OpenAI Warns of High Cybersecurity Risks from New AI Models
OpenAI has issued a warning regarding the potential cybersecurity risks posed by its upcoming artificial intelligence models. The company highlighted that these models could develop capabilities to execute zero-day remote exploits against well-defended systems or assist in complex enterprise or industrial intrusion operations. As the capabilities of these AI models advance, OpenAI is investing in strengthening them for defensive cybersecurity tasks. This includes creating tools to help defenders audit code and patch vulnerabilities more efficiently. To mitigate these risks, OpenAI is implementing a combination of access controls, infrastructure hardening, egress controls, and monitoring. Additionally, the company plans to introduce a program offering tiered access to enhanced capabilities for users and customers focused on cyber defense. OpenAI is also establishing the Frontier Risk Council, an advisory group that will initially focus on cybersecurity and later expand into other areas. 
