CyberSip™ — Issue 131 — September 16, 2026

Today's picture

Cisco confirmed active exploitation of CVE-2026-76461, a CVSS 9.8 vulnerability in AsyncOS for Cisco Secure Email Gateway where an unauthenticated remote attacker can achieve root OS command execution by sending a crafted email containing malicious SQL statements through the affected device, with CISA adding it to its Known Exploited Vulnerabilities catalog today. Apple shipped its largest security update in company history on September 14, releasing iOS 27, macOS Golden Gate 27, and eight companion platform updates that together address 273 unique CVEs including 20 kernel fixes in iOS alone, a CoreMedia image-triggered iPhone compromise path, remote code execution routes through CUPS and SMB, and six Gatekeeper bypasses. Japan's Digital Agency confirmed a data breach affecting the Government Solution Service, a shared IT platform providing infrastructure across multiple government ministries, with attackers gaining access to data held in the shared system.

Today's intelligence

3 items

01 CriticalCisco Secure Email GatewayCISA KEV

Cisco Secure Email Gateway CVE-2026-76461: an unauthenticated attacker sends one crafted email and gets root command execution on the gateway — confirmed exploited, on CISA KEV today

The attack vector is the email itself. CVE-2026-76461 is an insufficient validation flaw in the email parsing logic of AsyncOS that lets an attacker embed malicious SQL statements in a message, which the gateway executes against its backend database in a way that escapes to OS command execution with root privileges. No credentials, no user interaction, no network exploit beyond delivering the email. Patch AsyncOS now.

CVECVE-2026-76461
CVSS 9.8

Attack vectorSend a crafted
email through
the gateway
No credentials
No interaction

ResultArbitrary OS
commands as root
on the gateway

KEVAdded today
Sept 16, 2026

FixPatch AsyncOS
for Cisco Secure
Email Gateway

Sources The Hacker News SecurityWeek CISA KEV

What happened

Cisco published an advisory this week confirming active exploitation of CVE-2026-76461 in AsyncOS software for Cisco Secure Email Gateway. The flaw is in the email parsing logic: the gateway processes all inbound email, and a specially crafted message containing malicious SQL statements can cause the backend database handler to execute those statements. The SQL injection then escalates to OS command execution, running arbitrary commands with root privileges on the underlying operating system. The attack requires no authentication and no user interaction beyond the email being processed by the gateway, which happens automatically for every message it receives. CISA added CVE-2026-76461 to its Known Exploited Vulnerabilities catalog today. Cisco's advisory describes it as a case of insufficient validation in the email parsing logic, and specifically states that an attacker exploits this vulnerability by sending a crafted email message that contains malicious SQL statements through an affected device. The Cisco fix is an AsyncOS update. Cisco's cloud-based Email Security through Cisco's own infrastructure has been patched; organizations running Cisco Secure Email Gateway on-premises or on third-party cloud infrastructure must apply the AsyncOS update manually.

Why it matters

Email gateways occupy a uniquely exposed position: they are internet-facing by design and their core function is to process every inbound message without filtering by sender reputation before parsing begins. CVE-2026-76461's attack vector is not a network protocol or an administrative interface — it is the email itself. Any organization running Cisco Secure Email Gateway is processing an uncontrolled stream of inbound traffic from the public internet, and any item in that stream could be a crafted exploit attempt. Root access on the email gateway gives an attacker access to the email content of every message processed, the gateway's own configuration and credentials, and a foothold at the network perimeter for further access.

Potential actions

Apply the AsyncOS update for Cisco Secure Email Gateway immediately to all on-premises and customer-managed cloud deployments. Cisco's hosted cloud gateway has already been updated by Cisco. The CISA KEV addition confirms active exploitation; treat this as an emergency patch with no grace period. Review the Cisco advisory for the specific AsyncOS versions that include the fix.
For any deployment that was running a vulnerable AsyncOS version before the patch, review gateway logs for unexpected outbound connections, new user accounts, or process execution events that would not normally be associated with email processing. Root command execution on the gateway means the attacker can have modified configuration, extracted stored credentials, or established persistence that survives the patch.

The Sip

The attack is the email. Send a crafted message, get root on the gateway that processes every message your organization receives. Confirmed exploited. CISA KEV today. Patch AsyncOS now. If it was running vulnerable before the patch, treat it as a potentially compromised host and review the logs.

02 HighApple iOS 27 / macOS Golden Gate273 CVEs

Apple's largest security update ever: iOS 27 and macOS Golden Gate 27 patch 273 CVEs including 20 iOS kernel fixes, image-triggered RCE, and six Gatekeeper bypasses

Released September 14 across every current Apple platform simultaneously. No CVEs confirmed exploited in the wild — but this is a kernel-heavy release with multiple RCE paths worth prioritizing. CVE-2026-64752 in CoreMedia lets an attacker compromise an iPhone by getting a malicious image in front of a user; Apple removed the flawed code entirely rather than patching it. Nine SMB client flaws include two that can corrupt kernel memory on connection to a malicious server. CUPS carries a remotely exploitable RCE. Update the entire Apple fleet.

Total CVEs273 unique
(1,038 platform
listings before
deduplication)

iOS 27 kernel20 kernel fixes
Memory corruption
Privilege escalation
Info leaks

Notable RCECVE-2026-64752
CoreMedia
Image-triggered
iPhone RCE
(code removed)

Other pathsCUPS remote RCE
SMB: 9 CVEs
6 Gatekeeper
bypasses
WebKit: 6 CVEs

ExploitedNone confirmed
in the wild

Sources CyberSecurityNews Gadget Review IT-Connect

What happened

Apple released iOS 27, iPadOS 27, macOS Golden Gate 27, macOS Tahoe 26.7, macOS Sequoia 15.8, iOS and iPadOS 26.7, watchOS 27, tvOS 27, visionOS 27, Safari 27, and Xcode 27 on September 14, 2026. Across all ten advisories, 273 unique CVEs are addressed. macOS Golden Gate 27 carries the broadest coverage with 210 CVEs. iOS 27 addresses around 126 CVEs with 20 of those in the kernel, covering memory corruption, privilege escalation, and information leak classes. Standout findings include CVE-2026-64752 in CoreMedia, where a malicious image could compromise an iPhone; Apple chose to remove the affected code entirely rather than patching it. Nine CVEs in the macOS SMB client include two where simply connecting to a malicious server can corrupt kernel memory. CUPS, the printing system, has nine flaws including one remote code execution path. Six separate Gatekeeper bypass routes are addressed across autofs, copyfile, the kernel, System Settings, and WebDAV. WebKit's six CVEs include paths that could enable universal cross-site scripting and sensitive data exposure during web content processing. Apple's advisories do not flag any of the 273 as exploited in the wild. Eight vulnerability discoveries in the macOS 27 advisory credit Calif.io "in collaboration with Claude and Anthropic Research," representing one of the first documented cases of Anthropic-assisted AI research credited in Apple's official security notes.

Why it matters

No CVEs confirmed as exploited in the wild reduces the urgency relative to today's Cisco story, but the kernel-heavy nature of this release changes the calculus for enterprise fleets. Twenty iOS kernel fixes in a single release is substantial. Jamf's Adam Boynton framed this accurately to SecurityWeek: the raw CVE count matters less than where the fixes sit, and this is a kernel release rather than a browser release. An unpatched iOS 27 kernel vulnerability is a much more valuable acquisition for an advanced threat actor than an unpatched WebKit flaw, because kernel-level code execution enables the kind of deep device persistence that commercial spyware relies on. Patch the entire Apple fleet now, prioritizing managed devices in sensitive environments.

Don't miss

The eight vulnerability credits to Calif.io "in collaboration with Claude and Anthropic Research" in the macOS Golden Gate 27 notes are worth a brief note. Apple's security credits are one of the few systematic public records of who finds what. Eight credited AI-assisted discoveries in a single Apple advisory, joining eleven from July 2026, suggests that AI-assisted vulnerability research is producing discoveries that clear Apple's bar for inclusion in security advisories at a pace that is accelerating. The same dynamic is referenced in The Hacker News's sidebar today: in H1 2026, Anthropic's Mythos-class models surfaced 26,153 vulnerability candidates in open-source software, of which 421 were patched upstream. The gap between candidates surfaced and patches shipped is the current bottleneck, not the discovery rate.

Potential actions

Update all Apple devices to iOS 27, iPadOS 27, macOS Golden Gate 27 (Apple Silicon Macs), macOS Tahoe 26.7 (Intel Macs), and corresponding watchOS, tvOS, and visionOS releases. With same-day MDM support now standard per Apple, a delayed fleet patching schedule is a policy decision, not a tooling constraint. Prioritize devices handling sensitive data in high-risk environments given the kernel-heavy nature of this release.
Evaluate exposure to the SMB client vulnerabilities specifically if macOS devices connect to external or untrusted SMB shares. Two of the nine SMB CVEs can corrupt kernel memory on connection to a malicious server without further user interaction. Restrict macOS SMB connections to trusted internal shares as an interim control on devices that cannot immediately update.

The Sip

273 CVEs. 20 iOS kernel fixes. A CoreMedia image-triggered iPhone RCE that Apple patched by deleting the code entirely. Nine SMB CVEs, two of which corrupt kernel memory on connection. None exploited in the wild yet. This is a kernel release — prioritize the fleet update and pay attention to SMB exposure.

03 HighJapan GSSGovernment Breach

Japan's Digital Agency confirmed a breach of the Government Solution Service, the shared IT platform used across multiple ministries

The Government Solution Service is Japan's centralized shared IT infrastructure platform serving multiple government ministries and agencies. A breach of a shared government platform affects every tenant on the platform, not only a single agency. Japan's Digital Agency confirmed attackers gained access to data held in the GSS. Full scope, affected ministries, and attacker attribution have not been confirmed publicly as of today.

PlatformGovernment Solution
Service (GSS)
Japan shared IT
infrastructure

Confirmed byJapan Digital
Agency

ScopeMultiple ministries
and agencies
as GSS tenants

AttributionNot confirmed

Data affectedGovernment data
held in shared
infrastructure
Full scope TBD

Sources CyberSecurityNews SecurityWeek

What happened

Japan's Digital Agency confirmed a significant data breach affecting the Government Solution Service, a centralized IT infrastructure platform that provides shared computing services to multiple Japanese government ministries and agencies. The GSS was established as part of Japan's digital government initiative to consolidate and standardize IT infrastructure across the public sector. The Digital Agency's confirmation states that attackers gained access to data held within the shared platform. The specific ministries and agencies affected, the volume and type of data accessed, the attack vector, and any attribution have not been publicly disclosed as of today. The breach of a shared multi-tenant government platform is structurally similar in scope to the incidents this brief has covered involving managed service provider platforms like N-central (Issue 127): a single platform compromise is access to every tenant's data and systems on that platform.

Why it matters

Shared government IT platforms represent a concentration of sensitive administrative data across many agencies in a single exploitable target. Japan has been a consistent target of China-aligned espionage operations; in August 2024, the Japanese government attributed a multi-year campaign to MirrorFace, a China-linked group, that targeted government ministries, defense contractors, and research institutions. A successful breach of the GSS, if the scope extends beyond a single agency's data, would represent access to data from across the Japanese government's shared infrastructure in a single incident. The Digital Agency's disclosure is early-stage; the operational picture will become clearer as Japan's incident investigation progresses.

Potential actions

Organizations with formal or informal data-sharing relationships with Japanese government agencies should assess whether data they have shared through GSS-connected channels could be in scope for the breach. Monitor the Japan Digital Agency's public disclosures for additional detail on which agencies and data categories were affected as the investigation progresses.
Security teams tracking China-aligned or North Korea-linked espionage operations against Japan should note the GSS breach as a potential intelligence collection event and assess whether operational data for current engagements may have been exposed. Attribution is not yet confirmed; the scope of compromised data, when disclosed, will inform the likely threat actor profile.

The Sip

Japan's shared government IT platform, multiple ministries as tenants, confirmed breach. One platform, many agencies' data. Attribution and full scope not yet public. Watch Japan's Digital Agency disclosures as the investigation proceeds.

Cross-source standouts

Email as an attack surface: when the protocol itself is the exploit delivery mechanism

CVE-2026-76461 in Cisco Secure Email Gateway belongs to a class of vulnerabilities where the attack vector cannot be meaningfully blocked by network controls because the malicious payload is delivered through the channel the device exists to process. Email gateways receive and parse email from the public internet as their core function. Any attempt to block email before parsing is self-defeating. The SQL injection that achieves root execution is carried inside a message that looks like ordinary email to every network control that operates above the application layer. The only defenses are patching the parsing logic and treating any pre-patch deployment as potentially compromised. This is the same structural constraint that makes web application vulnerabilities difficult to block at the network perimeter: if the service must process arbitrary input from the public internet, a vulnerability in how it processes that input cannot be mitigated by controlling what reaches it.

AI-assisted vulnerability research in Apple's credits: what the acceleration of discovery means for the gap between finding and fixing

Apple's security notes for macOS Golden Gate 27 credit eight vulnerabilities to Calif.io "in collaboration with Claude and Anthropic Research." Eleven similarly credited discoveries appeared in July 2026's Apple update. The Hacker News reported today that Anthropic's Mythos-class models surfaced 26,153 vulnerability candidates in open-source software in H1 2026, with 421 patched upstream. The ratio is the story: 26,153 surfaced, 421 fixed. AI-assisted discovery is scaling the rate at which potential vulnerabilities are identified faster than the rate at which maintainers can review, validate, and patch them. The downstream security implication is not immediately obvious. More discovered vulnerabilities that are responsibly disclosed and patched is a better outcome than the same vulnerabilities sitting undiscovered. But the gap between surfacing and patching is also a gap during which a sophisticated attacker with the same AI-assisted discovery capability, or with access to the disclosed-but-unpatched candidates, could exploit them ahead of the fix. The bottleneck is not discovery speed anymore. It is the human and organizational capacity to triage, validate, and ship fixes for the volume AI discovery is producing.

Still watching

Days 4–7+

GitLab CVE-2026-85706 (Issue 129 · max-severity path traversal, mass exploitation confirmed) — patch immediately. Hunt POST to /api/v4/projects/{id}/repository/commits/ with file.path parameter. CVE-2026-19478 active simultaneously; verify patch status for both independently.

Day 4

ShieldCrash (ShieldBreak bypass) (Issue 126 · Nightmare Eclipse, no CVE, SYSTEM file read on latest patched Windows) — ShieldBreak CVE-2026-69414 patched. ShieldCrash unpatched. Monitor MSRC. Behavioral detection for the series must remain active.

Day 7+

SAP OVERPASS CVE-2026-44756 (Issue 126 · CVSS 10.0, unauthenticated SAP kernel EPP RCE) — apply patch per SAP Note 3500180. Restrict EPP endpoint to trusted networks. Still no confirmed exploitation; prior CVSS 10.0 SAP flaws moved to exploitation quickly.

Day 7+

StyleSmuggler CVE-2026-75650 (Issue 125 · CVSS 10.0, Magento/Adobe Commerce, actively exploited since September 4) — apply composer patch VULN-39341. Rotate encryption key and all derived credentials. Run eComscan for the Linux backdoor at the OS level.

Day 7+