Skip to main content

TaskingAI CVE-2026-51906

| EUVDEUVD-2026-91701 CRITICAL
Path Traversal (CWE-22)
2026-10-02 mitre GHSA-w3f3-w8fw-rqq5
9.1
CVSS 3.1 · Vendor: mitre
View

Attack ChainAIDerived

Hypothetical attack flow derived from CVE metadata – not a confirmed exploit chain

  1. Access
    Access TaskingAI tool/assistant API
  2. Delivery
    Invoke DALL-E 3 image generation
  3. Exploit
    Inject traversal sequence in project_id
  4. Execution
    save_url_image writes image outside target dir
  5. Impact
    Overwrite file or drop payload for code execution

Vulnerability Assessment

Weakness
CWE
OWASP Top 10:2025
A01:2025 - Broken Access Control
ASVS 5.0
V5 - File Handling Path canonicalization, traversal prevention
MITRE ATT&CK

2 techniques across 2 tactics: Discovery, Collection

The technique list with ATT&CK links and the kill-chain view is part of the Basic plan and up.

Exploitation Conditions

Exploitation requires the DALL-E 3 image generation tool to be enabled and configured (with valid image-generation API credentials) within TaskingAI v0.3.0, and the attacker must be able to invoke that tool and control the project_id parameter passed to save_url_image. The attacker supplies traversal sequences in project_id to redirect the written image outside the intended project directory. Limiting factors: the write content is an image file whose bytes are only partially attacker-controlled, so escalation to code execution depends on a writable web root or overwriting an executable/config/startup path; the DALL-E 3 tool must be active; and the issue is pinned to v0.3.0. Despite the vector's PR:N, invoking the tool pipeline realistically implies application-level access rather than fully anonymous exploitation - verify with vendor.

Impact

Signals conflict, and the gap matters. The input CVSS is 9.1 Critical with AV:N/AC:L/PR:N/UI:N/C:H/I:H, which paints a remote, unauthenticated, high-confidentiality-and-integrity picture. But the description is a path-traversal (CWE-22) arbitrary file WRITE in the DALL-E 3 save_url_image routine - a write primitive, not a read primitive, so the C:H rating is hard to justify and appears inflated (verify confidentiality impact with vendor). PR:N is also questionable: the DALL-E 3 tool must be invoked through TaskingAI's assistant/tool-calling feature with a configured image-generation API key, which in most deployments implies some level of application access rather than fully anonymous reach. Exploitation-probability signals all point LOW: EPSS is 0.14% (3rd percentile), there is no CISA KEV listing, and no explicit POC flag is set in the input (though a referenced researcher gist may document the technique). Net assessment: this is a genuine, realistically exploitable arbitrary-file-write bug that can escalate to code execution if images land in a web root or executable/config path, but it is NOT a mass-exploitation emergency - it affects a niche, emerging product, has no evidence of in-the-wild use, and the headline 9.1 overstates real-world risk. Prioritize patching for anyone running TaskingAI v0.3.0 with the DALL-E 3 tool enabled; treat it as medium-high operationally rather than critical.

Exploit ScenarioDerived

An attacker able to invoke TaskingAI's DALL-E 3 image generation tool calls save_url_image with a project_id containing directory-traversal sequences (e.g. ../../), causing the downloaded image to be written outside the intended project directory to an arbitrary server path such as the web root or a startup/config location. By overwriting a sensitive file or dropping a crafted payload into an executable path, the attacker can corrupt data or potentially achieve code execution. A gist referenced in the advisory appears to document the technique, though a formal POC is not confirmed in the input.

Act Now Prioritize within 24-48h. Why: CRITICAL severity.
No patch listed yet 0.1% exploitation probability (EPSS)

Get an e-mail when this CVE is exploited, patched or re-scored. Free account.

Alert me

AnalysisAIv1 – Oct 06, 2026

Path traversal in TaskingAI v0.3.0's DALL-E 3 image-generation tool lets an attacker who can invoke that tool write downloaded images to arbitrary locations on the server filesystem by injecting traversal sequences into the project_id parameter passed to save_url_image (CWE-22). The independent assessment scores this as an integrity-only file-write issue (CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:N/I:H/A:N), which is materially more conservative than the CNA's 9.1 (AV:N/AC:L/PR:N/UI:N/C:H/I:H/A:N) that asserts no privileges required and full confidentiality plus integrity impact - the vector conflict matters because it changes who is realistically in scope. Exploitation is not unconditional: the DALL-E 3 tool must be enabled and configured with valid image-generation API credentials and the attacker must be able to drive that pipeline with a controlled project_id; EPSS is 0.14% (3rd percentile), there is no public exploit identified at time of analysis, and the issue is not listed in CISA KEV.

RemediationAI

No vendor-released patch or fixed version is identified in the available data, so treat upstream issue https://github.com/TaskingAI/TaskingAI/issues/375 and the NVD entry https://nvd.nist.gov/vuln/detail/CVE-2026-51906 as the tracking points and confirm whether the maintainers have shipped a corrected build before rolling forward. Until a fix is confirmed, the highest-value control is to disable the DALL-E 3 image-generation tool where it is not required, accepting the loss of that generation capability as the trade-off. If the tool must stay enabled, restrict who can invoke tools so that only trusted principals can supply a project_id, and validate project_id server-side against a strict allowlist (for example an opaque UUID/hex pattern) combined with canonicalizing the joined path and confirming it still resolves inside the intended project image directory - reject anything containing '..', absolute prefixes, or symlink escapes. Run the service as a dedicated unprivileged account with the application/web root, sys.path, PYTHONPATH directories, and configuration/startup files all non-writable to that account, and mount the container or filesystem read-only with only a single dedicated image directory writable; this contains the write primitive but requires an explicit writable volume to be provisioned for legitimate image persistence. As a secondary check, validate the downloaded object's content type and magic bytes before writing, and monitor for image files appearing outside the configured image directory, which is a reliable detection signal for exploitation attempts.

Affected VersionsENISA

n/a n/a

Affects your stack

Find out whether the stacks and servers you run are on an affected version.

Checking this CVE against your own stacks and servers is part of the Basic plan and up.

Technical ContextAI · show details

The flaw is a textbook CWE-22 (Improper Limitation of a Pathname to a Restricted Directory) instance: the save_url_image helper in TaskingAI's DALL-E 3 image-generation tool composes a destination filesystem path from a caller-supplied project_id without canonicalizing it or rejecting '..' segments, so a crafted value causes image bytes fetched from the DALL-E 3 response to land outside the intended project image directory. TaskingAI is an open-source Python platform for building LLM applications and tool integrations, so the vulnerable code is in the tool-output persistence layer rather than in the model API itself. The write primitive is constrained - the file content is a generated image that only the upstream API response partially governs, and the destination is a single derived path - so escalating the primitive into code execution requires an additional condition such as a writable web/doc root, a directory on sys.path or PYTHONPATH, or an overwritable config, template, or startup file; that constraint is why the independent assessment stops at integrity impact with no confidentiality impact. The CPE supplied in the source data is a placeholder (cpe:2.3:a:n/a:n/a:*:*:*:*:*:*:*:*) and EUVD's affected-version field is likewise n/a n/a, so there is no authoritative vendor CPE binding to rely on; the description pins the affected code to TaskingAI v0.3.0.

Know when the next one hits your stack

  • Watch products and get an email when something new is published
  • Check this CVE against your own stack and servers
  • Exploit and PoC references, threat intelligence from 22 sources

No credit card · 7-day full trial

Severity by source

Vendor (mitre) PRIMARY 9.1 CRITICAL
AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:N
vuln.today AI 6.5 MEDIUM

Invoking the DALL-E 3 tool requires application access so PR:L; traversal yields arbitrary file write giving I:H; no direct read so C:N and no inherent DoS so A:N.

3.1 AV:N/AC:L/PR:L/UI:N/S:U/C:N/I:H/A:N
4.0 AV:N/AC:L/AT:N/PR:L/UI:N/VC:N/VI:H/VA:N/SC:N/SI:N/SA:N

Primary rating from Vendor (mitre).

CVSS VectorVendor: mitre

Attack Vector
Network
Attack Complexity
Low
Privileges Required
None
User Interaction
None
Scope
Unchanged
Confidentiality
High
Integrity
High
Availability
None

Lifecycle Timeline

3
CVSS changed
Oct 06, 2026 - 03:22 NVD
9.1 (CRITICAL)
CVE Published
Oct 02, 2026 - 00:00 cve.org
CRITICAL 9.1
CVE Published
Oct 02, 2026 - 00:00 cve.org
UNKNOWN (no severity yet)

DescriptionCVE.org

In TaskingAI v0.3.0 in the DALL-E 3 image generation tool save_url_image function, a path traversal vulnerability allows attackers to write downloaded images to arbitrary locations on the server filesystem by manipulating the project_id parameter.

Executive ViewAI

Why It Matters

TaskingAI v0.3.0's DALL-E 3 image-generation tool introduces a critical file-integrity risk for organizations that have enabled and credentialed that feature. The scoring conflict is material: the CNA vector indicates unauthenticated (PR:N) and full confidentiality/integrity impact, while the independent vector indicates authenticated (PR:L) and integrity-only impact, so the realistic attacker scope remains disputed. No vendor-released patch identified at time of analysis, but no public exploit identified at time of analysis and exploitation is conditional on tool configuration and valid image-generation API credentials.

Business Impact: High
Compensating Controls

Disable the DALL-E 3 image-generation tool in TaskingAI v0.3.0 if not business-essential; restrict tool invocation to trusted users and service accounts; validate project_id with an allow-list and canonical path checks to block traversal sequences; confine save_url_image writes to a dedicated non-executable directory; run the TaskingAI service with least privilege and remove write access to sensitive paths; monitor logs and file integrity for anomalous writes. No vendor-released patch identified at time of analysis.

Threat Intelligence Overview

EPSS Score
0.1%
Top 97% exploitable
Prevalence
Low
emerging open-source AI-agent platform, limited production footprint
Data Quality
5 sources
Confidence: Medium

How this decision was reached

The engine reads each signal below and applies a fixed list of rules from top to bottom. The first rule that matches sets the recommended action – like a statement, every line is shown.

SignalValue
Severity CRITICAL
CISA KEV No
Public POC No
EPSS 0.1%
Patch available No
Active threat No
Deciding rule
CRITICAL severity
What every action level means
Emergency Immediate action required. Criteria: CRITICAL + in CISA KEV (actively exploited) without patch, OR CRITICAL + EPSS > 30% without patch, OR CRITICAL + active threat signals without patch.
Act Now Prioritize within 24-48h. Criteria: in KEV with patch available, OR CRITICAL + public POC, OR HIGH + EPSS > 10%, OR CRITICAL + active threat, OR any CRITICAL severity.
This Week Schedule remediation this week. Criteria: HIGH severity (CVSS 7.0-8.9). Patch within normal change window.
This Month Plan within 30 days. Criteria: MEDIUM severity (CVSS 4.0-6.9), lower exploitation probability. Include in regular patch cycle.
Monitor Track and reassess. Criteria: LOW severity (CVSS 0.1-3.9), minimal exploitation risk. No immediate action needed.

SSVC AssessmentENISA

SSVC Decision
Scheduled
Exploitation
Poc
System Exposure· derived
Open
Automatable
Yes
Human Impact
High
Technical Impact
Total

Priority Score

46
Low Medium High Critical
KEV: 0
EPSS: 0
CVSS: +46
PoC: 0
Weaponized: 0
Threat intel: 0
Gov advisory: 0

Attack Technique

Path traversal exploits occur when applications use user-controlled input to construct file system paths without proper validation.

This site uses cookies essential for authentication and security. No tracking or analytics cookies are used. Privacy Policy