Governed AI begins with controlled access,
clear authority, and traceable action.
Quantum is designed around dealer-controlled authorization, least privilege, consent-aware communication, tenant isolation, human ownership, data minimization, and auditable system activity.
Security, privacy, agent governance, and operational trust in one place.
Quantum publishes only current, verifiable security and compliance information. Certifications and formal attestations are listed only after completion.
Every consequential action should answer six questions.
- 01
Identity
Which dealership, user, buyer, vehicle, opportunity, and agent are involved?
- 02
Authority
What is the actor allowed to read, recommend, change, or communicate?
- 03
Consent
Which channel, purpose, timing, and disclosure are permitted?
- 04
Ownership
Who currently controls the customer conversation or operational decision?
- 05
Evidence
Which verified facts support the action?
- 06
Audit
What happened, why, through which system, and with what result?
The AI model does not decide its own permissions. Quantum evaluates authority before the model can use a consequential tool.
Access is activated, scoped, monitored, and revocable.
Dealer controls may include
- 01Authorized dealer group and rooftops
- 02Approved systems and accounts
- 03Permitted users and departments
- 04Data categories
- 05Read permissions
- 06Write permissions
- 07Communication channels
- 08Allowed agent actions
- 09Operating hours and quiet hours
- 10Retention requirements
- 11Activation date
- 12Expiration
- 13Suspension and revocation
Quantum uses the least access required for the approved workflow and records the dealership’s current authorization.
Dealer approval does not bypass vendor terms, API requirements, platform controls, or applicable law.
Every human, service, connector, and agent receives a defined identity.
Access controls
- 01Unique user identity
- 02Role-based access
- 03Dealer and rooftop scope
- 04Service-to-service identity
- 05Short-lived credentials where supported
- 06Multi-factor authentication where available
- 07Environment separation
- 08Least-privilege tool access
- 09Session expiration
- 10Credential rotation
- 11Access review
- 12Revocation
Shared personal employee credentials should not be used for production integrations. Quantum favors vendor-permitted service identities and dealer-controlled activation.
Every request is evaluated against the authorized dealer and rooftop.
Quantum separates dealership records through tenant-aware data access, scoped credentials, role rules, and audit context.
Authorized dealer + rooftop boundary
- Organizations and rooftops
- Users and roles
- Customers and opportunities
- Vehicles and listings
- Conversations and recordings
- Appointments
- Consent
- Recommendations
- Outcomes
- Integration credentials
- Knowledge sources
- Audit records
Isolation applies to every record class above, in every request path.
An agent never receives unrestricted cross-dealer context. The authorized tenant and opportunity are assembled for each execution.
Sensitive access belongs in managed security controls, not prompts or workflow notes.
Controls
- 01TLS for data in transit
- 02Encryption at rest through approved cloud services
- 03Managed secret storage
- 04Credential separation by environment
- 05Restricted production access
- 06Secret rotation
- 07No credentials in source code
- 08No credentials in public documents
- 09Sensitive-field redaction
- 10Encrypted backups
- 11Access and change logging
Quantum does not publish passwords, API keys, tokens, certificates, or dealer credentials in agent instructions, support messages, or shared documents.
Every buyer-facing action is evaluated against channel, purpose, timing, and ownership.
Consent record
- Buyer or contact point
- Channel
- Purpose
- Source
- Timestamp
- Disclosure version
- Expiration when applicable
- Opt-out
- Suppression state
- Dealership and rooftop
- Evidence
Communication controls
- Quiet hours
- Channel-specific permission
- Identity confirmation
- Opt-out enforcement
- Frequency limits
- One-voice ownership
- Human escalation
- Complaint handling
- Delivery status
- Record retention
A valid phone number or email address is not, by itself, permission to contact a person for every purpose.
Hannah and Coach share context, not permissions.
Buyer-facing
Hannah
May communicate with the buyer only through approved channels and only while buyer-facing authority is active.
Rep-facing
Coach
May advise the salesperson or manager. Coach does not contact the buyer.
Evaluated before either agent acts
- 01Current identity
- 02Dealer scope
- 03Opportunity
- 04Official CRM state
- 05Operational evidence
- 06Consent
- 07Ownership
- 08Allowed tools
- 09Action authority
- 10Escalation rule
If required context is missing, stale, conflicting, or outside authority, the agent asks, waits, or escalates.
Document what is collected, why it is used, and when it is deleted.
Published for every deployment
- 01Data categories collected
- 02Purpose of processing
- 03Source systems
- 04Dealer and Quantum roles
- 05Storage locations
- 06Retention periods
- 07Deletion process
- 08Export process
- 09Backup retention
- 10Legal hold behavior
- 11Anonymization
- 12Service provider involvement
- 13Post-termination handling
Dealership and customer data is not used for unrelated purposes or model training unless the applicable agreement, authorization, disclosure, and provider controls explicitly permit it.
Cloud regions, storage and backup locations, cross-border processing, and the retention schedule by data category are documented and provided in writing during security review and in the applicable agreement.
The model reasons. Quantum controls the context, tools, policy, and record.
Governance controls
- 01Approved model providers
- 02Model and prompt versioning
- 03Structured inputs and outputs
- 04Authorized context assembly
- 05Sensitive-data redaction
- 06Tool-level permissions
- 07Cost and latency limits
- 08Fallback behavior
- 09Human escalation
- 10Quality evaluation
- 11Unsupported-claim detection
- 12Provider retention settings
- 13Change testing before release
Quantum may use different models for voice, chat, recommendation, extraction, or evaluation. Agent identity and dealership authority remain inside Quantum’s governed layer rather than depending on one model provider.
A model response is not a completed action. Consequential actions require policy approval, tool execution, destination confirmation, and audit.
Every important system and agent event should leave evidence.
Audit record
- 01Actor
- 02Dealer and rooftop
- 03Buyer and opportunity
- 04Vehicle
- 05Action
- 06Source system
- 07Policy result
- 08Consent state
- 09Ownership state
- 10Tool used
- 11Timestamp
- 12Destination result
- 13Retry or reconciliation
- 14Correlation ID
A customer-facing audit view answers
- Who acted?
- What changed?
- Why was it allowed?
- Which system confirmed it?
- What happened next?
Quantum may redact sensitive content while preserving the evidence required for support, customer review, incident analysis, reconciliation, and applicable compliance obligations.
Detect, contain, communicate, recover, and learn.
- 01
Detect
Receive alerts from system health, security monitoring, customer reports, or operational reconciliation.
- 02
Classify
Determine severity, affected tenants, data, systems, workflows, and customer impact.
- 03
Contain
Pause affected connectors, credentials, communications, or agent authority.
- 04
Investigate
Preserve evidence, identify cause, and determine scope.
- 05
Communicate
Notify internal owners and affected customers according to contract and applicable requirements.
- 06
Recover
Restore safely, rotate credentials, reconcile data, and verify normal operation.
- 07
Improve
Document root cause, corrective actions, ownership, and follow-up testing.
Security contact: security@quantumconnectai.com
A system problem should reduce authority, not create uncontrolled action.
Continuity controls
- 01Encrypted backups
- 02Documented restore process
- 03Environment separation
- 04Connector safe pause
- 05Workflow retry
- 06Dead-letter handling
- 07Provider fallback where approved
- 08Communication shutdown controls
- 09CRM reconciliation
- 10Recovery testing
- 11Customer notification process
- 12Manual operating procedure
Safe-failure behavior
- Stale inventoryHold vehicle-specific claims
- CRM unavailablePreserve events and queue approved writes
- Ownership uncertainBlock buyer-facing outreach
- Communication provider degradedStop duplicate retries
- Model unavailableUse approved fallback or escalate
- Consent unavailableBlock contact
Give researchers and customers a clear path to notify Quantum.
If you believe you identified a security vulnerability affecting Quantum, report it privately with the affected system, reproduction steps, potential impact, and safe contact information.
Disclosure program elements
- Security email
- Expected acknowledgement time
- Testing boundaries
- Prohibited testing
- Counsel-approved safe-harbor language
- Disclosure coordination process
Published today
- Security email: security@quantumconnectai.com
- Reports are acknowledged by a named owner before a formal acknowledgement window is published.
- Formal safe-harbor terms and testing boundaries are published once approved by counsel.
TCPA, Safeguards, GLBA, 10DLC, and state AI disclosure, mapped to the control we run.
Each rule, what it covers, and how Quantum operates against it. Reviewed September 9, 2026.
| Rule | What it covers | What we run |
|---|---|---|
| TCPA | Federal rules on calling and texting consumers, including consent, revocation, and time-of-day limits. Consent practice is currently unsettled after the FCC one-to-one consent rule was vacated in January 2025 and subsequently repealed. | Consent is captured with evidence of when and how it was given, checked before every send, and revocation is applied across channels on receipt. Quiet hours are enforced per store and per state. |
| FTC Safeguards Rule | Mandatory since June 9, 2023 for dealers that extend or arrange credit. Assessors ask about vendor access, access control, MFA, and how a vendor is cut off. | Least-privilege access with per-action policies, no shared administrator logins, credentials held as managed secrets, and the QConnect appliance running outbound-only with a documented revocation path. |
| GLBA | Protection of nonpublic personal information collected in connection with a financial product or service. | Highly restricted data is out of scope for the assistant. Credit, applications, and finance details are routed to a person at the store rather than handled in an automated conversation. |
| 10DLC and A2P registration | Carrier registration for business text messaging in the United States. Unregistered traffic gets filtered or blocked. | Messaging runs on registered brand and campaign records with the dealership named, opt-out keywords honored, and message content aligned to the registered use case. |
| Utah AI Policy Act | Disclosure duties in consumer transactions, effective May 7, 2025. | The assistant discloses that it is an AI assistant. Disclosure wording and version are recorded per store, so you can show what was said and when it changed. |
| California SB 243 | Chatbot disclosure obligations effective January 1, 2026, part of a wave of state chatbot laws. | Same disclosure control, configured per state. Where a state requires specific wording, that wording is set at the store level rather than assumed nationally. |
| Colorado ADMT statute | Colorado repealed and replaced its AI Act in 2026 with a narrower automated decision-making statute effective January 1, 2027. | The assistant makes no credit, pricing, or eligibility decision. Every consequential decision stays with a person at the dealership. |
Summary of how our controls line up with rules dealerships ask about, current as of the review date above. It is not legal advice, and it does not make your dealership compliant on its own. Confirm your obligations with your own counsel.
Trust Center FAQ
Does Quantum train public AI models on dealership data?+
Quantum does not use dealership or customer data for unrelated model training unless the applicable agreement, authorization, disclosure, and provider controls explicitly permit it. The answer in your deployment matches your actual product configuration and contract.
Who owns dealership data?+
The dealership retains its rights in dealership and customer data. Quantum processes data only for documented services, subject to the applicable agreement and law.
Can a dealership revoke Quantum’s access?+
Yes. Every production connection has a documented suspension and revocation process.
How do we review Quantum's security for our vendor process?+
Quantum runs on enterprise cloud infrastructure with independently audited providers, and the controls on this page are implemented today. For your vendor process we provide a written security package covering the controls summary, subprocessor list, data flow description, retention and deletion commitments, and line-by-line answers to your questionnaire.
How does Quantum handle opt-outs?+
Consent and suppression state are evaluated before approved buyer-facing communication. Opt-out events are recorded and written back where the connection supports it.
Where is data stored?+
Cloud regions, storage and backup locations, cross-border processing, and the retention schedule by data category are documented and provided in writing during security review and in the applicable agreement.
How long is data retained?+
Cloud regions, storage and backup locations, cross-border processing, and the retention schedule by data category are documented and provided in writing during security review and in the applicable agreement.
How are Hannah and Coach controlled?+
Each agent has a separate identity, context, tools, permissions, audience, and escalation model. Neither agent decides its own authority.
Related solutions across the platform.
Bring security, IT, legal, and dealership operations into the same conversation.
Quantum can map the proposed systems, data categories, permissions, communication workflows, agent authority, retention, and incident responsibilities before deployment.
Clear access. Clear authority. Clear evidence.