OpenAI introduces a new cyber model amid fears of AI cyberattacks
Summary
OpenAI is releasing a new cyber-focused version of its AI model called GPT-5.6-Cyber to trusted cybersecurity defenders. This version is designed to help companies prepare for automated cyberattacks by allowing advanced research on system weaknesses.Key Facts
- OpenAI is introducing GPT-5.6-Cyber as part of its Daybreak program for cybersecurity defenders.
- Daybreak has two levels: Daybreak Blue offers GPT-5.6 Sol without strict cyber safety limits, and Daybreak Red includes GPT-5.6-Cyber for deeper vulnerability research.
- GPT-5.6-Cyber answered 95% of advanced cybersecurity requests during testing, compared to 1.5% for GPT-5.6-Sol.
- The model reached a "High" level of cyber capability in OpenAI's safety evaluation but is less capable than another delayed model, Astra.
- Companies like Accenture, IBM, CrowdStrike, Cisco, and Palo Alto Networks can use these models to build better security tools and services.
- OpenAI is investigating an earlier hacking incident involving its tools breaking into the platform Hugging Face.
- AI companies like Anthropic and OpenAI are developing similar tools to help cybersecurity defenders find and fix vulnerabilities.
Read the Full Article
This is a fact-based summary from The Actual News. Click below to read the complete story directly from the original source.
