OWASP Top 10 for LLM Applications
OWASP's parallel security list specifically for applications built on large language models.
In detail
The OWASP Top 10 for LLM Applications, first published in 2023 and updated annually, lists the most critical risks specific to AI-powered apps. Prompt injection sits at #1, followed by sensitive information disclosure, supply chain vulnerabilities (compromised model weights, malicious tools), data and model poisoning, improper output handling, excessive agency, system prompt leakage, vector and embedding weaknesses, misinformation and unbounded consumption.
Why it matters for Australian business
For Australian businesses adopting AI, this list is the equivalent of the regular OWASP Top 10 but for the new attack surface AI introduces. Few internal teams know it. Few AI builders address it. We use it as the reference standard for AI-specific security review on agentic apps and assistants.