Security
Last updated: March 18, 2026
Our security commitment
Contracts are among the most sensitive business documents. We built ClauseGuard with a security-first architecture: documents are processed in memory and never stored as raw files, all stored data is encrypted at rest with AES-256, encrypted in transit with TLS 1.3, and never used for AI model training. Row-level security ensures users can only access their own data. We use enterprise-grade infrastructure from providers with SOC 2 Type 2 certification.
Data handling
Two modes of operation
ClauseGuard operates in two distinct modes, each with different storage behaviors:
Upload a contract and get AI analysis instantly. The document is processed in memory, analyzed, and the results are shown in your browser. Nothing is stored — when you close the tab, the analysis is gone. No account required.
When you sign in and save a contract, the AI analysis, extracted text, and metadata are stored in your encrypted database. This enables lifecycle management: status tracking, approvals, signatures, collaboration, and alerts. You control your data — delete any contract at any time.
Document processing
When you upload a contract, it is sent over TLS 1.3 encrypted connections to our Vercel serverless infrastructure. The document is parsed in memory, sent to the Anthropic API for AI analysis, and the structured results are returned to your browser. The raw document file (PDF/DOCX) is not stored as a file on our servers.
What we store
| Data | Quick Scan | CLM (Saved) |
|---|---|---|
| Raw document file | No | No |
| Extracted text | No | Yes (encrypted) |
| AI analysis results | No | Yes (encrypted) |
| Tasks, comments, activity | N/A | Yes (encrypted) |
| Signatures | N/A | Yes (encrypted, with audit) |
| Account info | N/A | Yes |
| Payment data | No | No (Stripe PCI L1) |
Data deletion
You can delete individual contracts and all associated data (tasks, comments, signatures, activity) at any time from your contracts page. If you delete your account, all associated data is permanently removed within 30 days.
AI model security
Zero-training guarantee
Your contracts are never used to train, fine-tune, or improve any AI model. We use Anthropic's Claude API, which has a contractual commitment that API inputs and outputs are not used for model training. Your data goes in, analysis comes out, nothing is retained by the AI provider.
Contract text is sent to Anthropic's API over encrypted connections for real-time analysis. Anthropic does not store API inputs or outputs beyond the duration of the request. See Anthropic's privacy commitments for details.
Infrastructure security
Application hosting
Edge network with automatic TLS, DDoS protection, serverless isolation
Authentication & database
PostgreSQL with row-level security, encrypted at rest (AES-256), hosted on AWS
Payment processing
We never see or store credit card numbers
AI analysis engine
No data retention, no model training on API inputs
Transactional email
DKIM, SPF, and DMARC authenticated sending
Encryption
- In transit: All data transmitted between your browser and ClauseGuard is encrypted using TLS 1.3. All API calls to third-party services (Anthropic, Stripe, Supabase) are also encrypted in transit.
- At rest: Database storage is encrypted at rest using AES-256 encryption, managed by Supabase/AWS infrastructure.
- Authentication tokens: Session tokens are encrypted, HTTP-only, and SameSite protected. API keys are stored as environment variables, never exposed to the client.
Access control
- Authentication: Google OAuth 2.0 or email magic link (passwordless). No passwords stored.
- Authorization: Row-level security (RLS) enforced on all 34 database tables. Users can only access their own data. Team data is scoped to team membership.
- Middleware auth gating: 28 API routes require authentication. 11 pages redirect to login when not authenticated.
- Input validation: Zod schema validation on all AI-calling and data-modifying API routes prevents injection and malformed requests.
- Immutable audit trail: All sensitive actions are logged to an append-only audit table that cannot be modified or deleted.
- Security headers: HSTS, X-Frame-Options: DENY, CSP (report-only), X-Content-Type-Options, Referrer-Policy, and Permissions-Policy on all responses.
- Counterparty access: External parties access contracts only via 256-bit cryptographic tokens with configurable expiry and revocation.
- Signature integrity: Electronic signatures are captured with IP address, user agent, timestamp, and legal consent — creating a legally binding audit trail.
Incident response
In the event of a security incident involving your data, we will notify affected users within 72 hours of discovery, provide details about the nature and scope of the incident, describe the measures taken to address it, and recommend steps you can take to protect yourself. We maintain an incident response plan and conduct regular reviews of our security posture.
Responsible disclosure
If you discover a security vulnerability in ClauseGuard, please report it to security@theclauseguard.com. We take all reports seriously and will respond within 48 hours. We ask that you give us reasonable time to address the issue before public disclosure.