Security & trust

Evidence before promises.

FACOS is being built as a governed AI workforce platform with business isolation, constrained worker authority, controlled external execution, incident response, recovery controls and fail-closed production gates. This Trust Center explains what is active now and what FACOS will not claim until the evidence exists.

Current lifecycle: Controlled preview / pre-production.
FACOS has working security controls and deployed preview services. Broad customer-production certification, independent penetration testing, production KMS/HSM encryption and production recovery evidence are not yet claimed.

Security foundation

The platform uses multiple independent layers rather than relying on a model, password, network rule or hosting provider as a single security boundary.

Business-scoped authorization

Runtime identity, jobs, permissions and protected references carry business scope. Cross-business access is treated as a security violation.

Active foundation

Nonhuman authority ceiling

AI workers cannot promote themselves into Owner authority. Material actions use deterministic authorization and human/Owner boundaries outside the model.

Active foundation

Secrets stay outside prompts

FACOS uses secret-scanning and credential-broker/handle patterns so permanent provider credentials are not intended to be handed directly to AI workers.

Active foundation

Fail-closed releases

Release gates run syntax, policy, security, hostile-input, dependency, network-route, recovery and trust-claim checks before deployment.

Active

SBOM & release evidence

CycloneDX software bills of materials and SHA-256 release evidence are generated from the locked dependency/source state.

Active · unsigned

Independent availability checks

FACOS combines internal observability with GitHub Actions availability monitoring outside the Render runtime.

Active

AI-worker governance

FACOS treats the model as a worker inside a control system—not as the control system itself.

Worker boundaries

Exact capability scope

Workers operate against explicit business, capability, resource, data-class and environment boundaries.

ACTIVE
Prompt-injection defense

External content is treated as data, not authority. Hostile override, secret-extraction and escalation patterns are tested during releases.

OFFLINE TESTED
Governed outbound network

Known routes use host/method allowlists, HTTPS, request/response bounds and private-network protections.

APP LAYER
Live model/tool adversarial certification

The exact future production model and complete tool stack still require live adversarial validation.

PENDING

Human authority

Owner-gated actions

High-risk authority cannot be created merely because a model, web page, document or tool result asks for it.

ACTIVE
Containment

FACOS supports worker, business and integration containment when a serious security condition is detected.

ACTIVE
Raw permanent secret access

Worker policy rejects direct raw-secret access patterns in favor of controlled application bridges.

DENIED
Infrastructure-level egress

Independent network enforcement outside the application process remains a production-readiness requirement.

PENDING

Data protection & resilience

FACOS is designed to minimize unnecessary copies of sensitive data and preserve enough evidence to investigate, recover and prove what happened.

Data classification

Protected source data and AI-derived material such as prompts, outputs, caches, evaluations and tool results retain protected handling expectations.

Incident response

Critical response covers evidence preservation, containment, session/credential revocation, rotation, recovery validation, monitoring and post-incident review.

Recovery tooling

FACOS includes backup-integrity and restore-verification tooling. Real production restore evidence remains a required production gate.

Encryption transparency: FACOS has application envelope-encryption architecture using AES-256-GCM and an external KMS/HSM design, but live production KMS/HSM activation and protected-object-storage rollout are still pending. FACOS does not claim that every preview record is already protected by application-level envelope encryption.

Production readiness

These items stay visible because a serious trust program should show the work that remains—not hide it behind marketing language.

Implemented / evidenced foundation

Private source + locked dependenciesACTIVE
Dependency and committed-secret scansACTIVE
Business isolation + worker authority controlsACTIVE
Prompt-injection offline defense testsACTIVE
Incident-response tabletop and monitoringACTIVE
SBOM + hash-bound release evidenceACTIVE

Still required before broad production

Production database + tested PITR/restorePENDING
Live KMS/HSM + protected object storagePENDING
Live model/tool adversarial validationPENDING
Signed/trusted release provenancePENDING
Independent penetration/security testPENDING
Final counsel-reviewed privacy/DPA/subprocessor termsPENDING
Certification disclosure: FACOS does not currently claim a completed SOC 2 report, ISO 27001 certificate, HIPAA authorization, or equivalent independent security certification.

Current platform dependencies

This is a preview-environment dependency disclosure, not yet the final contractual subprocessor register.

Render

Preview web services and preview PostgreSQL hosting.

Production database and recovery evidence still pending.

GitHub

Private source control, CI/security checks, release evidence and external availability monitoring.

Active engineering/security dependency.

GoDaddy

Domain registration and DNS administration.

Infrastructure dependency; not an application data store.

OpenAI

AI-provider integration path in FACOS.

Live customer-production model/tool certification not yet completed.

Security reporting

Security reports should be private-first and should never contain live passwords, API keys, recovery codes or bearer tokens.

Contact

Current FACOS security contact: thebrysonmethod@gmail.com

This Trust Center reflects the current FACOS controlled-preview posture as of October 1, 2026. Controls and production gates will be updated as evidence changes.