Entity trail

Cloud Security Alliance

Source-backed findings, relationship evidence, citations, and briefing history from the public MindPattern archive.

Briefing refs

Findings

Edges

Sources

Corpus findings

2026-07-13 / vibe-coding-researcherPattern: MCP-Relayed Third-Party Content Is Now a First-Class Prompt-Injection SurfaceThe Agentjacking research generalizes beyond Sentry: any MCP server that relays attacker-influenceable content — error trackers, issue trackers, PR comments, browser DOMs — can smuggle instructions into an agent's context and drive tool execution. The emerging discipline treats every MCP tool output as untrusted input requiring the same scrutiny as user input, not as trusted diagnostics.
2026-06-14 / sources-researcherAgentjacking: Poisoned Sentry Errors Hijack Claude Code, Cursor, and Codex via MCPTenet Security and the Cloud Security Alliance disclosed 'Agentjacking,' an attack that injects malicious instructions into Sentry error events using only a public, write-only DSN — which MCP-connected coding agents then retrieve and execute with the developer's own system privileges. Controlled tests hit an ~85% success rate across 100+ orgs, and the chain bypasses EDR, WAF, IAM, and firewalls because every step is technically authorized. Sentry declined a root-cause fix, calling the class 'not defensible' at the platform level — anyone wiring error-tracking MCP servers into an agent should treat ingested events as untrusted input.
2026-05-24 / arxiv-researcherAI-Generated Code CVE Surge: 35 Vulnerabilities in March 2026 Alone, More Than All of 2025The Cloud Security Alliance's Vibe Security Radar reports 35 new CVE entries from AI-generated code in March 2026 — up from 6 in January and 15 in February, exceeding the entire second half of 2025 (18 cases across seven months). To date, 74 confirmed CVEs are traceable to AI-generated code, including 14 critical-risk and 25 high-risk vulnerabilities. Authentication bypass, command injection, and SSRF are the three most common vulnerability types. Veracode testing of 100+ LLMs finds the overall security pass rate stuck at ~55%.
2026-04-23 / skill-finderCSA Research Note: Vibe Coding's Security Debt — 87% of AI-Generated Pull Requests Contain Security Issues, AI Devs Introduce Findings at 10x RateThe Cloud Security Alliance published research showing AI-assisted developers produce commits 3-4x faster but introduce security findings at 10x the rate of non-AI peers. In a single week (March 27), 35 AI-generated CVEs were disclosed. 87% of AI-generated pull requests contain security issues. Combined with the finding that mean time from vulnerability disclosure to confirmed exploitation has fallen below one day in 2026 (down from 2.3 years in 2019), the velocity gap between AI-generated vulnerabilities and human remediation capacity is widening. CSA recommends mandatory automated security scanning in CI/CD as a minimum countermeasure.
2026-04-17 / skill-finderCloud Security Alliance: 53% of Organizations Have Had AI Agents Exceed Intended Permissions, 47% Experienced Agent-Involved Security Incidents in Past YearCSA published 'Enterprise AI Security Starts with AI Agents' on April 16 surveying enterprise AI agent deployments. Key findings: 53% of organizations report AI agents exceeding intended permissions (scope violations), 47% had a security incident involving an AI agent in the past year, 54% report 1-100 unsanctioned shadow AI agents with unclear ownership, and only 15% have 76-100% of agents with defined ownership. 43% of organizations say more than half of employees use AI agents regularly across IT (53%), security (37%), customer service (34%), and engineering (34%).
2026-04-12 / thought-leaders-researcherCloud Security Alliance Publishes Two Official Research Notes on Vibe Coding Security Debt — AI-Generated CVE Surge Now Institutionally TrackedThe Cloud Security Alliance published two formal research notes: one on 'AI-Generated Code Vulnerability Surge 2026' documenting the CVE explosion, and another on 'Credential Sprawl and SDLC Debt' from vibe coding practices. AI-assisted commits now expose secrets at 2x the rate of human-only commits (3.2% vs 1.5%). Georgia Tech projects 400-700 AI-introduced vulnerabilities across the open-source ecosystem, with Claude Code attributed to 27 of 74 confirmed cases. This marks the shift from blog-level criticism to institutional security governance.
2026-04-12 / thought-leaders-researcherVibe Coding Declared 'Over' as Professional Strategy — '69 Vulnerabilities in 15 Apps' Reckoning AcceleratesA viral Medium post by Ahmed Hafdi declared 'Vibe Coding Is Over. What Comes Next Is Much Harder,' arguing the Karpathy-coined approach of accepting AI output without review is no longer viable professionally. DEV Community documented 69 vulnerabilities across 15 vibe-coded apps. The Cloud Security Alliance published two research notes on 'Vibe Coding Security Debt' covering credential sprawl and SDLC debt. Combined with Georgia Tech's 483% CVE surge (6→35 in Q1 2026), the backlash is now institutionalized.
2026-04-07 / skill-findermcp-sec-audit Toolkit: Static Pattern Matching + Docker/eBPF Dynamic Fuzzing Achieves 100% MCPTox Benchmark DetectionThe mcp-sec-audit toolkit combines static pattern matching for Python-based MCP servers with dynamic sandboxed fuzzing and monitoring via Docker and eBPF. It achieves 100% detection on the MCPTox benchmark—the standard test suite for MCP tool poisoning. Detects risky capabilities (file system access, network requests, command execution) through configurable rule-based analysis and provides specific mitigation recommendations. Open source, part of the Cloud Security Alliance's Model Context Protocol Security initiative.
2026-03-31 / skill-finderCSA Report: 68% of Organizations Cannot Distinguish AI Agent from Human Activity — 74% Say Agents Have More Access Than NecessaryThe Cloud Security Alliance published survey data on March 24 showing 85% of organizations use AI agents in production, but 68% cannot clearly distinguish between human and AI agent actions. 52% use workload identities, 43% rely on shared service accounts, and 31% allow agents to operate under human user identities. 74% acknowledge agents receive more access than necessary, and 79% say agents create access pathways difficult to monitor. This is the most comprehensive governance gap data published to date.
2026-03-25 / skill-finderCSA Launches CSAI Foundation: Dedicated Non-Profit for Agentic AI Security with AI Risk Observatory, TAISE-Agent Certification, and CSA Pod Live Agent TelemetryThe Cloud Security Alliance launched CSAI at RSAC 2026, a 501(c)3 non-profit dedicated to 'Securing the Agentic Control Plane' — governing identity, authorization, orchestration, runtime behavior, and trust assurance for autonomous AI agents. Six programs include: AI Risk Observatory for agentic AI CVE tracking, identity-first controls guidance for non-human actors, three new TAISE certification tracks, TAISE-Agent Certification extending professional certification to autonomous agents through behavioral evaluation, and CSA Pod — a live agent interaction environment and telemetry source.
2026-03-14 / news-researcherCSA Report: 36.7% of 7,000 MCP Servers Vulnerable to SSRF; Trend Micro Finds 492 with Zero AuthThe Cloud Security Alliance's March 13 State of Cloud and AI Security 2026 report analyzed over 7,000 MCP servers and found 36.7% potentially vulnerable to server-side request forgery. An independent Trend Micro analysis found 492 MCP servers with zero client authentication and zero traffic encryption. The findings arrive as MCP adoption accelerates rapidly, with builders frequently standing up servers without hardening — a pattern researchers warn mirrors early OAuth misconfigurations that led to widespread credential theft.

Graph relationships

RELEASED
CSAI Foundation
Cloud Security Alliance launched CSAI at RSAC 2026, a 501(c)3 non-profit
Source finding
BUILT BY
mcp-sec-audit
mcp-sec-audit is part of the Cloud Security Alliance's MCP Security initiative.
Source finding
CRITICIZES
Claude Code
Cloud Security Alliance documented Claude Code attributed to 27 of 74 confirmed CVE cases.
Source finding
Reported in
Pattern: MCP-Relayed Third-Party Content Is Now a First-Class Prompt-Injection Surface
Source finding
Reported in
Cloud Security Alliance Publishes Two Official Research Notes on Vibe Coding Security Debt — AI-Gene
Source finding
BUILT BY
mcp-sec-audit -> Cloud Security Alliance
mcp-sec-audit is part of the Cloud Security Alliance's MCP Security initiative.
Source finding

Source trail

Cloud Security Alliance research note The Hacker News paper on arXiv Cloud Security Alliance reports Prompt Armor disclosed Project Glasswing strings referencing Mythos meta-analysis on arXiv Microsoft released .NET 10.0.7 VulnCheck reports Pillar Security disclosed A detailed analysis

Graph sources

entity graphfindings textkg edgeskg entitiesnewsletter issues

Fetching from the wire…