OpenAI has released a new artificial intelligence model for cybersecurity at a time when AI hacking and digital attacks are on the rise. GPT-6 Cyber will help security researchers find software weaknesses, test systems and find potential vulnerabilities before cybercriminals can exploit them.
The launch shows how AI is becoming an integral part of the cybersecurity industry. Advanced AI tools can be used by organisations to protect their digital systems but the same technology could also be misused to develop more sophisticated cyberattacks.
OpenAI says GPT-6 Cyber is very good at finding security holes in common software. In testing it found 81 previously unknown high severity vulnerabilities in Mozilla Firefox (apparently).
The company also said the model helped researchers come up with working exploits for 948 Linux kernel vulnerabilities (of which 588 were previously unknown). Some of the weaknesses could allow attackers with local access to gain root privileges so they have more control over affected systems.
These results are important because Firefox and Linux are used on computers, servers and digital infrastructure everywhere. Security testing is done regularly but software updates can hide vulnerabilities for years even though they’re known. AI models that can look at a lot of code could find problems faster.
But technology also has risks. The ability to find vulnerabilities and work out working exploits could make it easier for malicious users to attack computer systems.
Businesses, government organisations and other institutions could have greater security problems if such capabilities fall into the wrong hands.
To address these concerns, OpenAI is being cautious about when it releases the model. GPT-6 Cyber will be made available to a small group of approved cybersecurity professionals and organisations via its Daybreak Red programme first.
The programme is intended to help with authorised security testing (AST), penetration testing and vulnerability research. OpenAI is also working on defensive applications through its Daybreak Blue program which does things like code reviews, malware analysis and investigating security incidents.
OpenAI has classified GPT-6 Cyber as having Critical cybersecurity capability in its safety framework — that’s because it can identify and possibly exploit software bugs more effectively than any other model.
The company is also making its earlier GPT-5.6 Cyber model available to approved Daybreak Red customers.
The announcement comes as technology companies explore how artificial intelligence can be used for digital security. From flagging suspicious activity to checking software for weaknesses, AI is changing how cybersecurity teams work.
The challenge for organisations will be to use these tools in ways that increase protection but don’t create new security risks. As AI capabilities improve, controlling access to powerful cybersecurity models is likely going to become more important.
OpenAI’s latest move reflects this new reality. GPT-6 Cyber could help security professionals find and fix vulnerabilities faster but its restricted availability also shows why powerful AI cybersecurity tools need oversight.
Comments