| PocketOS: Cursor agent running Claude Opus 4.6 deletes production database and backups in nine seconds |
2026-04-27 |
Real Estate |
High
|
PocketOS |
— |
Opus 4.6 |
Claude |
| Open VSX: ‘GlassWorm’ campaign seeds 73 lookalike extensions targeting AI coding IDEs |
2026-04-25 |
Technology |
High
|
Open VSX Registry (Eclipse Foundation) |
— |
— |
— |
| PocketOS: Cursor AI coding agent deletes production database and backups |
2026-04-25 |
Technology |
High
|
PocketOS |
United States |
Opus 4.6 |
Claude |
| Metropolitan Police: Live facial-recognition cameras wrongly identify roofer as furniture thief |
2026-04-22 |
Government & Public Sector |
Medium
|
Metropolitan Police |
United Kingdom |
— |
— |
| Google: ‘Groundfall’ prompt-injection flaw in Antigravity IDE bypasses Secure Mode, enables remote code execution |
2026-04-21 |
Technology |
Low
|
Google |
— |
— |
Google |
| SGLang: Critical command-injection flaw in AI inference server enables remote code execution via malicious model files |
2026-04-21 |
Technology |
Low
|
SGLang |
— |
— |
— |
| Lovable: Tenant-isolation flaw exposed source code and credentials for every project created before November 2025 |
2026-04-20 |
Technology |
Critical
|
Lovable |
Sweden |
— |
— |
| Anthropic: Unsafe MCP SDK defaults expose 7,000+ AI agent servers to remote code execution |
2026-04-20 |
Technology |
Medium
|
Anthropic |
United States |
— |
Claude |
| Vercel: Third-party AI vendor breach exposes customer credentials via OAuth token theft |
2026-04-20 |
Technology |
Critical
|
Vercel |
United States |
— |
— |
| Belo: Claude AI wrongly suspends 60+ fintech accounts in Argentina |
2026-04-18 |
Finance & Banking |
Medium
|
Belo |
Argentina |
— |
Claude |
| Anthropic, Google, GitHub: Cross-vendor ‘Comment and Control’ prompt injection leaks AI coding agents’ API keys |
2026-04-16 |
Technology |
Low
|
Anthropic, Google, GitHub |
United States |
— |
Claude, Gemini |
| Bee Cheng Hiang: AI-generated email script exposes 95,000+ customer addresses |
2026-04-15 |
Retail & E-commerce |
Low
|
Bee Cheng Hiang |
Singapore |
— |
— |
| Greg Lake: Omaha Attorney Suspended Indefinitely Over AI-Hallucinated Court Brief |
2026-04-15 |
Other |
High
|
Greg Lake |
United States |
— |
— |
| The New York Times: Reporter’s AI tool fabricates quote attributed to Pierre Poilievre |
2026-04-14 |
Media & Entertainment |
Low
|
The New York Times |
United States |
— |
— |
| Malicious LLM Routers: Research Finds Credential Theft, One Customer Loses $500,000 Crypto Wallet |
2026-04-13 |
Technology |
Medium
|
— |
— |
— |
— |
| Sullivan & Cromwell: AI-hallucinated citations found throughout federal bankruptcy court filing |
2026-04-09 |
Consulting & Professional Services |
Medium
|
Sullivan & Cromwell |
United States |
— |
— |
| Multiple organizations: Ransomware affiliate uses Cursor agent and Claude Sonnet to automate intrusion across 20+ companies |
2026-04-08 |
Other |
High
|
Multiple organizations (Christeyns, Teckentrup, Helideck Certification Agency, Bayou Title, and others) |
— |
Sonnet 4.5 |
Claude |
| FlowiseAI: Critical RCE flaw in Flowise agent builder actively exploited across 12,000+ instances |
2026-04-07 |
Technology |
High
|
FlowiseAI |
United States |
— |
— |
| Grafana Labs: ‘GrafanaGhost’ prompt-injection chain could have silently exfiltrated dashboard data |
2026-04-07 |
Technology |
Low
|
Grafana Labs |
— |
— |
— |
| KBS: AI real-time translation subtitles broadcast Korean profanity during Artemis II livestream |
2026-04-02 |
Media & Entertainment |
Low
|
KBS (Korean Broadcasting System) |
South Korea |
— |
— |
| Purolea Cosmetics Lab: FDA issues first warning letter citing inappropriate AI use in drug manufacturing |
2026-04-02 |
Manufacturing |
Medium
|
Purolea Cosmetics Lab |
United States |
— |
— |
| Mercor: AI startup breached via LiteLLM supply-chain attack |
2026-04-02 |
Technology |
High
|
Mercor |
United States |
— |
— |
| Anthropic: Claude Code deny-rule bypass flaw could let malicious commands evade permission checks |
2026-04-01 |
Technology |
Medium
|
Anthropic |
United States |
— |
Claude |
| U.S. Special Operations Command: AI chatbot hallucination nearly triggers strike on Chinese vessel |
2026-04-01 |
Government & Public Sector |
Critical
|
U.S. Special Operations Command |
United States |
— |
— |
| Google Cloud: ‘Vertex AI Double Agent’ research shows compromised agents can exfiltrate cloud credentials |
2026-04-01 |
Technology |
Low
|
Google Cloud (Vertex AI) |
United States |
— |
— |