TRUST

Intelligence should work under your authority.

TAWALI is being built around explicit permissions, bounded access, human approval for sensitive actions, a security decision before execution, fail-closed behavior, and evidence that can be verified later. Readiness here is stated precisely, not with badges.

Ask a security question Explore governed execution

TAWALI AI

Intellectual Property & Corporate Trust

Saudi-registered Establishment

TAWALI is developed and operated by a Saudi-registered establishment under Unified National Number 7053042599.

Registered Software Work

TAWALI AI is registered in Saudi Arabia as a computer software and applications work with the Saudi Authority for Intellectual Property (SAIP).

Registration No. 26-12-106719360Registered 16 September 2026

Official Corporate Presence

TAWALI AI’s official LinkedIn Company Page.

Follow TAWALI AI on LinkedIn (opens in a new tab)

Official Domain

The official website and corporate contact for TAWALI AI.

tawali-ai.com support@tawali-ai.com

Registered work title: TAWALI AI – منصة توالي للذكاء الاصطناعي
Certificate type: Work registration certificate
Registration authority: Saudi Authority for Intellectual Property (SAIP)
Category: Computer Software and Applications

Registration documents the software work and its intellectual-property record. It does not constitute product, security, or regulatory certification.

CORE AUTHORITY MODEL

Connecting an account is not unlimited access

UnderstandSuggestRequest approvalAct within limits

Connecting an account does not grant unlimited authority.

Reading is not sending. Monitoring is not account management.

Drafting is not publishing.

Permissions must be explicit, and access is least-privileged.

Sensitive execution requires appropriate authority.

Access is bounded by limits the customer defines.

COMMUNICATION AUTHORITY

Every channel, its own authority level

Communication and social messaging follow the same canonical governance sequence above. Reading is not sending, and drafting is not publishing — each stage requires explicit permission to move past it. An example for a message-handling task:

Understand

Reading and summarizing an inbound message or mention.

Suggest

Organizing the conversation and preparing a draft reply.

Request approval

Nothing is sent or published until approval is explicitly requested.

Act within limits

Sending or publishing happens only within the limits you approved.

OPERATIONAL PROTECTIONS

Review, revocation, and safe behavior

Review and approval

Actions of consequence are designed to be reviewable, with approval requested before they take effect.

Revocable access

Every connection can be reviewed and revoked. Access never becomes permanent by default.

Safe failure and expiry

Expired or failed connections default to the safest behavior, never to broader access.

DATA & CONNECTION PRINCIPLES

Least privilege, most visibility for the customer

Access follows explicit, customer-defined permissions.

Connections are scoped to what a task actually needs.

Connections remain visible and manageable by the customer.

Internal security architecture is not exposed publicly.

GOVERNANCE PRINCIPLES

Capability ≠ Authority ≠ Readiness

What an AI model can do, what this user, workspace, and connected account are permitted to do right now, and whether the credentials and controls are safe at this moment are three different questions. TAWALI keeps them apart and checks all three inside the execution path itself.

Least privilege

Connecting an account grants only the scope a task needs. Reading is not sending, and drafting is not publishing.

Human control

Sensitive governed actions require human approval where required, and that approval is bound to the exact draft, recipient, and operation.

Fail-closed behavior

When authority, credentials, or a required control is missing or uncertain, the action stops rather than proceeding with broader access.

SENTRA before execution

SENTRA evaluates risk before a governed action runs and signals proceed, review, or stop. Its decision is not execution authority.

Fresh authority and duplicate protection

Authority is checked again at execution time. Execution is designed to prevent duplicate execution through idempotency and to surface uncertain outcomes rather than retry blindly.

Evidence before claims

Outcomes are recorded with honest states: verified, attention, not observed, or not applicable. Absence is shown as absence, never as success.

CURRENT STATUS

What is verified, in which environment, and what that does not mean

These states are static and reviewed from TAWALI's public claim register; nothing here is read from a live system. Each item shows when it was last reviewed.

VerifiedObserved working in the named environment under controlled conditions. Not a statement of public or Production availability.
BuiltImplemented and tested within its defined boundary, not necessarily activated for customers.
Not availableNot offered today. Merged code does not change this.
Default-offCode and controls exist but are disabled and not customer-facing.

Email intelligence: reading, prioritizing, extracting tasks, dates, and risks, and drafting for your review
Verified
EnvironmentStaging, under controlled conditions
What this does not meanNot available to customers in Production, and not an open Early Access offer.
Last reviewed: 2026-09-10
One governed real email send through the governed execution path
Verified
EnvironmentStaging, one controlled experiment
What this does not meanGeneral sending for customers is not activated anywhere.
Last reviewed: 2026-09-10
One governed scheduled send through the full control plane
Verified
EnvironmentStaging, one authorized run, all grants revoked afterwards
What this does not meanNo continuous scheduler is active. Scheduling is not a customer feature today.
Last reviewed: 2026-09-10
SENTRA risk decisions (proceed, review, stop) evaluated before governed execution
Verified
EnvironmentStaging, observed in controlled execution
What this does not meanA SENTRA decision is not execution authority, and SENTRA is not sold as a separate product today.
Last reviewed: 2026-09-10
Approval binding, fresh authority at execution time, governed execution, duplicate protection, and evidence records
Built
EnvironmentRepository and CI, exercised on Staging
What this does not meanDoes not mean Production activation. Human approval applies to sensitive governed actions where required, not as a blanket step on each action.
Last reviewed: 2026-09-10
Files and untrusted-content security: database authority foundation
Built
What this does not meanThe file-scanning runtime is not live in any environment.
Last reviewed: 2026-09-10
Customer file upload
Not available
What this does not meanNo public or customer upload path exists. Untrusted intake stays disabled and fails closed.
Last reviewed: 2026-09-10
Voice: governed voice intake foundations
Default-off
What this does not meanNo customer voice experience is live.
Last reviewed: 2026-09-10
Public Early Access
Not available
What this does not meanAccess is by consideration only. There is no open self-service offer.
Last reviewed: 2026-09-10
Production activation of governed capabilities
Not available
What this does not meanMerged code and Staging verification never imply Production activation.
Last reviewed: 2026-09-10

SECURITY DISCLOSURES

What we do not claim

TAWALI does not claim that every attack or every AI error can be prevented. Governance reduces and surfaces risk; it does not eliminate it.

TAWALI holds no third-party security certification today. None is claimed on this site.

Verification on Staging never implies Production activation or public availability. Public Early Access is not open.

Internal security architecture details, reason codes, and thresholds are not published.

SECURITY MATURITY

Security maturity expands alongside platform capability

Security maturity is expanding alongside platform capability, with controls introduced, tested, and validated in stages. Policy management, audit exports, risk tiers, and compliance readiness remain planned enterprise controls, not capabilities offered today.