What Never to Send to Claude

I když je Claude bezpečný, některé informace by nikdy neměly opustit váš perimetr.

Absolutní zákaz ❌

1. Secrets & Credentials

# ❌ NIKDY
prompt = f"""
Debug my database connection:
Connection string: postgresql://admin:SuperSecretPass123@prod-db.company.com:5432/production
"""

# ✅ SPRÁVNĚ
prompt = f"""
Debug my database connection:
Connection string: postgresql://[USER]:[PASSWORD]@[HOST]:5432/[DATABASE]
Error: Connection timeout after 30 seconds
"""

2. API Keys & Tokens

# ❌ NIKDY
prompt = f"Why doesn't this work? api_key = 'sk-ant-api03-xxxxx...'"

# ✅ SPRÁVNĚ
prompt = "Why doesn't my API call work? I'm getting 401 Unauthorized."

3. Private Keys

# ❌ NIKDY
prompt = f"""
-----BEGIN RSA PRIVATE KEY-----
MIIEpAIBAAKCAQEA...
-----END RSA PRIVATE KEY-----
Why can't I decrypt with this key?
"""

# ✅ SPRÁVNĚ
prompt = "I'm getting 'invalid key format' when decrypting. What are common causes?"

4. Personally Identifiable Information (PII)

# ❌ NIKDY
prompt = f"""
Analyze this customer data:
Name: John Smith
SSN: 123-45-6789
Credit Card: 4111-1111-1111-1111
"""

# ✅ SPRÁVNĚ
prompt = f"""
Analyze this anonymized customer data:
customer_id: C12345
purchase_history: [electronics, clothing]
loyalty_tier: gold
"""

5. Health Information (PHI)

# ❌ NIKDY
prompt = f"Patient Jane Doe, DOB 1985-03-15, diagnosis: ..."

# ✅ SPRÁVNĚ
prompt = "Patient ID P-1234, age group 35-40, condition category: chronic"

Vysoké riziko ⚠️

Interní business data

# ⚠️ ZVÁŽIT
prompt = "Our Q4 revenue was $50M, how do we present this to investors?"

# ✅ BEZPEČNĚJŠÍ
prompt = "How to present strong quarterly results to investors? General best practices."

Proprietary algorithms

# ⚠️ ZVÁŽIT
prompt = f"""
Our secret recommendation algorithm:
{full_algorithm_code}
How can we optimize it?
"""

# ✅ BEZPEČNĚJŠÍ
prompt = """
I have a recommendation algorithm with these characteristics:
- Collaborative filtering based
- Uses user embeddings
- Current latency: 50ms
How can I optimize performance?
"""

Unreleased product info

# ⚠️ ZVÁŽIT
prompt = "Our new product 'SecretProject' launches in March with features X, Y, Z..."

# ✅ BEZPEČNĚJŠÍ
prompt = "How to plan a product launch for a B2B SaaS tool?"

Sanitization helpers

import re

class DataSanitizer:
    PATTERNS = {
        'email': r'\b[A-Za-z0-9._%+-]+@[A-Za-z0-9.-]+\.[A-Z|a-z]{2,}\b',
        'phone': r'\b\d{3}[-.]?\d{3}[-.]?\d{4}\b',
        'ssn': r'\b\d{3}-\d{2}-\d{4}\b',
        'credit_card': r'\b\d{4}[-\s]?\d{4}[-\s]?\d{4}[-\s]?\d{4}\b',
        'api_key': r'\b(sk-|api_|key_|token_)[a-zA-Z0-9]{20,}\b',
        'ip_address': r'\b\d{1,3}\.\d{1,3}\.\d{1,3}\.\d{1,3}\b',
    }
    
    @classmethod
    def sanitize(cls, text: str) -> str:
        for name, pattern in cls.PATTERNS.items():
            text = re.sub(pattern, f'[REDACTED_{name.upper()}]', text)
        return text
    
    @classmethod
    def check(cls, text: str) -> list:
        """Returns list of found sensitive data types"""
        found = []
        for name, pattern in cls.PATTERNS.items():
            if re.search(pattern, text):
                found.append(name)
        return found

# Použití
text = "Contact john@company.com or call 555-123-4567"
sanitized = DataSanitizer.sanitize(text)
# "Contact [REDACTED_EMAIL] or call [REDACTED_PHONE]"

Pre-flight check

def safe_prompt(prompt: str) -> str:
    """Check prompt before sending to Claude"""
    
    issues = DataSanitizer.check(prompt)
    
    if issues:
        print(f"⚠️  WARNING: Found sensitive data: {issues}")
        print("Sanitizing before sending...")
        prompt = DataSanitizer.sanitize(prompt)
    
    return prompt

# Automaticky v client wrapper
class SafeClaudeClient:
    def __init__(self):
        self.client = Anthropic()
    
    def messages_create(self, **kwargs):
        if 'messages' in kwargs:
            for msg in kwargs['messages']:
                if isinstance(msg.get('content'), str):
                    msg['content'] = safe_prompt(msg['content'])
        
        return self.client.messages.create(**kwargs)

Bezpečnost dat je základ důvěryhodného AI systému.