Loading…
Security & trust
Enterprise identity, least-privilege access, tenant isolation and complete traceability — plus AI-specific controls that keep agents inside the lines you draw.
Independently audited security, availability and confidentiality controls.
Certified information security management system.
Safeguards for protected health information.
Data-subject export and erasure tooling built into the platform.
Platform security
AI-specific controls
Traditional controls protect people and data. ZeaIQ adds the controls agents need: what they may do, when a person must approve, and how to stop them instantly.
Written as
“Credits over $100 need approval from a supervisor.”
when tool = issue_credit and args.amount > 100 USD on chat, whatsapp, teams, api then require_approval(role="supervisor")
Held: issue_credit · $240.00
Customer Service Concierge · WhatsApp · waiting 2m
Every tool call is checked by middleware. Block, require approval or allow — a prompt can't override it.
Versions must pass gating eval suites and an approver before they go live.
Suspend any agent or workflow on every surface at once, with a reason shown to callers.
Security filter, PII redaction and a guardrail activity log per agent.
Secrets come from the org vault; browser passwords go into the page, never to the model.
Browser agents treat page content as data, not instructions, and ask before sensitive actions.
ZeaIQ connects to the model providers you choose with your own or platform credentials. Contact us for the data-processing terms that apply to your deployment and providers.
Yes. ZeaIQ can be deployed fully air-gapped with local models through Ollama, so prompts, documents, memories and audit data never leave your environment.
Every agent turn is recorded in the Agent Ops run ledger, and every administrative and runtime action is written to a hash-chained audit log that can be exported and verified.
Yes. Contact our team to request audit reports, policies and our security questionnaire responses.
Get audit reports, policies and questionnaire responses, or set up a session with our security team.
