diff --git a/docs/my-website/docs/eu-ai-act-compliance.md b/docs/my-website/docs/eu-ai-act-compliance.md
index 93ecada079f..c8f4d339433 100644
--- a/docs/my-website/docs/eu-ai-act-compliance.md
+++ b/docs/my-website/docs/eu-ai-act-compliance.md
@@ -151,7 +151,7 @@ class ComplianceLogger(CustomLogger):
 litellm.callbacks = [ComplianceLogger()]
 ```
 
-Whichever option you choose, connect to a persistent backend with a retention policy of at least 6 months (Article 26(6) for deployers; Article 18 requires 10 years for providers).
+Whichever option you choose, connect to a persistent backend with a retention policy of at least 6 months (Article 26(5) for deployers; Article 18 requires 10 years for providers).
 
 ## Article 13: Transparency
 
@@ -184,6 +184,30 @@ What you need to build for Article 14 compliance:
 
 LiteLLM provides the **logging and hook infrastructure** to build human oversight on top of. The oversight logic itself — review queues, approval workflows, kill switches — lives in your application layer.
 
+## Article 50: Transparency for AI-interacting systems
+
+Article 50 applies to **all AI systems that interact directly with users**, not just high-risk systems. If your LiteLLM deployment powers a chatbot, virtual assistant, or any interface where a user communicates with AI-generated responses, you must:
+
+1. **Inform users they are interacting with an AI system** — before or at the start of the interaction
+2. **Mark AI-generated content** — if the system generates text, audio, images, or video that could be mistaken for human-created content, it must be machine-readable as AI-generated (Article 50(2))
+3. **Disclose deepfakes** — if the system generates or manipulates content depicting real people or events, this must be disclosed (Article 50(4))
+
+LiteLLM itself does not handle user-facing disclosure — it operates at the API gateway layer. Your application must implement the disclosure:
+
+```python
+# Example: Add AI disclosure header to responses
+response = litellm.completion(model="gpt-4", messages=messages)
+
+# Your application layer adds the disclosure
+user_response = {
+    "content": response.choices[0].message.content,
+    "ai_disclosure": "This response was generated by an AI system.",
+    "model_used": response.model,  # Transparency: which model answered
+}
+```
+
+Article 50 obligations exist **independently of Annex III classification**. Even if your system is not high-risk, this section applies.
+
 ## GDPR considerations
 
 LiteLLM processes user prompts. If those prompts contain personal data: