OpenAI has launched a bug bounty program focused on identifying potential safety issues in its technology. The program targets three main areas: agentic vulnerabilities, where an AI system might act independently of its intended purpose; prompt injection, where an attacker could manipulate the input to influence the AI's behavior; and data exfiltration, where sensitive information is extracted from the system without authorization. --- Why it matters: This matters because identifying these types of risks can help prevent unintended consequences of AI development and deployment. By incentivizing researchers and security experts to find vulnerabilities, OpenAI aims to improve the safety and reliability of its technology. Source: https://openai.com/index/safety-bug-bounty