Unprotected data source for ai context rag

The application uses data from Cloud Firestore or Cloud Storage to provide run-time context to Firebase AI Logic (e.g., via Retrieval Augmented Generation), but the data source is not protected by strong Firebase Security Rules. This can lead to inference-time data poisoning (a variant of OWASP LLM03), where an attacker taints the data source, causing the AI to produce biased, inaccurate, or malicious outputs.