According to Decrypt, an automated system developed by OpenAI discovered 948 security flaws in Linux software and produced functional exploits for each flaw. The system also drafted corresponding fixes for the identified issues. The cost per flaw for this automated process was approximately $1,279. The findings raise concerns about the ability of AI to rapidly locate and exploit vulnerabilities in critical infrastructure software at relatively low cost.
The tool operated automatically, scanning for vulnerabilities and generating both exploit code and draft remediation suggestions. No human review or additional testing is mentioned in the source details. The reported figure of $1,279 per flaw reflects the estimated expense associated with the automated process for each flaw found, based on the source’s summary. The discovery involves Linux security issues, and the exploits produced by the system are described as working. The implications noted in the summary concern the potential for AI to find holes in critical software quickly and cheaply at scale.
