Security
Last updated: January 2026
1. Security Overview
Security is fundamental to how we build and operate Crawfish API. We employ industry-standard practices to protect your data, API keys, and ensure the integrity of our service.
🛡️ Our Security Principles:
- Minimize data collection - we don't store your prompts or responses
- Encrypt everything - at rest and in transit
- Defense in depth - multiple layers of security
- Principle of least privilege - minimal access by default
- Transparency - clear policies and honest commitments
2. Encryption
🔒 At Rest
AES-256
All stored data encrypted with AES-256-GCM
🌐 In Transit
TLS 1.3
All connections encrypted with TLS 1.3 (minimum TLS 1.2)
What's Encrypted:
- Database contents
- API keys (encrypted before storage)
- File system data
- Backups
- All API communications
- Dashboard sessions
3. API Key Security
API keys are the most sensitive piece of data in our system. We treat them with extra care:
🔑 Key Protection Measures:
- Encrypted at rest using AES-256 before storage
- Never logged in full - only last 4 characters shown
- Transmitted only over encrypted connections
- Hashed using SHA-256 for verification
- Regeneratable at any time
- Individual keys per user/team with granular permissions
⚠️ Key Security Best Practices:
- Never commit API keys to version control
- Use environment variables, not hardcoded values
- Rotate keys periodically
- Use separate keys for development and production
- Set spending limits on keys
- Monitor usage for anomalies
4. Data Privacy & Prompt Handling
We have a strict policy regarding your prompts and responses:
✅ What We DO:
- Forward your requests directly to model providers
- Log token usage (counts only, not content) for billing
- Store minimal data required for service operation
- Support zero-logging mode (where available)
❌ What We DON'T DO:
- Store the content of your prompts
- Store model responses
- Cache or retain request/response data
- Analyze your prompts for any purpose
- Share your prompts with third parties
- Use your data to train models
Model Provider Data Policies:
Your prompts are sent directly to model providers (OpenAI, Anthropic, etc.). Their data policies apply during processing:
- OpenAI: Enterprise Privacy
- Anthropic: Trust & Safety
- Google: Cloud Security
5. Access Control
We implement strict access controls at multiple levels:
🔐 Multi-Layer Access Control:
- Role-Based Access Control (RBAC): Users have defined roles with specific permissions
- API Key Hierarchy:
- - Master Key: Full account access
- - User Keys: Individual user permissions
- - Spend Limit Keys: Capped spending
- IP Allowlisting: Optional IP-based access restrictions (enterprise)
- Multi-Factor Authentication: Available for account security
Internal Access:
- Employees require MFA for internal systems
- SSH access uses key-based authentication only
- Production access is logged and audited
- Least privilege principle applied
6. Audit Logging
Comprehensive logging for security and compliance:
Logged Events:
- Authentication events (login, logout, failures)
- API key creation, rotation, deletion
- Account changes (settings, team members)
- Billing and payment events
- Admin actions (internal)
- Security events (suspicious activity)
What We Don't Log:
The actual content of your prompts or model responses. We only log metadata such as token counts for billing purposes.
7. Infrastructure Security
Our infrastructure is hosted on secure, enterprise-grade cloud services:
🏢 Hosting:
- Provider: Tencent Cloud (Singapore region)
- Server Location: Singapore
- Type: Cloud servers with managed services
Infrastructure Protections:
- Network firewalls and security groups
- DDoS protection
- WAF (Web Application Firewall)
- Automatic security patching
- Regular vulnerability scanning
- Intrusion detection systems
8. Compliance & Certifications
✅ Currently Implemented:
- GDPR compliance for EU users
- CCPA compliance for California residents
- Data Processing Agreements (DPAs) available
- Industry-standard security practices
- Privacy-by-design principles
📋 In Roadmap:
- SOC 2 Type II: In planning phase, expected timeline TBD
- ISO 27001: Consideration phase
- Penetration Testing: Scheduled quarterly
Contact us for security questionnaires or additional compliance documentation.
9. Vulnerability Reporting
We welcome responsible disclosure of security vulnerabilities:
Report a Vulnerability:
If you discover a security vulnerability, please report it responsibly:
- Email: security@crawfish.top
- Response Time: Within 48 hours
- Bounty Program: Under consideration
What to Include:
- Description of the vulnerability
- Steps to reproduce
- Potential impact assessment
- Any suggested fixes (optional)
Out of Scope:
- Denial of service attacks
- Social engineering
- Physical security issues
- Vulnerabilities in third-party services
10. Contact
For security-related inquiries:
General Security: admin@crawfish.top
Urgent Issues: security@crawfish.top