ASR AI Security Radar
Back to homepage

Recent AI Security Incidents

This archive includes all published incident pages. Page 12 of 18.

Each page is intended to help a security team answer three questions quickly: why the issue is AI-relevant, what part of the workflow may be exposed, and what actions should happen first.

Selection criteria and correction policy are documented in Methodology & Editorial Policy.

AI security incident: OpenClaw has a LFI in BlueBubbles media path handling (GHSA-rwj8-p9vq-25gv)

Incident date: February 18, 2026 | Published: February 25, 2026

Summary The BlueBubbles extension accepted attacker-controlled local filesystem paths via mediaPath and could read arbitrary local files from disk before sending them as media attachments.

Read details

AI security incident: OpenClaw inter-session prompts could be treated as direct user instructions (GHSA-w5c...

Incident date: February 18, 2026 | Published: February 25, 2026

Summary Inter-session messages sent via sessions_send could be interpreted as direct end-user instructions because they were persisted as role: "user" without provenance metadata.

Read details

AI security incident: Libredesk has a SSRF Vulnerability in Webhooks (GHSA-wgm6-9rvv-3438)

Incident date: February 18, 2026 | Published: February 25, 2026

**Date:** 2025-12-07 **Vulnerability:** Server-Side Request Forgery (SSRF) **Component:** Webhooks Module ## Executive Summary A critical security vulnerability exists in the LibreDesk Webhooks module that allows an authenticated...

Read details

AI security incident: OpenClaw: Command hijacking via unsafe PATH handling (bootstrapping + node-host PATH...

Incident date: February 18, 2026 | Published: February 25, 2026

Command hijacking via PATH handling **Discovered:** 2026-02-04 **Reporter:** @akhmittra ## Summary OpenClaw previously accepted untrusted PATH sources in limited situations.

Read details

AI security incident: OpenClaw affected by denial of service via unbounded webhook request body buffering (...

Incident date: February 18, 2026 | Published: February 25, 2026

Summary Multiple webhook handlers accepted and buffered request bodies without a strict unified byte/time limit. A remote unauthenticated attacker could send oversized payloads and cause memory pressure, degrading availability.

Read details

AI security incident: OpenClaw affected by denial of service through unguarded archive extraction allowing...

Incident date: February 18, 2026 | Published: February 25, 2026

Summary Archive extraction lacked strict resource budgets, allowing high-expansion ZIP/TAR archives to consume excessive CPU/memory/disk during install/update flows. ## Affected Packages / Versions - openclaw (npm): <= 2026.2.

Read details

AI security incident: OpenClaw: denial of service through large base64 media files allocating large buffers...

Incident date: February 18, 2026 | Published: February 25, 2026

Summary Base64-backed media inputs could be decoded into Buffers before enforcing decoded-size budgets. An attacker supplying oversized base64 payloads can force large allocations, causing memory pressure and denial of service.

Read details

AI security incident: OpenClaw exec approvals: safeBins could bypass stdin-only constraints via shell expan...

Incident date: February 18, 2026 | Published: February 25, 2026

Summary OpenClaw's exec-approvals allowlist supports a small set of "safe bins" intended to be stdin-only (no positional file arguments) when running tools.exec.host=gateway|node with security=allowlist .

Read details

AI security incident: OpenClaw has a command injection in maintainer clawtributors updater (GHSA-m7x8-2w3w-...

Incident date: February 18, 2026 | Published: February 25, 2026

Summary Command injection in the maintainer/dev script scripts/update-clawtributors.ts . ### Impact Affects contributors/maintainers (or CI) who run bun scripts/update-clawtributors.

Read details

AI security incident: OpenClaw Google Chat spoofing access with allowlist authorized mutable email principa...

Incident date: February 17, 2026 | Published: February 25, 2026

Summary Google Chat allowlisting supports matching by sender email in addition to immutable sender resource name ( users/ ). This weakens identity binding if a deployment assumes allowlists are strictly keyed by immutable principals.

Read details