TOMs ALAI AI Services v1
Technical and Organizational Measures (TOMs)
ALAI Holding AS — AI Services
Version: 1.0 | Date: 2026-05-01 | GDPR Reference: Article 32
Overview
This document describes the technical and organizational measures (TOMs) implemented by ALAI Holding AS to ensure the security of personal data processed on behalf of clients in connection with AI Services.
ALAI acts as a Data Processor when delivering AI services (AI audits, AI development, AI agent orchestration) to clients. This document satisfies GDPR Article 28(3)(c) requirement to demonstrate appropriate security measures.
Technical Measures
2.1 Encryption
| Measure | Implementation | Purpose |
|---|---|---|
| Data in Transit | TLS 1.3 for all HTTP connections; SSH for server access | Protect data during transmission |
| Data at Rest | AES-256 encryption for PostgreSQL databases, file storage, and backups | Prevent unauthorized access to stored data |
| API Keys and Secrets | Stored in Bitwarden (encrypted vault); environment variables in production; never committed to git | Protect credentials |
| TLS for SMTP/IMAP; PGP available for sensitive communications | Secure email in transit |
Implementation Details:
- All client-facing web services enforce HTTPS via Cloudflare SSL
- Database connections use SSL/TLS (Azure PostgreSQL enforces encrypted connections)
- Backups encrypted at rest using Azure Storage encryption (Microsoft-managed keys)
2.2 Pseudonymization
| Measure | Implementation | Purpose |
|---|---|---|
| Development/Test Data | Client production data is anonymized before use in dev/test environments | Minimize exposure of real personal data |
| Logging | Personal identifiers (emails, names) are redacted or hashed in system logs | Prevent leakage via logs |
| AI Training Data | Client data used for AI model fine-tuning is pseudonymized where feasible | Protect individual identities in training datasets |
2.3 Access Control
| Measure | Implementation | Purpose |
|---|---|---|
| Multi-Factor Authentication (MFA) | Required for all production system access (Azure Portal, SSH, Bitwarden, Documenso, BookStack admin) | Prevent unauthorized access |
| Role-Based Access Control (RBAC) | Azure AD roles limit production access to designated personnel only | Need-to-know principle |
| SSH Key Authentication | Password authentication disabled; only SSH keys allowed for server access | Prevent brute-force attacks |
| API Token Rotation | Quarterly rotation of service API tokens | Limit token exposure window |
2.4 Logging and Monitoring
- Audit Logs: All access to production databases and personal data logged with timestamps, user ID, action
- Retention: Logs retained for 90 days (compliance requirement)
- Alerting: Failed login attempts, unauthorized access attempts trigger alerts to CEO
- Review: Monthly log review for anomalies
2.5 Security Updates
- Patch Cycle: Monthly security updates for OS and dependencies
- Dependency Scanning: Automated vulnerability scanning via Dependabot (GitHub) and npm audit
- Critical Patches: Applied within 48 hours of disclosure for high-severity vulnerabilities
2.6 Penetration Testing
- Frequency: Annual external penetration testing (planned Q4 2026)
- Scope: Web applications, API endpoints, authentication mechanisms
- Remediation: High/critical findings remediated within 30 days
Organizational Measures
3.1 Personnel Security
| Measure | Implementation | Purpose |
|---|---|---|
| GDPR Training | Annual GDPR training for all staff with data access | Ensure awareness of data protection obligations |
| Confidentiality Agreements | All employees and contractors sign NDAs covering client data | Legally binding confidentiality |
| Background Checks | Reference checks for all hires with production access (Norway/Bosnia) | Vet trustworthiness |
| Access Termination | All access revoked within 24 hours of employee/contractor departure | Prevent ex-employee access |
3.2 Access Management
- Need-to-Know Principle: Access granted only when documented business need exists
- Access Review: Quarterly review of who has production access; revoke unnecessary permissions
- Temporary Access: Time-limited credentials for contractors (expire after engagement)
3.3 Backup and Recovery
| Measure | Implementation | Target |
|---|---|---|
| Backup Frequency | Daily automated backups of all databases | RPO: 24 hours (max data loss) |
| Backup Location | Azure Blob Storage (geo-redundant, EU region) | Survive regional outage |
| Recovery Time | Tested quarterly restore procedures | RTO: <24 hours (time to restore) |
| Backup Encryption | AES-256 encryption at rest | Protect backup data |
3.4 Incident Response
Data Breach Response Plan:
- Detection: Automated alerts + manual log review
- Containment: Immediate isolation of affected systems (within 1 hour)
- Assessment: Determine scope: what data, how many records, what breach type
- Notification:
- Client notification within 24 hours of breach discovery (per DPA)
- Datatilsynet (Norwegian DPA) notification within 72 hours if required by GDPR
- Remediation: Patch vulnerability, restore from backup if needed
- Documentation: Full incident report with timeline, root cause, remediation steps
Incident Contact: [email protected] (CEO, available 24/7 for critical incidents)
3.5 Data Retention and Deletion
| Data Type | Retention Period | Deletion Method |
|---|---|---|
| Client personal data (production) | Duration of contract + 30 days post-termination | Secure deletion (multi-pass overwrite or Azure storage deletion) |
| Backups | 90 days rolling window | Automatic expiry |
| Audit logs | 90 days | Automatic expiry |
| Signed contracts (NDA, Retainer, DPA) | 7 years (Norwegian accounting law) | Archived at archive.alai.no per ZAKON ARCHIVE FIRST |
Data Deletion Verification: Upon contract termination, ALAI provides written confirmation of data deletion within 30 days (per DPA section 3.7).
Sub-Processor Security
ALAI relies on sub-processors for infrastructure and AI services. Each sub-processor has been vetted for GDPR compliance:
| Sub-Processor | Certifications | Data Location |
|---|---|---|
| Anthropic PBC | SOC 2 Type II, GDPR DPA, SCCs | USA (AWS us-east-1) |
| Microsoft Azure | ISO 27001, SOC 2, GDPR compliant | EU West / Norway East |
| Cloudflare Inc. | ISO 27001, SOC 2 Type II, GDPR DPA | Global (EU data residency) |
| Brevo | GDPR compliant, ISO 27001 | EU (Frankfurt) |
See DPA Template for full sub-processor details and 30-day notice policy.
Compliance and Audit
- Annual TOMs Review: CEO reviews and updates this document annually or upon material security changes
- Client Audit Rights: Clients may audit ALAI's compliance with TOMs (1x/year free, additional by agreement)
- External Audit: SOC 2 Type I audit planned Q4 2026 (post-AI Services launch)
Limitations and Disclaimers
⚠️ Current Status: DRAFT
This TOMs document is based on ALAI's existing infrastructure and planned security posture. Final validation pending:
- Security audit: External review not yet conducted (planned Q4 2026)
- ISO 27001: Not yet certified (est. cost 150K NOK, 6-month timeline if client requires)
- SOC 2: Type I audit planned Q4 2026 (Type II requires 6-12 month observation period)
If client requires formal certification (ISO 27001, SOC 2 Type II), CEO will assess feasibility and cost impact.
Document History
- v1.0 (2026-05-01): Initial version (Lexicon + Proveo 19/20 PASS)
- Next Review: 2027-05-01 (annual) or upon first security audit
Source File: ~/Public/legal/ai-services/TOMs-ALAI-AI-Services-v1.md
Full document available at source location (13K file).
Referenced by DPA Template v1 as Annex B.
No comments to display
No comments to display