regard is limited and does not describe any realized harm or incident. Therefore, it represents a plausible future risk rather than an actual incident. This fits the definition of an AI Hazard, as the development or use of the AI system could plausibly lead to harm, but no harm has yet occurred.[AI generated]
AI principles
SafetyRobustness & digital securityAccountability
Industries
Healthcare, drugs, and biotechnologyGovernment, security, and defence
Business function:
ICT management and information security
AI system task:
Content generationEvent/anomaly detection

Articles about this incident or hazard