Introduction
An unauthenticated attacker who can send crafted High Availability (HA) join request packets to a vulnerable FortiMail appliance may be able to execute unauthorized code or commands, according to Fortinet advisory FG-IR-26-200. FortiMail sits directly in the mail path of the organizations that deploy it, so a CVSS 9.8 command injection in its clustering logic deserves prompt attention from email and infrastructure teams, even though Fortinet currently reports no known exploitation.
FortiMail is Fortinet's email security product. It is built to address unwanted bulk email, phishing, ransomware, business email compromise, and targeted attacks, and it integrates with other Fortinet products, third-party components, and cloud email environments. Fortinet, founded in 2000 in the San Francisco Bay Area, announced in September 2026 that Gartner named it a Leader in the 2026 Magic Quadrant for Hybrid Mesh Firewall. That recognition concerns firewalls, and the sources we reviewed do not establish FortiMail-specific market share. Whatever its market position, an email gateway is a sensitive asset wherever it runs, and a flaw reachable without credentials changes its risk profile considerably.
The CVE data we started from lists the vendor as unknown. The advisory and NVD record make clear that the affected vendor is Fortinet.
Technical information
Root cause
NVD classifies CVE-2026-126148 as CWE-78, improper neutralization of special elements used in an OS command. MITRE defines CWE-78 as constructing an OS command using externally influenced input without correctly neutralizing special elements that could modify the intended command. The FortiGuard advisory uses the broader parent category, CWE-77 (command injection). CWE-78 is a child of CWE-77, so the two labels describe the same flaw at different levels of specificity.
Fortinet labels the affected component CLI, but the delivery mechanism is a crafted HA join request packet sent by an unauthenticated attacker. The component label does not mean the attacker must first obtain a CLI login.
Based on the patch analysis available to us, when a node asks to join a FortiMail HA cluster, values from that join request reach an operating system command without proper neutralization. This happens before authentication, so an attacker who can reach the HA channel can run commands with no credentials. The fix therefore sits in the HA join request handling code, not in the web UI or the mail processing path. That also explains the branch scoping: the vulnerable HA code exists only in the 7.6 and 8.0 branches, which is why FortiMail 7.4 and 7.2 are not affected.
What is not public
FortiMail is closed source, and neither the advisory nor NVD publishes:
- The HA join request packet layout or the specific vulnerable field
- The command string, shell, executable, or internal function involved
- The required port or transport protocol
- Whether HA must be enabled, whether cluster membership is required, or whether a particular interface configuration is necessary
- The operating system account or privilege level under which injected commands run
The advisory does not support any claim of root execution. The published details also do not establish either universal Internet exposure or safety for appliances with HA disabled. We have not filled these gaps with assumptions.
Severity scoring
Fortinet, acting as the CVE Numbering Authority, supplies the following base vector. NVD shows its own assessment as not yet provided, so this is the CNA score rather than a separately calculated NIST score.
CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H
That translates to network attack vector, low attack complexity, no privileges required, no user interaction, unchanged scope, and high potential confidentiality, integrity, and availability impact. These values describe the vulnerability itself and do not indicate that any compromise has occurred.
The advisory also publishes temporal metrics:
AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H/E:P/RL:O/RC:R
E:P denotes Proof of Concept exploit maturity. Under the CVSS v3.1 specification, this can mean proof-of-concept code is available, or that an attack demonstration is not practical for most systems and may need substantial modification. No public exploit link or payload accompanies the advisory, so it would be wrong both to say no PoC exists and to say a reliable weaponized exploit is public. RL:O denotes an official fix, consistent with the published upgrades.
Attack flow
Based on the documented details, exploitation proceeds as follows:
- The attacker locates a FortiMail appliance running 7.6.0 through 7.6.6 or 8.0.0 through 8.0.1 whose HA join request handling is reachable from the attacker's network position.
- The attacker sends a crafted HA join request packet containing special elements in one or more request values.
- Before any authentication takes place, the appliance passes those values into an operating system command without neutralizing them.
- The injected content alters the intended command, and the appliance executes unauthorized code or commands.
The exact field, payload syntax, and network prerequisites for step 1 and step 2 are not published.
Patch information
Fortinet fixed CVE-2026-126148 in new FortiMail firmware. There is no standalone hotfix, no configuration change, and no public source diff. Advisory FG-IR-26-200, which NVD links as the authoritative reference, sets these minimum fixed versions:
- FortiMail 8.0: vulnerable 8.0.0 through 8.0.1, fixed in 8.0.2 or above (feature release, build 263)
- FortiMail 7.6: vulnerable 7.6.0 through 7.6.6, fixed in 7.6.7 or above (mature release, build 858)
- FortiMail 7.4: not affected, not applicable
- FortiMail 7.2: not affected, not applicable
The build numbers come from the published 8.0.2 and 7.6.7 release notes. Use them to confirm that an appliance is running the fixed image, rather than relying on a matching version string alone.
What the code change likely involves
Fortinet has published no commit, diff, or patch notes for this code path, so nobody outside Fortinet can verify exactly how it was changed. Fixes for this class of bug usually involve one or more of the following:
- Strictly validating HA request fields against an allowlist
- Passing values to the program as separate arguments instead of building a shell command string
- Rejecting unauthenticated or malformed join requests before parsing them
These are typical approaches, not Fortinet's confirmed implementation.
Release notes do not list this fix
The Common Vulnerabilities and Exposures tables in the 7.6.7 and 8.0.2 release notes list fixes for path traversal (CWE-22), SQL injection (CWE-89), and resource consumption in a loop (CWE-1050), plus CWE-472 in 7.6.7 only and CWE-321 and CWE-522 in 8.0.2 only. Neither table contains a command injection entry (CWE-77 or CWE-78). Fortinet notes that these lists do not cover every bug fixed in a release, so treat the advisory version table, not the release notes bug list, as proof that a build is fixed.
Classification mismatch
The FortiGuard advisory uses CWE-77, while the CVE and NVD record, and Fortinet's CSAF file as reported by Rescana, use CWE-78. Both describe the same fix, so don't let the differing labels split tracking in vulnerability management tooling.
Release timing
The fixed builds shipped weeks before the advisory as part of Fortinet's coordinated security hardening cycle. Fortinet's October 21, 2026 blog listed FortiMail 8.0.2 and 7.6.7 as available that day, and the advisory was first published on November 10, 2026 (NVD lists November 11, 2026). Fortinet also reports that it fixed FortiMail Cloud on October 5, 2026 by moving all tenants to 7.6.7 GA or 8.0.2 GA, so hosted customers need to do nothing.
Fortinet's own blog warns that once updates are public, attackers can compare patched and unpatched builds to find vulnerable code, which it calls patch diffing risk. For an unauthenticated network bug like this one, the interval between disclosure and a working exploit may be short.
No virtual patch
FG-IR-26-200 marks the virtual patch field as No, meaning Fortinet has published no IPS signature for this issue. The advisory also provides no workaround or compensating configuration. Installing the fixed firmware is the only vendor fix.
Affected systems and versions
- FortiMail 8.0.0 through 8.0.1: affected, upgrade to 8.0.2 (build 263) or above
- FortiMail 7.6.0 through 7.6.6: affected, upgrade to 7.6.7 (build 858) or above
- FortiMail 7.4: not affected by this CVE
- FortiMail 7.2: not affected by this CVE
- FortiMail Cloud: remediated by Fortinet on October 5, 2026 (firmware moved to 7.6.7 GA or 8.0.2 GA), no customer action required
The vulnerable entry point is the HA join request handling. The advisory does not specify whether HA must be configured for an appliance to be exploitable, so version, not configuration, should drive the exposure assessment. The unaffected status of 7.4 and 7.2 applies to this CVE only, not to every vulnerability in those branches.
Vendor security history
Fortinet credits Theo Leleu of its Product Security team, who found and reported the vulnerability internally using frontier AI models. The vendor shipped fixed firmware before public disclosure and reports completed cloud remediation ahead of the advisory. Leleu is the discoverer, not a threat actor, and the AI models were used for discovery, not exploitation.
FortiMail also has a recent exploited vulnerability on record. CVE-2026-104286, a path traversal issue, was added to the CISA Known Exploited Vulnerabilities Catalog on October 1, 2026 based on evidence of active exploitation. The Cyber Security Agency of Singapore described it as allowing unauthenticated arbitrary file writes through crafted HTTP or HTTPS requests and reported active exploitation.
That flaw uses a different mechanism, arrives over a different channel, and has its own advisory context. It shows that FortiMail flaws have progressed to real attacks before, but it is not attribution or evidence of exploitation for CVE-2026-126148, which FG-IR-26-200 lists as "Known Exploited: No". The reviewed sources associate no CVE-specific indicators of compromise, actor names, or campaigns with this issue, and the CVE.org record was still shown as RESERVED at the time of review.
References
- NVD: CVE-2026-126148
- CVE Record: CVE-2026-126148
- Fortinet PSIRT Advisory FG-IR-26-200
- Fortinet Blog: Security Hardening Release for Multiple Products
- FortiMail 8.0.2 Release Notes: Introduction and Supported Models
- FortiMail 7.6.7 Release Notes: Introduction and Supported Models
- FortiMail 8.0.2 Release Notes: Resolved Issues
- FortiMail 7.6.7 Release Notes: Resolved Issues
- Rescana: Fortinet November 2026 PSIRT, FortiMail and FortiManager Unauthenticated RCE
- HKCERT: Fortinet Products Multiple Vulnerabilities
- CISA: CISA Adds One Known Exploited Vulnerability to Catalog (October 1, 2026)
- Cyber Security Agency of Singapore: Active Exploitation of Vulnerability in Fortinet FortiMail
- MITRE CWE-78: Improper Neutralization of Special Elements used in an OS Command
- FIRST: CVSS v3.1 Specification Document
- Fortinet: FortiMail Data Sheet
- Fortinet: About Us
- Fortinet: Named a Leader in the 2026 Gartner Magic Quadrant for Hybrid Mesh Firewall



