What is CWE-94?
The product constructs all or part of a code segment using externally-influenced input from an upstream component, but it does not neutralize or incorrectly neutralizes special elements that could modify the syntax or behavior of the intended code segment.
CWE-94 is a base-level weakness in MITRE’s Common Weakness Enumeration, with a MITRE likelihood of exploit of Medium. Applicable platforms: Language: Interpreted; Technology: AI/ML.
Source: MITRE CWE (CWE-94 definition, reproduced verbatim). Counts and linkage below are Threadlinqs data.
Consequences
- Access Control — Bypass Protection Mechanism. In some cases, injectable code controls authentication; this may lead to a remote vulnerability.
- Access Control — Gain Privileges or Assume Identity. Injected code can access resources that the attacker is directly prevented from accessing.
- Integrity, Confidentiality, Availability — Execute Unauthorized Code or Commands. When a product allows a user's input to contain code syntax, it might be possible for an attacker to craft the code in such a way that it will alter the intended control flow of the product. As a result, code injection can often result in the execution of arbitrary code. Code injection attacks can also lead to loss of data integrity in nearly all cases, since the control-plane data injected is always incidental to data recall or writing.
- Non-Repudiation — Hide Activities. Often the actions performed by injected control code are unlogged.
Source: MITRE CWE, common consequences.
How CWE-94 is exploited in the wild
Threadlinqs maps 191 CVEs to CWE-94, published between 2008-10-23 and 2026-11-14. 30 are listed in CISA’s Known Exploited Vulnerabilities catalog, the authoritative record of exploitation in the wild, and 6 are tied to ransomware campaigns. By CVSS v3 severity the set splits into 74 critical, 57 high, 23 medium, 19 low. The highest EPSS score in the set is 99.9% (CVE-2017-9841), the modelled probability of exploitation in the next 30 days. 331 tracked threats reference CWE-94 directly or through a CVE it covers; the most recent is “CISA KEV Addition: Perforce Puppet Enterprise Report-Processing Deserialization RCE (CVE-2026-115731)” (2026-11-18). Affected products concentrate in IBM (24), Microsoft (10), SourceCodester (6), among 105 vendors in total.
Vulnerabilities (CVEs)
Showing 40 of 191 CVEs mapped to CWE-94, CISA KEV first, then by CVSS score.
- CVE-2025-32432 — CISA KEV · CVSS 10 critical · EPSS 99.7% · published 2025-04-25
- CVE-2021-22205 — CISA KEV · CVSS 10 critical · EPSS 99.7% · published 2021-04-23
- CVE-2022-22947 — CISA KEV · CVSS 10 critical · EPSS 98.2% · published 2022-03-03
- CVE-2017-9841 — CISA KEV · CVSS 9.8 critical · EPSS 99.9% · published 2017-06-27
- CVE-2025-3248 — CISA KEV · CVSS 9.8 critical · EPSS 99.9% · published 2025-04-07
- CVE-2023-3519 — CISA KEV · CVSS 9.8 critical · EPSS 99.7% · published 2023-07-19
- CVE-2022-22965 — CISA KEV · CVSS 9.8 critical · EPSS 99.6% · published 2022-04-01
- CVE-2008-4250 — CISA KEV · CVSS 9.8 critical · EPSS 98.7% · published 2008-10-23
- CVE-2026-1281 — CISA KEV · CVSS 9.8 critical · EPSS 98.6% · published 2026-01-29
- CVE-2026-1340 — CISA KEV · CVSS 9.8 critical · EPSS 98.6% · published 2026-01-29
- CVE-2023-25717 — CISA KEV · CVSS 9.8 critical · EPSS 98.0% · published 2023-02-13
- CVE-2025-54068 — CISA KEV · CVSS 9.8 critical · EPSS 97.3% · published 2025-07-17
- CVE-2026-9198 — CISA KEV · CVSS 9.8 critical · EPSS 28.6% · published 2026-07-17
- CVE-2026-33017 — CISA KEV · CVSS 9.8 critical · EPSS 24.7% · published 2026-03-20
- CVE-2026-60004 — CISA KEV · CVSS 9.8 critical · EPSS 23.9% · published 2026-08-26
- CVE-2026-72530 — CISA KEV · CVSS 9 critical · EPSS 1.6% · published 2026-08-19
- CVE-2025-49704 — CISA KEV · CVSS 8.8 high · EPSS 99.9% · published 2025-07-08
- CVE-2012-0158 — CISA KEV · CVSS 8.8 high · EPSS 99.9% · published 2012-04-10
- CVE-2021-22894 — CISA KEV · CVSS 8.8 high · EPSS 41.2% · published 2021-05-27
- CVE-2026-34197 — CISA KEV · CVSS 8.8 high · EPSS 15.4% · published 2026-04-01
- CVE-2026-65660 — CISA KEV · CVSS 8.8 high · EPSS 2.1% · published 2026-08-11
- CVE-2026-3910 — CISA KEV · CVSS 8.8 high · EPSS 1.0% · published 2026-03-13
- CVE-2025-62593 — CISA KEV · CVSS 8.8 high · EPSS 1.0% · published 2025-11-26
- CVE-2020-8243 — CISA KEV · CVSS 7.2 high · EPSS 90.7% · published 2020-09-29
- CVE-2025-4428 — CISA KEV · CVSS 7.2 high · EPSS 86.9% · published 2025-05-13
- CVE-2026-15410 — CISA KEV · CVSS 7.2 high · EPSS 76.3% · published 2026-07-14
- CVE-2021-22900 — CISA KEV · CVSS 7.2 high · EPSS 14.1% · published 2021-05-27
- CVE-2026-0322 — CISA KEV · CVSS 7.2 high · EPSS 1.7% · published 2026-10-14
- CVE-2021-22204 — CISA KEV · CVSS 6.8 medium · EPSS 99.9% · published 2021-04-23
- CVE-2024-20359 — CISA KEV · CVSS 6 medium · EPSS 19.4% · published 2024-04-24
- CVE-2025-59528 — CVSS 10 critical · EPSS 86.4% · published 2025-09-22
- CVE-2026-26216 — CVSS 10 critical · EPSS 6.4% · published 2026-02-12
- CVE-2026-26030 — CVSS 10 critical · EPSS 3.7% · published 2026-02-19
- CVE-2026-10134 — CVSS 10 critical · EPSS 2.3% · published 2026-06-30
- CVE-2026-45829 — CVSS 10 critical · EPSS 1.0% · published 2026-05-18
- CVE-2026-58231 — CVSS 10 critical · EPSS 0.8% · published 2026-08-11
- CVE-2026-73453 — CVSS 10 critical · EPSS 0.6% · published 2026-09-16
- CVE-2026-73456 — CVSS 10 critical · EPSS 0.6% · published 2026-09-16
- CVE-2026-61447 — CVSS 10 critical · EPSS 0.5% · published 2026-07-11
- CVE-2024-9264 — CVSS 9.9 critical · EPSS 97.5% · published 2024-10-18
Affected vendors
- IBM — 24 CVEs
- Microsoft — 10 CVEs
- SourceCodester — 6 CVEs
- Adobe — 3 CVEs
- Apache Software Foundation — 3 CVEs
- Apple — 3 CVEs
- CGM — 3 CVEs
- FlowiseAI — 3 CVEs
- Ivanti — 3 CVEs
- Johnson Controls — 3 CVEs
- Langflow — 3 CVEs
- SAP_SE — 3 CVEs
Threat activity
331 tracked threats cite CWE-94; the 25 most recent are listed.
- CISA KEV Addition: Perforce Puppet Enterprise Report-Processing Deserialization RCE (CVE-2026-115731)CRITICAL
- Malicious GitHub Repositories Hijack AI Coding Agents (Claude Code, Cursor) via Poisoned Instruction Files to Steal Developer CredentialsHIGH
- Atlassian November 2026 Security Bulletin: 318 vulnerability entries across Jira, Confluence, Bitbucket, Bamboo, Crowd, Fisheye and Crucible (52 critical), plus Confluence Axis RCE CVE-2026-90436CRITICAL
- Dragos Q3 2026: Industrial Ransomware Incidents Up 11% to 1,266, Manufacturing 69% of VictimsHIGH
- Kaspersky ICS CERT: APT and Financial Attacks on Industrial Organizations in Q2 2026 (DAEMON Tools Supply Chain, Iran-Linked PLC Attacks, ZionSiphon, INC Ransomware)HIGH
- GTIG warns fully-automated agentic AI attacks imminent: exploit_agent, CAIRN, ARTEX abuse and SOMBERMEME via modified Claude Code (MIDNIGHT NEPTUNE / UNC6836)HIGH
- CVE-2026-21656 - Johnson Controls Frick Controls Quantum HD Pre-Authentication Code Injection (ICSA-26-057-01)CRITICAL
- Ransomware Leak-Site Cadence Analysis: INCRANSOM, TheGentlemen and Audit Team Run Active Intrusion Pipelines Amid Group ConsolidationMEDIUM
- CVE-2026-21658: Pre-Authentication Input Validation / Code Injection Flaw in Johnson Controls Frick Controls Quantum HD (10.22 and earlier, legacy through 11)CRITICAL
- CVE-2026-21657 - Johnson Controls Frick Controls Quantum HD Insufficient Input Validation (Pre-Authentication Code Injection)CRITICAL
- CVE-2026-21656 - Johnson Controls Frick Controls Quantum HD Code Injection via Insufficient Input Validation (Pre-Authentication)CRITICAL
- ARTEX open-source agentic AI pentest framework abused by Chinese-nexus state and criminal actors to automate intrusions at machine speed (FBI Flash Alert)HIGH
- Palo Alto Networks PAN-OS (Panorama) Code Injection RCE as Root - CVE-2026-0330HIGH
- Malware and Credential Exposure in AI Agent Instruction Files (Skills, Hooks, MCP Configs) - Mitiga Skillgate ResearchHIGH
- ThreatsDay Roundup: Microsoft MDASH Linux Kernel Bugs (CVE-2026-64563), Falcon Browser Spyware, GSocket BackdoorsHIGH
- SGLang RCE Exploited Against NEAR AI Inference Server — Data Stolen From 18 Agents and ~100 UsersHIGH
- Microsoft Patch Tuesday November 2026: ~1,095 vulnerabilities, 64 critical, one exploited Outlook NTLM-leak spoofing zero-day (CVE-2026-124194)CRITICAL
- AI-agent-coordinated attacks and 575+ malicious OpenClaw skills delivering malware (Acronis TRU Cyber Pulse 2026)HIGH
- Kaspersky Q3 2026 Non-Mobile Threat Landscape: Qilin Ransomware via CVE-2026-0257, INC Ransom via SonicWall SMA1000 CVE-2026-15409/15410, macOS Stealers (PamStealer, MacSync, Overlord) and npm/PyPI/crates.io Supply-Chain BackdoorsHIGH
- CTO at NCSC Weekly Summary (week ending July 26, 2026): Cl0p PTC Windchill/FlexPLM exploitation (CVE-2026-12569), SonicWall SMA 1000 zero-days (CVE-2026-15409/CVE-2026-15410), and multi-actor state-linked activityHIGH
- Microsoft November 2026 Patch Tuesday: Outlook Zero-Day CVE-2026-124194 Exploited to Steal NTLM Hashes (1,235 Flaws Fixed)HIGH
- ForgeRock Attack Chain: Default admin:admin Felix Console Credentials and WAF Path-Encoding Bypass Lead to RCE and Remote Vehicle Unlock at Two AutomakersHIGH
- AI Reward Hacking and Autonomous Sandbox Escape: OpenAI ExploitGym Evaluation Models Breach Hugging Face (Schneier DEF CON 34 Keynote Op-Ed)HIGH
- Metasploit 6.5.9 Wrapup: Public Exploit Modules for Citrix NetScaler (CVE-2026-88771), Cisco Secure Email Gateway (CVE-2026-76461/76443), GitLab CI/CD Regex RCE (CVE-2026-97229), OpenBullet2, Flowise, SPIP and WordPress Bug LibraryCRITICAL
- LLMjacking Bills Reach $1M (Mandiant/GTIG; LiteLLM CVE-2026-40217), Trezor Safe 5 Reseller Implant, SAP Flaws Weaponized Within DaysHIGH
Mitigations
- Architecture and Design: Refactor your program so that you do not have to dynamically generate code.
- Architecture and Design: Run your code in a "jail" or similar sandbox environment that enforces strict boundaries between the process and the operating system. This may effectively restrict which code can be executed by your product. Examples include the Unix chroot jail and AppArmor. In general, managed code may provide some protection. This may not be a feasible solution, and it only limits the impact to the operating system; the rest of your application may still be subject to compromise. Be careful to avoid CWE-243 and other weaknesses related to jails.
- Implementation / Input Validation: Assume all input is malicious. Use an "accept known good" input validation strategy, i.e., use a list of acceptable inputs that strictly conform to specifications. Reject any input that does not strictly conform to specifications, or transform it into something that does. When performing input validation, consider all potentially relevant properties, including length, type of input, the full range of acceptable values, missing or extra inputs, syntax, consistency across related fields, and conformance to business rules. As an example of business rule logic, "boat" may be syntactically valid because it only contains alphanumeric characters, but it is not valid if the input is only expected…
- Testing: Use automated static analysis tools that target this type of weakness. Many modern techniques use data flow analysis to minimize the number of false positives. This is not a perfect solution, since 100% accuracy and coverage are not feasible.
- Testing: Use dynamic tools and techniques that interact with the product using large test suites with many diverse inputs, such as fuzz testing (fuzzing), robustness testing, and fault injection. The product's operation may slow down, but it should not become unstable, crash, or generate incorrect results.
- Operation / Compilation or Build Hardening: Run the code in an environment that performs automatic taint propagation and prevents any command execution that uses tainted variables, such as Perl's "-T" switch. This will force the program to perform validation steps that remove the taint, although you must be careful to correctly validate your inputs so that you do not accidentally mark dangerous inputs as untainted (see CWE-183 and CWE-184).
- Operation / Environment Hardening: Run the code in an environment that performs automatic taint propagation and prevents any command execution that uses tainted variables, such as Perl's "-T" switch. This will force the program to perform validation steps that remove the taint, although you must be careful to correctly validate your inputs so that you do not accidentally mark dangerous inputs as untainted (see CWE-183 and CWE-184).
- Implementation: For Python programs, it is frequently encouraged to use the ast.literal_eval() function instead of eval, since it is intentionally designed to avoid executing code. However, an adversary could still cause excessive memory or stack consumption via deeply nested structures [REF-1372], so the python documentation discourages use of ast.literal_eval() on untrusted data [REF-1373].
Source: MITRE CWE, potential mitigations.
Detection methods (MITRE CWE)
- Automated Static Analysis (effectiveness: High): Automated static analysis, commonly referred to as Static Application Security Testing (SAST), can find some instances of this weakness by analyzing source code (or binary/compiled code) without having to execute it. Typically, this is done by building a model of data flow and control flow, then searching for potentially-vulnerable patterns that connect "sources" (origins of input) with "sinks" (destinations where the data interacts with external components, a lower layer such as the OS, etc.)
Source: MITRE CWE, detection methods. Threadlinqs detection rules for the threats above are Blue tier and higher.