ai_security▌

2026-11-17

16 item(s) verified and swept on 2026-11-17, each citing a primary source.

NOTABLECVETOOL

GitLab CVE-2026-129253: developer role could bypass admin-configured AI tool governance in governed namespaces

GitLab's Nov 17 security release fixes CVE-2026-129253 (CVSS 4.3, medium): improper binding of authorization tokens to their intended workflow execution context allowed an authenticated developer-role user to bypass admin-configured AI tool governance controls in a governed namespace. Affects GitLab EE 19.1 before 19.3.9, 19.4 before 19.4.8 and 19.5 before 19.5.6.

NOTABLECVETOOL

AdminNeo AI plugins served prompts to unauthenticated callers using the operator's Claude/Gemini keys (CVE-2026-130358)

GHSA-g35j-5g93-mgff (CVSS 6.5): the ClaudeSqlPlugin, GeminiSqlPlugin and OpenWebUiPlugin sendHeaders() handlers in AdminNeo before 5.9.1 run on the login page without authentication, so an unauthenticated attacker could POST claude, gemini or openwebui fields to relay arbitrary prompts using the operator's API key or bearer token, including against internal Open WebUI endpoints. Fixed in 5.9.1.

NOTABLETOOLSUPPLY-CHAIN

arXiv: Compositional Exploits — coding agents chain individually harmless tools into attack chains in most exploit tasks

Coding agents (Antigravity, Claude Code, Cursor, OpenHands) are audited tool-by-tool at install time; arXiv 2611.19764 shows emergent compositional exploits instead: the agent stitches legitimate tool operations across turns over shared workspace files, environment variables and chat memory to execute an unauthorized chain, no malware involved. On a 560-task suite (360 exploit tasks across 8 hazard classes, 200 benign workflows), unprotected Gemini 3.5, 3.7 and 3.8 Flash execute exploit chains with 72.9-84.5% success; the paper proposes an ephemeral defense.

NOTABLERED-TEAM

arXiv: deep-research agent security systematized — planning stage under-studied; 'agenda seeding' beats prompt injection

arXiv 2611.20045 systematizes security for deep-research and agentic search agents along retrieval, planning, evidence integration and report generation: a unified threat model, attack/defense taxonomy and benchmark review. Coding 34 primary studies, only 26.5% involve the planning stage and none performs an end-to-end poisoning evaluation on commercial products; the authors' agenda-seeding attack (introduce a fictitious evaluation metric, write no instructions) makes the attacker's target the top pick across 3,520 sandboxed research runs where prompt injection and GEO are nearly ineffective.

NOTABLETOOL

arXiv: MCP tool-provider misbinding — an attacker-run MCP server can win agent tool calls through metadata alone

MCP clients compose tools advertised by multiple servers, but the tool metadata used for discovery and selection is not bound to a verified provider identity. arXiv 2611.21577 formulates tool-provider misbinding as an attacker-controlled tool-selection optimization: an adversary operating only another MCP server — no host, model, legitimate-server or channel compromise — gets a request intended for a legitimate provider resolved to its own tool, strongest when exact-name collision, task-aligned description and compatible schema reinforce each other; measured across five LLM-based MCP clients.

NOTABLERED-TEAM

arXiv: control-gated probes reconstruct incomplete-fix patch bypasses in four LLM agent frameworks

Replaying a patched vulnerability's original PoC proves the reported trigger closed, not that the invariant holds on other code paths. arXiv 2611.20514 runs benign, control-gated sibling-path probes across four LLM agent frameworks: at PraisonAI 4.6.31 the original tool-loading trigger is blocked while the tool_override path still violates the invariant — reproducing the publicly documented CVE-2026-44334 bypass of CVE-2026-40287 — and CrewAI release 1.11.0 shows a violation that is confined in 1.11.1; before/after reconstructions also cover Langroid and Agno.

NOTABLESUPPLY-CHAIN

arXiv: suppression-aware VLA backdoors switch malicious robot behavior off during tests via instruction tokens

arXiv 2611.19774 builds suppression-aware backdoors on Vision-Language-Action models: conjunctive trigger activation combined with instruction-conditioned suppression — in the Self-Suppressing Trigger variant a [DISRUPT] instruction token gates activation and a [RESTORE] token suppresses the attack — so checks built for trigger-based backdoors pass. Variants (Cross-Modal Trigger, Self-Suppressing Trigger, Multi-Trigger) are evaluated with SmolVLA on two SO-100 manipulation datasets against State Backdoor, BadVLA, TabVLA and SilentDrift baselines.

NOTABLEPROMPT-INJECTION

arXiv: A2A Agent Cards steer agent selection before invocation; 310 live cards measured, SECCARD validator proposed

Agent2Agent Agent Cards contain attacker-controllable natural-language fields that can influence an orchestrator's agent-selection decision before a provider is invoked. arXiv 2611.18887 taxonomizes ten Agent-Card manipulation attacks, probes 36.1 million Internet domains and collects 310 deployed Agent Cards (96.8% two-annotator agreement), and presents SECCARD, a layered protocol/deterministic/LLM-semantic validation framework whose strongest semantic judge reaches 97.0% security recall at 1.7% false-positive rate on controlled attacks injected into real cards.

NOTABLEPROMPT-INJECTION

arXiv: prompt injection against LLM-generated industrial compliance artifacts — Safety Data Sheets, QC documents, product passports

arXiv 2611.19090 studies prompt injection against LLMs used to skim and generate industrial documents — Safety Data Sheets, Quality-Control documents and Digital Product Passports — examining how adversaries can manipulate agents into producing compromised compliance artifacts, a deployment surface the chat, browser-addon and coding-agent injection literature has left largely unexplored.

NOTABLERED-TEAM

arXiv: frozen guardrail probes fail open as models update — false-positive rate jumps from 1.01% to 15.12%

arXiv 2611.19216 freezes a guardrail probe and its threshold, then re-measures the operating point across 18 cells covering 16 public language-model updates: at a 1% false-positive target the largest jump is 1.01% to 15.12% while AUC changes by no more than 0.0058 in any cell, and the benchmark's own staleness criteria classify none of them as stale; a label-free rate estimate flags 0.918 of the layer cells, and resetting the threshold using stored benign prompts restores the rate without new labels.

NOTABLEPOLICY

Anthropic cuts its coordinated vulnerability disclosure window to 30 days for high and critical findings

Anthropic's Frontier Red Team is shortening its coordinated disclosure timelines from 90 days to 30 days for high- and critical-severity issues and to 45 days for all other issues, under an updated policy starting January 1, 2027.

NOTABLEPOLICY

OpenAI starts pilot embedded assessments with Apollo Research, METR, Transluce and Goodfire

OpenAI named Apollo Research, METR, Transluce and Goodfire as its pilot embedded assessors, external evaluators that get deeper access to independently evaluate model alignment, safeguards and monitoring systems. The announcement sets out scoping protocols for the first phase of the program.

NOTABLECVEPROMPT-INJECTION

Timestream for InfluxDB MCP server sent its API token to attacker-chosen endpoints on an injected URL override

GHSA-2f9r-g4f3-wc2j: before 0.0.19 the AWS Labs MCP server accepted an InfluxDB URL introduced through the MCP client context, so a prompt-injected endpoint change sent the configured API token to an attacker-controlled host. Fixed in 0.0.19, with token rotation required.

NOTABLECVE

AWS Pricing MCP server report tool wrote files outside its output directory on agent-influenced paths

GHSA-mqm3-w68j-75mw: before 1.0.31 the cost-report generation tool accepted absolute paths and traversal sequences in the report output path, so anyone able to influence the MCP agent could write attacker-controlled content to arbitrary locations. Fixed in 1.0.31.

NOTABLECVE

DeepAudit unauthenticated config endpoint returned operator LLM provider keys and repository tokens

CVE-2026-129877: versions 2.0.0-beta.1 through 3.8.0 of the LLM-driven code-audit tool served server secrets at /api/v1/config/defaults, letting unauthenticated attackers retrieve the operator's LLM provider API keys and GitHub or GitLab tokens and abuse those services and reachable repositories.

NOTABLETOOL

Windows 11 telemetry reports which coding agents and developer tools a machine ran

A reverse-engineering write-up by Xusheng Li documents a Windows telemetry component that assigns one bit per tool category, covering codex.exe, claude.exe, copilot.exe and cursor.exe plus common developer tools, and includes runtime, time-in-focus and interaction counters in the report. The post identifies the reporting component and the privacy setting that governs the collection.

← 2026-11-18 · 2026-11-16 → · all days