Skip to main content

​
Content Safety Policy

Laozhang API follows strict content safety standards to ensure service compliance and security.

​
Prohibited Content

The following types of content are strictly prohibited:
Serious ViolationsThese behaviors will result in immediate account suspension:
  • Illegal content (fraud, money laundering, illegal transactions, etc.)
  • Child safety-related content (CSAM)
  • Violent extremism and terrorism
  • Malicious code and hacking activities
Prohibited Content Types
CategoryDescriptionExamples
Illegal ActivitiesViolating laws and regulationsFraud, drug trafficking, illegal weapon transactions
Harmful ContentEndangering public safetyCreating viruses, hacking tutorials, explosive manufacturing
Violent ContentExtreme violence, goreDetailed descriptions of torture, murder guides
Hate SpeechDiscrimination and incitementRacial discrimination, religious persecution incitement
HarassmentTargeted attacks on individualsCyberbullying, privacy violation, doxxing
Adult ContentPornographic, sexual contentExplicit sexual descriptions, pornographic images
MisinformationDeliberately spread false informationMedical fraud, financial fraud, political rumors
InfringementIntellectual property violationsPiracy, plagiarism, unauthorized use

​
Permitted Content

The following content can be used normally within reasonable scope:
Legitimate Use Cases
  • Educational and research purposes (e.g., security research, medical research)
  • Creative works (novels, scripts, artwork)
  • News reporting and commentary
  • Technical discussions (programming, cybersecurity)
  • Customer service and consulting
  • Content translation and summarization

​
Content Review Mechanism

​
Automated Review

All API requests undergo automated content review: 
User Request
    ↓
Automated Content Analysis
    ↓
Risk Assessment
    ↓
Decision: Pass / Warning / Block
Review Dimensions:
  • Content category classification
  • Sensitive word matching
  • Context semantic analysis
  • User behavior pattern
  • Historical violation records

​
Manual Review

High-risk content is manually reviewed:
  • Flagged suspicious requests
  • User appeals
  • Repeated violations
  • Edge cases

​
Appeal Process

1

Submit Appeal

If you believe content was misjudged, submit an appeal in console
2

Provide Evidence

Provide relevant evidence:
  • Request ID
  • Content description
  • Usage context
  • Compliance explanation
3

Wait for Review

Manual review team processes within 1-3 business days
4

Get Results

Receive review results via email:
  • Pass: Resume service
  • Reject: Maintain original decision

​
Violation Handling

​
Handling Measures

Depending on violation severity, corresponding measures will be taken:
Violation LevelHandling MeasuresDuration
First Minor ViolationWarning notice-
Repeated Minor ViolationsTemporary suspension24 hours
Moderate ViolationService restriction7 days
Serious ViolationAccount suspensionPermanent
Extremely SeriousAccount termination + legal reportPermanent

​
Violation Examples

Violation Content: User requests generation of phishing emails, fraud scriptsReview Result: Immediately blocked, account suspendedHandling Measures:
  • Permanently terminate account
  • Report to relevant authorities
  • Freeze remaining balance
Violation Content: Generate explicit pornographic textReview Result: Content blocked, record violationHandling Measures:
  • First offense: Warning
  • Second offense: 7-day suspension
  • Third offense: Permanent suspension
Violation Content: Generate false medical advice, cure claimsReview Result: Content blocked, evaluate intentHandling Measures:
  • Educational research purpose: Warning
  • Malicious spread: Account suspension
  • Serious consequences: Legal report
Request Content: Discuss web security vulnerabilities, SQL injection principlesReview Result: Within reasonable scope, permittedConditions:
  • Educational purpose
  • No specific attack targets
  • No malicious code
  • No detailed exploitation tutorials

​
Compliance Guidelines

​
Development Best Practices

Ensure application compliance:
  1. Content Filtering 
    def filter_user_input(text):
    """Filter user input"""
    # Check sensitive words
    if contains_sensitive_words(text):
        return "Input contains inappropriate content"
    
    # Validate content length
    if len(text) > MAX_LENGTH:
        return "Input too long"
    
    return text
  2. Logging and Auditing 
    import logging

def log_api_call(user_id, prompt, response):
    """Log API calls for compliance audit"""
    logging.info(f"User: {user_id}")
    logging.info(f"Prompt hash: {hash(prompt)}")
    logging.info(f"Response hash: {hash(response)}")
  3. User Consent 
    def get_user_consent():
    """Ensure user agrees to content policy"""
    consent_text = """
    By using this service, you agree to:
    1. Not generate illegal content
    2. Not use for fraudulent activities
    3. Comply with local laws and regulations
    4. Take responsibility for generated content
    """
    return user_accepts(consent_text)

​
User-facing Applications

If providing services to end users:
1

Clear Usage Terms

Display user agreement and content policy in prominent location
2

Input Validation

Validate and filter user input
3

Output Filtering

Review and filter AI-generated content
4

Reporting Mechanism

Provide reporting channel for users to report violations
5

Record Retention

Retain necessary logs according to legal requirements

​
Special Scenarios

​
Educational and Research

If used for education or research:
  • Clearly indicate purpose
  • Limit scope of use
  • Implement access controls
  • Record research process

​
Content Creation

If used for creative work:
  • Mark as fictional content
  • Add appropriate content warnings
  • Follow publishing platform policies
  • Respect intellectual property
AI-generated content cannot replace professional advice
  • Clearly indicate AI-generated
  • Recommend consulting professionals
  • Add disclaimer
  • Do not provide diagnostic advice

​
Technical Implementation

​
Content Filtering SDK

from laozhang_safety import ContentFilter

# Initialize content filter
filter = ContentFilter(api_key="your_api_key")

def safe_completion(prompt):
    """Safely call completion API"""
    
    # Pre-check input
    input_check = filter.check_input(prompt)
    if not input_check.is_safe:
        return {
            "error": "Input contains inappropriate content",
            "details": input_check.violations
        }
    
    # Call API
    response = client.chat.completions.create(
        model="gpt-4-turbo",
        messages=[{"role": "user", "content": prompt}]
    )
    
    # Post-check output
    output_check = filter.check_output(response.choices[0].message.content)
    if not output_check.is_safe:
        return {
            "error": "Generated content contains inappropriate material",
            "details": output_check.violations
        }
    
    return response

​
Custom Safety Rules

# Define custom safety rules
custom_rules = {
    "allowed_categories": [
        "education",
        "creative_writing",
        "customer_service"
    ],
    "blocked_keywords": [
        "your_custom_sensitive_words"
    ],
    "max_risk_score": 0.3  # Maximum acceptable risk score
}

filter = ContentFilter(
    api_key="your_api_key",
    custom_rules=custom_rules
)

​
Common Questions

False Positive Handling:
  1. Immediate Action:
    • Check error message details
    • Adjust request content
    • Try alternative expressions
  2. If Repeatedly False Positives:
    • Submit appeal request
    • Provide usage context
    • Explain legitimate purpose
    • Wait for manual review
  3. Prevention Suggestions:
    • Avoid sensitive words
    • Provide clear context
    • Use professional terminology
    • Add usage intent description
Compliance Checklist:
  • Read and understand content policy
  • Implement input validation
  • Add output filtering
  • Display user agreement
  • Implement logging
  • Set up reporting mechanism
  • Regular compliance audits
  • Staff training
Appeal Rights:Yes, can appeal within 30 days of decision. Submit:
  • Detailed explanation
  • Improvement measures
  • Compliance commitment
  • Supporting evidence
Appeal Process:
  1. Submit appeal form in console
  2. Provide necessary materials
  3. Wait for review (5-7 business days)
  4. Receive review results
Violation Record:
  • Records retained for 1 year
  • Considered in review
  • Affect trust level
Record Elimination:
  • Minor violations: Eliminated after 6 months without new violations
  • Moderate violations: Eliminated after 1 year without new violations
  • Serious violations: Permanent record

​
Contact Us

If questions about content safety: