Secure AI with Thales & Google Cloud | Mitigating OWASP LLM02 Sensitive Data Disclosure

Описание: Unlock AI with confidence.
In this demo, we show how to extend AI security from infrastructure and runtime protection into data input/output protection — addressing one of the OWASP Top 10 for LLMs: LLM02 (Sensitive Information Disclosure).

🔒 Context:
In our previous demo, we showcased how to secure AI pipelines on Google Kubernetes Engine (GKE) with Confidential Computing and Thales CipherTrust — protecting where AI runs.

🛡️ What’s new in this demo:

PII Tokenization at Ingress — Google DLP with Thales CipherTrust CRDP transforms sensitive data before it reaches Gemini.

Safe AI Analysis — Gemini provides insights from tokenized data, preserving utility without exposing PII.

Authorized Detokenization — cleartext is revealed only under policy, for approved users, with full audit trails.

✔️ Why it matters:
This directly mitigates OWASP LLM02:2025 Sensitive Information Disclosure, covering risks from PII, financial data, health records, and confidential business information leaking into AI contexts.

🚀 Takeaway:
Together with our earlier Hybrid RAG on GKE demo, this showcases the full Secure AI journey:

Protect where AI runs (infra + runtime).

Protect what AI consumes & outputs (data ingress + egress).

🔗 See the evolving story:

Hybrid RAG on GKE Demo (Infra & Runtime Security): [Insert link]

This PII Tokenization Demo (Data Protection & Compliance): [Insert link]

Hashtags for search visibility:
#AIsecurity #OWASP #ConfidentialComputing #DataProtection #Compliance #Thales #GoogleCloud #DevSecOps