Investigadores utilizan la IA de Anthropic para acceder al código interno de OpenAI, exponiendo dos vulnerabilidades

Independent cybersecurity researchers have successfully accessed OpenAI’s internal code systems by utilizing Claude, an artificial intelligence developed by Anthropic. This breach allowed the researchers to identify and expose two specific vulnerabilities within OpenAI’s infrastructure.

The incident highlights potential security risks associated with the integration and interaction of large language models with sensitive software environments. By leveraging the capabilities of one AI system to probe the internal architecture of another, the researchers have demonstrated a new vector for identifying system weaknesses.