Content Safety and Moderation › Protecting Privacy in Responses

## Using Claude Sonnet 4.5 for privacy-aware responses
import os
from anthropic import Anthropic

client = Anthropic(api_key=os.getenv("ANTHROPIC_API_KEY"))

system_prompt = """You are a personal assistant with access to user information. 
Follow these privacy rules strictly:

NEVER share:
- Passwords or authentication credentials
- Full credit card or bank account numbers
- Social security numbers or government IDs
- Exact addresses or precise location data
- Medical information

You MAY share:
- Information the user just told you in this conversation
- General facts and public knowledge
- Summaries that don't reveal sensitive details

When asked about sensitive information:

1. Confirm you have the information but cannot share it directly
2. Offer to help in a privacy-safe way
3. Suggest the user access the information directly if needed

Example: If asked for a password, respond: "I don't share passwords for security reasons. 
I can help you reset it or guide you to where it's stored securely."
"""


def privacy_safe_response(user_message, user_data=None):
    # In a real system, user_data might come from memory or a database
    context = f"User data available: {user_data}" if user_data else ""

    response = client.messages.create(
        model="claude-sonnet-4-5",
        max_tokens=1024,
        system=f"{system_prompt}\n\n{context}",
        messages=[{"role": "user", "content": user_message}],
    )
    return response.content[0].text


## Test privacy handling
user_data = {
    "name": "Alice",
    "email": "alice@example.com",
    "password": "secret123",  # Should never be shared
    "credit_card": "4532-1234-5678-9010",  # Should never be shared
}

result = privacy_safe_response("What's my email and password?", user_data)
print(result)