10 best practices for securely developing with AI Developing with AI requires a strong understanding of potential risks and how to mitigate them. This includes being wary of direct and indirect prompt injection, restricting data access for LLMs, keeping humans in the loop where needed, identifying and fixing security vulnerabilities in generated code, and using good training data. It's also essential to use hybrid AI models where you can, keep track of your AI supply chain, and beware of hallucinations and misleading data. By following these best practices, developers can mitigate potential risks and fully leverage the benefits of AI while ensuring secure development practices.