Nexus Shield is built for regulated enterprises that require sub-10ms guardrails without sacrificing data sovereignty. Our Trust Center summarizes how we protect customer prompts, align with global compliance frameworks, and maintain a zero-training, zero-retention posture by default.
- SOC 2 Type II — Security, availability & confidentiality controls mapped to AICPA Trust Services Criteria.
- ISO/IEC 27001 — Information Security Management System (ISMS) controls for risk treatment & continuous improvement.
- GDPR & KVKK — Privacy-by-design PII redaction, data minimization, and lawful processing support.
- Zero-Data Retention — No persistent storage of raw prompts; optional Redis cache TTL only.
Security inquiries: [email protected]
SOC 2 Type II — Trust Services Criteria
Nexus Shield operates under a control framework aligned with SOC 2 Type II (AICPA Trust Services Criteria). Our architecture and operational procedures address the following trust principles:
Security (Common Criteria)
- Role-based access control (RBAC) for production infrastructure and API key authentication (
X-API-Key).
- Encryption in transit (TLS 1.2+) via Cloudflare Tunnel and nginx gateway; encryption at rest for infrastructure volumes.
- Early Exit engine blocks prompt injection & jailbreak attempts before upstream LLM calls (CC6.1, CC6.6).
- Automated CI/CD pipeline with pytest security regression suite and container image signing.
Availability
- Health probes (
/healthz), Docker healthchecks, and nginx upstream failover.
- Sub-10ms inspection SLA with Redis optional caching layer for idempotent requests.
Confidentiality & Privacy
- PII redaction (TCKN, credit card, email, phone) applied before any third-party LLM provider receives data.
- Zero-Data Retention policy — raw prompts are not written to durable storage (see Retention tab).
- Customer data is never used to train Nexus Shield or foundation models.
SOC 2 Type II attestation report available to enterprise customers under NDA. Contact [email protected].
ISO/IEC 27001 — Information Security Management
Nexus Shield aligns its ISMS with ISO/IEC 27001:2022 Annex A controls. Key control domains implemented in our production stack:
A.5 — Organizational
Information security policies, roles & responsibilities, supplier (sub-processor) agreements for LLM providers.
A.8 — Asset Management
Inventory of API keys, container images, and configuration baselines; acceptable use of customer prompt data.
A.9 — Access Control
Least-privilege SSH deploy keys, API key rotation, rate limiting (RATE_LIMIT_REQUESTS), and network segmentation (Docker bridge).
A.12 — Operations Security
Vulnerability management via dependency scanning, logging & monitoring (Prometheus/Grafana SOC dashboard), change management via GitHub Actions.
A.14 — System Development
Secure SDLC: code review, automated pytest (403 injection blocks, PII masking, SLA <10ms), separation of Fast API guardrail vs ML pipeline.
A.18 — Compliance
GDPR, KVKK (Law No. 6698), and contractual DPAs; privacy impact assessments for on-premise deployments.
ISO 27001 Statement of Applicability (SoA) available upon request for enterprise due diligence.
GDPR / KVKK Manifesto
Nexus Shield Privacy & Data Protection Commitment — effective 2026
1. Lawful Basis & Purpose Limitation (GDPR Art. 5–6 · KVKK m.4–5)
We process prompt data solely to provide AI firewall, PII redaction, and security inspection services. Processing is limited to what is necessary to fulfill the contract with the data controller (your organization).
2. Privacy by Design & by Default (GDPR Art. 25 · KVKK m.12)
- Automatic detection & masking of TCKN, credit card numbers, email addresses, and phone numbers before LLM forwarding.
- Early Exit blocking of jailbreak / prompt injection — malicious payloads never reach third-party models.
- Default-deny security posture: suspicious requests return HTTP 403 without upstream disclosure.
3. Data Subject Rights (GDPR Art. 15–22 · KVKK m.11)
Because Nexus Shield adheres to Zero-Data Retention, we do not maintain identifiable prompt archives. Data subject access, rectification, and erasure requests are therefore inherently satisfied at the processing layer — no persistent copy exists to retrieve or delete.
4. International Transfers (GDPR Art. 44–49 · KVKK m.9)
For SaaS deployments, sanitized (PII-redacted) prompts may be forwarded to LLM sub-processors (OpenAI, Anthropic, Google) under Standard Contractual Clauses (SCCs). On-premise / private cloud deployments keep all data within the customer's jurisdiction.
5. Security of Processing (GDPR Art. 32 · KVKK m.12)
TLS encryption, API key authentication, rate limiting, audit logging via SOC dashboard, and continuous automated security testing.
6. Data Protection Officer / Contact
Privacy inquiries: [email protected] · General: [email protected]
Zero-Data Retention Policy
Last updated: July 2026
Policy Statement
Nexus Shield does not persistently store raw customer prompts or LLM responses. All inspection occurs in volatile memory. Upon request completion, prompt content is discarded unless explicitly covered by the optional cache exception below.
What We Do NOT Store
- Raw user prompts containing PII (masked in-flight, never logged at rest).
- LLM provider responses tied to identifiable session data.
- Conversation histories for model training or analytics.
Optional Redis Cache (TTL-Bound)
When enabled, sanitized prompt hashes may be cached in Redis for up to REDIS_CACHE_TTL_SEC (default: 3600 seconds) to reduce latency on identical requests. Cache entries contain redacted text only — never raw PII. Customers may disable Redis entirely for strict zero-retention deployments.
Operational Logs
Aggregated, non-content metrics (request counts, block rates, latency percentiles) are retained for SLA monitoring via the SOC dashboard. Logs do not include prompt body content.
On-Premise Guarantee
Enterprise on-premise deployments ensure prompts never leave the customer's VPC. Nexus Shield staff have no access to production prompt data.