AI agents show they can create exploits, not just find vulns

Recent research indicates that AI agents, such as those developed by Anthropic, OpenAI, and Google, are capable of not only identifying security vulnerabilities but also creating functional exploits. The study led by computer scientists from various institutions has resulted in the development of ExploitGym, a benchmark designed to evaluate the exploitative capabilities of AI models. This advancement raises important questions about the implications of AI in cybersecurity and the evolving threat landscape.