California (Web Desk): OpenAI Chief Executive Officer Sam Altman has expressed concern over the growing ability of artificial intelligence (AI) to identify security vulnerabilities.
In a recent interview, Altman said it was becoming increasingly difficult to confine AI technology to secure, isolated computing environments known as sandboxes.
He said that as AI models improve their capabilities, detecting signs that they may escape their secure environments could also become more difficult. According to Altman, monitoring can provide additional time to respond, but genuine security requires AI to adopt human values.
Altman warned that AI-related risks are extremely serious at present and must be addressed at an early stage to prevent potential harm.
He further said AI models could identify hidden hardware security vulnerabilities with extraordinary proficiency. According to him, hackers could exploit these capabilities to develop software bugs that would be extremely difficult to fix.