OpenAI releases restricted GPT-6 Cyber model

OpenAI has released GPT-6 Cyber, a restricted artificial intelligence model designed to identify software vulnerabilities and develop ways to exploit them, to a limited group of approved cybersecurity customers. The company said its testing found 81 previously unknown, high-severity vulnerabilities in Firefox and produced working exploits for 948 Linux kernel vulnerabilities, including 588 previously unknown flaws. Mozilla has begun releasing fixes for the Firefox vulnerabilities, according to OpenAI’s announcement. OpenAI said it reported its Linux findings to maintainers with proposed patches. The company also said it developed and validated a critical remote-code-execution vulnerability in the Brotli data-compression tool, and that the vulnerabilities it identified were reported to their owners and patched. The findings and performance claims are based on OpenAI’s own testing and have not been independently verified in the reporting. Finding vulnerabilities does not by itself establish how quickly affected systems can be fixed. OpenAI has not provided measurements showing how much its models accelerate patch deployment, while security professionals have raised concerns that...