SOC 2 Type II
99% policy coverage
{"audit":"token_audit","scope":"egress","ts":"2025-01-15T12:00:00Z"}UnblockLLM supports SOC 2 Type II with immutable audit logs and no PII in evidence.
Zero-trust proxy for OpenAI, Anthropic, and custom LLMs — block PII, enforce policies, and audit every token.
Token-level inspection (no full payload decryption)
0ms decryption overhead
100% token visibility
Custom rules: block credit_card if region=EU. Policy evaluated per token.
Per-token evaluation
YAML or dashboard config
On-the-fly PII masking (retain token count). No payload size change.
Token count preserved
Streaming redaction
Immutable logs for SOC 2 compliance. Metadata only, no PII.
90-day retention
SOC 2 evidence
curl -X POST https://api.unblockllm.com/v1/chat/completions \
-H "Authorization: Bearer YOUR_API_KEY" \
-d '{"messages":[{"role":"user","content":"Hello"}]}'99% policy coverage
{"audit":"token_audit","scope":"egress","ts":"2025-01-15T12:00:00Z"}UnblockLLM supports SOC 2 Type II with immutable audit logs and no PII in evidence.
98% policy coverage
{"entity":"PHI","action":"redacted","ts":"2025-01-15T12:00:00Z"}UnblockLLM passes HIPAA by never logging PHI and enforcing redaction before egress.
100% policy coverage
{"request_id":"req_xyz","entity_types":["EMAIL"],"region":"EU"}UnblockLLM passes GDPR via data minimization and configurable retention (90-day default).
97% policy coverage
{"consumer_data":"masked","jurisdiction":"CA"}UnblockLLM passes CCPA by masking PII in real time and supporting right-to-deletion workflows.
Your Skills Stay Powerful, Now Privacy-Compliant
Blocks GITHUB_TOKEN in all skill inputs
No code changes required
Risk:
Exposes repo tokens in prompts
How we secure it:
Blocks GITHUB_TOKEN in all skill inputs
$ npx skills add skills/github
→ [UNBLOCKLLM] Scanning skill dependencies...
→ [POLICY] Detected: GITHUB_TOKEN exposure risk
→ [ACTION] Injected redaction layer
→ [STATUS] ✅ Skill secured & installedRisk:
Leaks PII to Anthropic
How we secure it:
Redacts names/emails in real-time
$ npx skills add skills/claude
→ [UNBLOCKLLM] Scanning skill dependencies...
→ [POLICY] Detected: PII in prompt risk
→ [ACTION] Injected redaction layer
→ [STATUS] ✅ Skill secured & installedRisk:
Sends sensitive queries to Google
How we secure it:
Scrubs query parameters pre-request
$ npx skills add skills/websearch
→ [UNBLOCKLLM] Scanning skill dependencies...
→ [POLICY] Detected: query parameter exposure risk
→ [ACTION] Injected redaction layer
→ [STATUS] ✅ Skill secured & installed0ms
Latency overhead
100K
RPS throughput
99.999%
Uptime
Live policy enforcement (verifiable):
curl -X POST "${UNBLOCKLLM_BASE_URL}/v1/chat/completions" \
-H "Authorization: Bearer ${OPENAI_API_KEY}" \
-d '{"model":"gpt-4o-mini","messages":[{"role":"user","content":"My SSN is 123-45-6789."}]}'
# → 400 or redacted per policyDeploy in 8 minutes — no code changes to your app.
For enterprise inquiries, support, or partnerships.