Atlas › 07 Security Testing for AI Apps › Supply Chain Vulnerabilities, Overreliance, and Model Theft☰ Read as one page
Supply Chain Vulnerabilities, Overreliance, and Model Theft
3.1LLM05: Supply Chain VulnerabilitiesThe ML supply chain introduces risks absent from traditional software. Compromised model weights, poisoned fine-tuning data, malicious…3.2LLM06: Sensitive Information DisclosureThe LLM may reveal confidential information from its training data, system prompt, or conversation context.3.3LLM07: Insecure Plugin DesignLLM plugins (tool use, function calling) with excessive permissions allow the model to perform dangerous operations.3.4LLM08: Excessive AgencyThe LLM is given too much autonomy without adequate human oversight.3.5LLM09: OverrelianceUsers or systems trust LLM output without verification, leading to decisions based on hallucinated or incorrect information.3.6LLM10: Model TheftUnauthorized access to model weights, or extraction of model behavior through systematic querying.3.7OWASP LLM Top 10 Summary TableEvery entry in this table should have corresponding automated tests in your CI pipeline. The OWASP LLM Top 10 is the minimum security test…