A poisoned .AI file that pops a designer's laptop when they double-click it
CVE-2026-48334 is an improper input validation flaw (CWE-20) in Adobe Illustrator Desktop 2026 file parsing. An attacker crafts a malicious Illustrator document (likely .AI, .EPS, .PDF, or .SVG-derived) that, when opened by a user, executes arbitrary code with that user's privileges. Adobe's advisory APSB26-79 covers Illustrator 2026 (28.x) and Illustrator 2025 on Windows and macOS; fixed versions ship via Creative Cloud desktop. Adobe reports no known exploitation and credits researcher Kieran (kaiksi).
Vendor scored this 9.3 CRITICAL with AV:N/AC:L/PR:N/UI:R/S:C. The AV:N and S:C (scope change) inflate the number, but the reality is a UI:R client-side file-parse bug in a niche creative tool. This is not remote-server-reachable; it requires a human at a workstation to open an attacker-supplied file. That is squarely medium-severity turf for enterprises. The score is defensible under CVSS mechanics but misleading as a patch-priority signal.
5 steps from start to impact.
Craft a malicious Illustrator file
- Access to patched and unpatched Illustrator binaries
- Reverse-engineering capability
- No public PoC 24 hours after disclosure
- Adobe file formats are complex and undocumented — high dev cost
Deliver the file to a designer
- Target uses Illustrator
- Attacker knows a designer's email or a shared asset repo
- Email gateway attachment filtering (.AI is uncommon MIME, often quarantined)
- MOTW / SmartScreen prompt on Windows
Victim opens the file
UI:R gate — no auto-preview vector is known for Illustrator.- Victim opens the file in vulnerable Illustrator version
- MOTW bypass or user clicks through warning
- Trained users don't open unsolicited design files
- Preview handlers don't invoke full Illustrator parser
Code execution as the logged-in user
- Successful memory corruption chain (ASLR/DEP/CFG bypass baked into exploit)
- Windows CFG and Arbitrary Code Guard on Illustrator process
- EDR memory-scan and behavior heuristics
- Application-control (AppLocker/WDAC) blocks unsigned child processes
Post-exploitation on a creative workstation
- Successful implant persistence
- Least-privilege on user accounts
- Network segmentation between creative VLAN and prod
The supporting signals.
| In-the-wild exploitation | None reported. Adobe explicitly states no known exploits at disclosure. |
|---|---|
| Public PoC | None public as of 2026-07-15. Researcher Kieran (kaiksi) has not released details. |
| EPSS | 0.00392 (~0.4%) — bottom quartile of exploit-likelihood forecasting. |
| CISA KEV | Not listed. |
| CVSS 3.1 vector | AV:N/AC:L/PR:N/UI:R/S:C/C:H/I:H/A:N = 9.3 CRITICAL. UI:R and no A:H — the S:C is what pushes it over 9.0. |
| Affected versions | Illustrator 2026 (28.x) and Illustrator 2025 (29.x) on Windows and macOS per APSB26-79. |
| Fixed versions | Latest release via Creative Cloud desktop updater (APSB26-79, 2026-07-14). Refer to bulletin for exact build numbers. |
| Scanner/exposure telemetry | Not internet-reachable — Illustrator is a fat-client desktop app. Shodan/Censys/GreyNoise are N/A for this class. |
| Disclosure | 2026-07-14 as part of Adobe Patch Tuesday (APSB26-79 Illustrator bundle). |
| Credit | Reported to Adobe by researcher Kieran (kaiksi). |
noisgate verdict.
Downgraded from CRITICAL to MEDIUM because the single most decisive factor is that exploitation requires a human to open an attacker-supplied file in a desktop-only creative app — there is no network-reachable service, no auto-preview vector, and blast radius is bounded by one non-privileged designer account. Role-multiplier analysis produces no high-value-role deployment: Illustrator does not run on domain controllers, hypervisors, or CI/CD infrastructure.
Why this verdict
- UI:R is a hard gate. No network path, no listener, no auto-preview — a user must double-click a hostile file. That knocks the exploitable population from
all internet-reachable Illustrator installs(which is zero — it's desktop-only) tousers who open untrusted design files. - Deployment population is narrow. Illustrator is on designer/marketing workstations, not on the general fleet. In a 10,000-host estate, expect 2–8% installed base concentrated in creative teams.
- Role multiplier: low-value role only. Illustrator ships exclusively on end-user creative workstations — there is no server, DC, hypervisor, IdP, or CI role for it. Chain ends at one non-admin user's session; no fleet, tenant, or supply-chain amplification is architecturally available. Floor stays at MEDIUM.
- EPSS 0.00392 and no KEV — telemetry consensus is that exploitation is unlikely in the next 30 days.
- No public PoC and no ITW activity. Adobe file-format bugs are historically slow to weaponize outside of nation-state and commercial-exploit-broker use.
- Impact ceiling is user-context RCE (A:N). No SYSTEM escalation is provided by the bug itself; attacker inherits the designer's privileges only.
Why not higher?
HIGH would require either (a) active exploitation / KEV listing, (b) a public weaponized PoC, or (c) a high-value-role deployment such as a server, headless render farm, or admin workstation. None of these apply — Illustrator is a UI-driven desktop app for creative users, disclosure is 24 hours old, and no PoC exists.
Why not lower?
LOW would require the bug to be non-exploitable in practice or to have a trivial workaround. It's a real memory-safety-adjacent RCE with a plausible phishing delivery path against a well-defined user population (designers who open unsolicited files daily). MEDIUM correctly reflects a real-but-narrow risk that belongs on the 365-day remediation clock, not the 3-day critical queue.
What to do — in priority order.
- Block Illustrator file types at the email gateway from external senders — Add
.ai,.eps, and Illustrator-flavored PDFs to attachment quarantine rules on Proofpoint/Mimecast/Defender for O365 for messages originating outside the org or from unauthenticated senders. Since this is a MEDIUM verdict there is no noisgate mitigation SLA — treat as best-practice hardening but implement within the standard change window. - Enforce Mark-of-the-Web and SmartScreen on creative workstations — Ensure Attachment Manager and MOTW are enabled via GPO so downloaded .AI files trigger the untrusted-file warning. Verify via
Get-ItemProperty -Path 'Zone.Identifier'on a downloaded sample. Roll out within your normal endpoint-config cycle. - Deploy EDR behavioral rule: Illustrator.exe spawning script/shell interpreters — Alert on
Illustrator.exe -> cmd.exe|powershell.exe|wscript.exe|rundll32.exe|mshta.exe. This is a high-fidelity RCE signal across most Adobe file-parse CVEs — deploy once, benefit for years. Ship the rule within 30 days. - Roll the Creative Cloud update through your standard workstation patch ring — Push the APSB26-79 update via Adobe Admin Console remote update feature or your endpoint management (Intune/Jamf/SCCM). MEDIUM verdict fits within the noisgate remediation SLA of ≤ 365 days; realistically fold into the next monthly Creative Cloud maintenance ring.
- Perimeter WAF / IDS — Illustrator is a desktop parser, not a network service. WAF rules are irrelevant.
- Network segmentation of creative VLAN — helpful for lateral movement after compromise but does nothing to prevent the initial file-open.
- Disabling macros — Illustrator does not use Office macros; this is a native file-format parser bug.
- MFA on Adobe accounts — the exploit runs in the local Illustrator process; Adobe SSO status is unrelated.
Crowdsourced verification payload.
Run on each Windows creative workstation (or via your endpoint tool at scale). Requires local read access to Program Files. Invoke as: .\Check-CVE-2026-48334.ps1. For macOS fleets, adapt to read /Applications/Adobe Illustrator 2026/Adobe Illustrator.app/Contents/Info.plist with defaults read.
#requires -Version 5.1
# Check-CVE-2026-48334.ps1
# Detects vulnerable Adobe Illustrator installs per APSB26-79.
# Exit codes: 0 = PATCHED, 1 = VULNERABLE, 2 = UNKNOWN / not installed
$ErrorActionPreference = 'SilentlyContinue'
# Fixed builds per APSB26-79 (verify against vendor bulletin for exact minor)
$fixedBuilds = @{
'2026' = [Version]'28.7.1' # Illustrator 2026 fixed build family
'2025' = [Version]'29.9.1' # Illustrator 2025 fixed build family
}
$paths = @(
'C:\Program Files\Adobe\Adobe Illustrator 2026\Support Files\Contents\Windows\Illustrator.exe',
'C:\Program Files\Adobe\Adobe Illustrator 2025\Support Files\Contents\Windows\Illustrator.exe'
)
$found = $false
$vulnerable = $false
foreach ($p in $paths) {
if (Test-Path $p) {
$found = $true
$ver = (Get-Item $p).VersionInfo.FileVersion
try { $v = [Version]($ver -replace '[^0-9\.].*$','') } catch { $v = $null }
$family = if ($p -match '2026') { '2026' } elseif ($p -match '2025') { '2025' } else { $null }
$fixed = $fixedBuilds[$family]
Write-Host "Found Illustrator $family at $p (version $ver)"
if ($v -and $fixed -and $v -lt $fixed) {
Write-Host " -> VULNERABLE (fixed build: $fixed)" -ForegroundColor Red
$vulnerable = $true
} elseif ($v -and $fixed) {
Write-Host " -> PATCHED" -ForegroundColor Green
} else {
Write-Host " -> UNKNOWN version parse" -ForegroundColor Yellow
}
}
}
if (-not $found) {
Write-Host 'UNKNOWN: Adobe Illustrator not installed in default locations'
exit 2
}
if ($vulnerable) {
Write-Host 'VULNERABLE'
exit 1
} else {
Write-Host 'PATCHED'
exit 0
}
If you remember one thing.
Sources
- Adobe Security Bulletin APSB26-79 (Illustrator)
- OffSeq Threat Radar — CVE-2026-48334
- CIS Advisory — Multiple Adobe Vulnerabilities (July 2026)
- Adobe Security Bulletins index
- CVE Details — Adobe Illustrator vulnerabilities
- The Hacker News — Adobe July 2026 patches
- SecurityWeek — Adobe patches critical vulnerabilities
- CWE-20: Improper Input Validation
What defenders are saying.
Crowdsourced verification outputs.
Results submitted by users who ran the verification payload against their environment.