Loading story…
Semgrep reported that AI coding agents from Anthropic and OpenAI identified genuine vulnerabilities in large open-source web applications, but produced high false-positive rates and inconsistent results across repeated scans. In a study of 11 actively maintained Python web apps, researchers manually triaged 445 findings and confirmed 46 true vulnerabilities from Claude Code and 21 from OpenAI Codex, with roughly 20 of the validated issues rated high severity. The company said responsible disclosure is still underway and did not name the affected applications.
The results showed uneven detection performance by vulnerability class: Claude Code performed best on IDOR, Codex performed best on path traversal, and both models struggled significantly with injection flaws such as SQL injection and XSS. The findings align with the broader role of intentionally vulnerable projects such as OWASP Juice Shop and OWASP Benchmark, which are widely used to evaluate security testing tools against exploitable web application weaknesses across SAST, DAST, and IAST scenarios.
See affected versions and whether adversaries are exploiting it.
1 event from the most recent confirmed update back to the earliest known activity.
Semgrep published research evaluating Anthropic Claude Code and OpenAI Codex for vulnerability discovery across 11 large open-source Python web applications. The study reported 445 manually triaged findings, 46 true vulnerabilities from Claude Code, 21 from Codex, and noted that responsible disclosure was still underway so the tested applications were not named.
Vulnerabilities, threat actors, malware, products, organizations, and breaches Mallory has linked to this story.
See whether adversaries are exploiting this yet, and where the affected versions run in your environment.
3 references tracked. Mallory keeps watching after this page renders.
github.com
Open sourcegithub.com
Open sourcesemgrep.dev
Open sourceMap indicators from this story to your assets and identify affected systems in minutes.
Every observed campaign, victim, and pivot linked to actors named in this story.
Malware, exploits, and IOCs connected to the activity described here.
YARA, Sigma, and Snort rules deployed to your SIEM as soon as they’re published.
Get matching new stories delivered to your team as they break — not the next morning.
Ask questions about this story and take action on the answers.
Other stories Mallory is tracking that overlap on entities, sources, or topic labels.
Two new studies found that AI-generated application code frequently contains exploitable vulnerabilities, with risk varying sharply by model, framework, and application size. Secure Code Warrior said testing of 1,760 complete codebases produced by 16 frontier models uncovered more than **27,000 vulnerabilities**, averaging **15 confirmed flaws per codebase** and **4.3 critical or high-severity issues**. Researchers said no single model was consistently safest across all environments; instead, each model showed a repeatable security fingerprint, with recurring **OWASP** weakness patterns and common omissions such as authentication checks and input validation. Xint.io separately identified **434 exploitable flaws** in AI-generated or AI-hardened applications, reporting that denial-of-service and resource exhaustion bugs, authorization failures, insecure direct object reference, and traversal or **SSRF** issues were common. The most severe findings were tied to hardcoded or default secrets and debug-mode remote code execution, while authorization weaknesses worsened as application size increased, particularly in a larger brownfield **Gnuboard7**-based app. Both reports said organizations should expect continued AI-assisted development and strengthen security review around recurring weaknesses such as hard-coded credentials, sensitive data leakage into logs, cross-site scripting, predictable session tokens or reset codes, and path traversal.
Jul 24, 2026
Researchers examined 9,041 open-source applications built with Claude Code and Lovable, plus 200 publicly deployed applications, and identified 1,186 vulnerabilities. Their audit found that **91%** of deployed applications had at least one flaw and that **65.77%** of identified issues were Critical or High severity, led by broken access control, injection vulnerabilities, and authentication failures. The study attributes recurring defects to eight failure modes involving AI-agent memory, objectives, and knowledge; stronger prompts and agent harnesses reduced, but did not eliminate, exposure. The findings align with reported failures in AI-generated applications, including `CVE-2025-48757` (CVSS 9.3), where Lovable applications using Supabase could expose database tables without authentication when Row Level Security policies were absent. Organizations adopting AI coding agents should enforce human security review, automated application testing, secret scanning, least-privilege authorization, and explicit database access controls before deployment; they should also treat coding assistants and their dependencies as potential supply-chain and prompt-injection targets.
Sep 11, 2026
Multiple studies and industry reports found that **AI-generated or "vibe-coded" software is frequently shipped with serious security flaws**, including XSS, SQL injection, hardcoded secrets, unsafe `eval` usage, weak authentication logic, wildcard CORS, exposed `.env` data, and insecure dependencies. One analysis of 5,600 publicly accessible AI-built applications reported more than 2,000 high-risk vulnerabilities and over 400 exposed secrets, while a separate review of 549 self-described AI-generated repositories found widespread committed secrets, injection issues, hardcoded passwords, and other common sanitization failures. Veracode and other researchers similarly warned that code produced by large language models often maps to **OWASP Top 10** weakness classes and that repeated AI-assisted edits, poor prompts, and excessive agent permissions can further increase risk. At the same time, researchers and vendors highlighted both the promise and fragility of AI-assisted security tooling. Trail of Bits reported that OpenAI Codex goal-based prompting helped uncover previously unknown bugs in projects including **Rust**, **curl**, **zlib**, and **Keycloak**, but stressed that human experts were still needed to scope goals, validate findings, and handle disclosure. New academic work also showed that **LLM-based vulnerability detectors can be easily misled**: the ALIBI framework used adversarial code comments to evade four detectors with success rates above 90%. In response to growing concern over agent reliability, Amazon backed the Lean language project to strengthen formal verification for AI agents, while security experts argued that theorem proving can improve assurance for high-risk components but should be applied selectively as part of broader defense-in-depth.
Aug 4, 2026