OWASP Top 10 for LLM

OWASP has observed and documented vulnerabilities in LLMs 

LLMs like OpenAI ChatGPT, GPT-4, Google BARD, and Microsoft BING have practical uses in various fields because they can understand and generate text that resembles human language, enabling numerous possibilities for applications.

OWASP has identified and documented vulnerabilities in LLMs that are selected based on their potential impact, exploitability, and how common they are in the LLM landscape. Some notable weaknesses in the list include prompt injections, data leakage, inadequate sandboxing, and unauthorized code execution. Certain vulnerabilities, like prompt injections, can be carried out even without extensive coding experience.

As with other OWASP lists, the main purpose of this one is to be a learning resource for developers, designers, architects, managers, and organizations involved in deploying and managing LLM applications.