An enterprise email security platform with native detection in English, Arabic, Urdu, Roman-Urdu, Hindi and Sindhi, and Gulf/Pakistan/India regional threat intelligence. A multi-engine AI pipeline — rules, machine learning, regional-language models, malware delivery-vector analysis, network-layer monitoring, and QR Shield — reaches a single, explainable risk decision. Built for governments, banks, and critical infrastructure. Deploys on-prem or air-gapped. Your data never leaves your network.
Every inbound email hits the MailGuard AI SMTP edge and is simultaneously scored by several independent detection engines — rules-based analysis, machine learning, regional-language models, malware delivery-vector analysis, and threat intelligence. The ensemble combines their outputs into a single 0–100 risk score. Score ≥85: auto-blocked. Score 50–84: quarantined. Below 50: delivered. Every decision is logged with a full forensic trail, SMTP headers, CVE mappings, and a plain-language AI explanation your SOC team can actually read.
This is not a signature-matching product. It uses pre-trained, continuously-updated models rather than requiring you to train anything yourself — so there is no lengthy tuning period before it is useful. The regional-language engines know Gulf, Pakistani and Indian scam patterns your existing SEG is completely blind to.
Each engine contributes a specialist signal. The ensemble stops threats that any single engine would miss — especially attacks with no prior signature.
Domain age, TLS validity, registrar reputation, redirect chains, look-alike and punycode domain detection, URL shortener unwrapping.
Pre-trained models detect BEC urgency patterns, executive impersonation signals, and phishing lures across English, Arabic, Urdu, Roman-Urdu, Hindi and Sindhi.
Detects urgency, authority, fear, and reward-trigger patterns across every language the platform supports, plus US-market-specific scam patterns (IRS, SSA, delivery, gift-card BEC).
Public CVEs mapped to email delivery vectors, auto-synced with the CISA Known Exploited Vulnerabilities catalog. CVSS-weighted scoring flags emails exploiting actively-exploited CVEs.
Native Arabic, Urdu, Roman-Urdu, Hindi and Sindhi lure detection — GCC and South Asian brand impersonation, government-authority impersonation, and mobile-money/UPI scam patterns. No other commercial platform we are aware of covers Sindhi at all.
Every image attachment decoded: OCR → QR decode → URL extraction → ML scoring on the extracted link. Catches QR-code phishing that bypasses standard URL filters entirely.
When several engines independently flag the same email, the combined confidence rises — catching attacks that each individual engine alone would rate as only borderline, including ones with clean SPF/DKIM/DMARC that a standard SEG would let straight through.
Autopilot lets MailGuard AI act immediately on very-high-confidence threats, instead of waiting on an analyst — but only for actions that are always safe to reverse. It quarantines a message rather than deleting it, and creates a removable block policy rather than a silent, untraceable change.
Most global email security products are English-first. An attacker who sends «تحويل عاجل» (urgent wire transfer) in Arabic, or an equivalent lure in Urdu, Roman-Urdu, Hindi or Sindhi, bypasses standard English-trained detection entirely. MailGuard AI is built with purpose-made detection for each of these languages — not a translation layer bolted onto an English model.
MailGuard AI embeds the complete NIST NVD + CISA KEV database — no internet required. Every email attachment and link is scored against 300+ CVEs covering 2010–2026, with 54 active CISA Known Exploited Vulnerabilities triggering auto-block regardless of other scores.
Executive impersonation, wire transfer lures, gift-card BEC, writing-style baseline analysis. Detects when an email doesn't match the sender's established communication patterns.
Real-time detectionOCR → QR decode → URL ML scoring on every image attachment. QR phishing routinely bypasses standard URL filters entirely — QR Shield closes that gap.
Every image attachment scanned34 magic byte signatures. VBA macro scan. CVE mapping: RTF→CVE-2017-11882, LNK→CVE-2017-8464, RAR→CVE-2023-38831. Archive bomb detection.
EXE, LNK, RTF, DOCX, PDFCyrillic homoglyphs, Turkish dotless-i tricks, zero-width characters, and right-to-left override characters — the same Unicode manipulation techniques used to hide prompt-injection attacks targeting AI assistants. Catches microsofft.com before it reaches the inbox.
Every message scannedSSNs, IBANs, credit card numbers (Luhn-validated), health-record patterns, and passport numbers — detected in email body and attachments before leaving the network.
Every outbound message scannedWriting-style baseline per executive. Impersonation attempt detection. Wire-fraud correlation. Look-alike domain monitoring for every C-suite member you configure.
Configurable per executiveEmail arrives at the MailGuard AI SMTP edge. TLS enforced. SMTP smuggling blocked (CVE-2023-51764).
SPF, DKIM, DMARC evaluated. Failures feed the ensemble; no auth alone doesn't trigger block.
All detection engines run in parallel. Attachments decoded, QR codes scanned, URLs extracted.
Weighted combination across engines produces a single 0–100 risk score.
≥85: auto-block. 50–84: quarantine. <50: deliver. All decisions logged with full forensic trail.
AI explanation in plain language. SMTP headers, CVE mappings, risk breakdown. SIEM/webhook export.
MailGuard AI's own controls are mapped against the frameworks your procurement and audit teams will ask about, to support your own compliance work. This is evidence mapping to help your audit, not a certification held by Cyber Zeus Global — ask us for our current certification status.
Saudi Arabian Monetary Authority Cybersecurity Framework. Relevant for Saudi financial institutions.
Saudi National Cybersecurity Authority Essential Cybersecurity Controls.
UAE Information Assurance Regulation. Relevant for UAE critical infrastructure and government.
SOC 2 is in preparation, not yet certified. Ask us for our current status.
PHI-pattern detection in the outbound DLP engine, for organisations that handle health data.
Evidence-mapped globally. Ask us for our current certification status.
For organisations handling EU personal data. On-premises/air-gapped deployment keeps data under your own control.
This page maps MailGuard AI's controls against these frameworks to support your own compliance and audit work. It is not a claim that Cyber Zeus Global holds any of these certifications today — ask us directly for current status.
MailGuard AI ships as a Docker Compose stack. Five minutes from zero to scanning. Air-gapped deployments ship with pre-trained models and the full CVE database embedded — no internet connection ever required. Email data never leaves your network.
Fully isolated networks. No outbound connection. Pre-trained models. Embedded CVE database. Required for classified environments.
Runs in your own datacentre. Linux x86/ARM. Docker Compose or bare metal. 8 services, single compose command.
AWS, Azure, or GCP single-tenant deployment. Data residency enforced per region. No shared infrastructure.
Transparent gateway mode. Point your MX records at MailGuard AI — nothing changes for end users.
Real-time webhook to Splunk, Microsoft Sentinel, QRadar, Cortex XSOAR, PagerDuty, Slack, Teams.
Full white-label. Per-tenant branding, billing API, MSSP console. One platform, your brand, your clients.
loss-avoided estimate on your own report, based on threats actually stopped and your own configured incident costs — not a generic industry multiple.
pilot against your own email traffic — see exactly what your current gateway is missing, before you commit.
to deploy. Docker Compose, any Linux server, pre-trained models shipped in the package — no lengthy tuning period.
detection thresholds, so you can balance catch-rate against false positives for your own environment and traffic pattern.
Actual results depend on your environment and traffic — request a live pilot to see real numbers on your own mail, not an industry-average estimate.
MailGuard AI natively detects Arabic, Urdu, Roman-Urdu, Hindi and Sindhi attacks alongside English — and is evidence-mapped against the compliance frameworks that matter across the Gulf, Pakistan, India and North America. Same codebase, same detection depth, wherever you deploy it.
Request a live demo. We will run MailGuard AI against your own email traffic and show you exactly what is getting through — regional-language BEC, QR phishing, CVE-linked exploits, and attacks your existing gateway never flags.
Request a DemoMailguard AI is designed to deliver value on day one for a small team, and to scale to a governed, multi-site enterprise deployment on the same platform.
Illustrative walkthroughs of how Mailguard AI is used and the value it creates. Your figures are set on your own data during a proof-of-value engagement.
A scoped engagement on your own data shows the results before any wide rollout — measured, not promised.
Start with one team or site, then expand across the organization on the same platform — no re-buy, no re-build.
Every output is explainable and audit-ready, so risk and compliance teams can stand behind it.
A scoped engagement shows the results measured on your own operation before any wide rollout — then scale across the organization on the same platform.
Book a proof of value
