Weekly · Automated
AI Security — week of 2026-06-22
Automated digest. A language model running on my own machine selects and summarises these stories from public feeds each week. I do not review them item by item, so a headline or summary can be wrong. The labels, such as "corroborated", are generated too. Check the linked source before relying on anything here.
This week's stories
AI Agent Hijacking
The Hacker News · 2026-06-19 · corroborated · reporting
AI agents can be hijacked for code execution, highlighting access control threats.
Also: The Hacker News · BleepingComputer
Malicious AI Plugins
The Hacker News · 2026-06-17 · corroborated · reporting
Malicious plugins can steal AI API keys and compromise supply chains.
Also: The Hacker News · BleepingComputer
Vertex AI SDK Flaw
The Hacker News · 2026-06-16 · single-source · reporting
Vertex AI SDK flaw allows hijacking model uploads, posing security risk.
LiteLLM Vulnerability
The Hacker News · 2026-06-15 · single-source · reporting
LiteLLM vulnerability exposes provider keys and secrets, highlighting security concerns.
Microsoft 365 Copilot Flaw
The Hacker News · 2026-06-15 · single-source · reporting
Microsoft 365 Copilot flaw could let attackers steal emails and files, posing security risk.
AI-Generated Content Misuse
BleepingComputer · 2026-06-19 · corroborated · reporting
AI-generated content can be misused for harassment and other malicious purposes.
Also: The Register Sec
Fable 5 Banned
Simon Willison · 2026-06-16 · corroborated · primary research
AI model banned for fixing code vulnerabilities raises security concerns.
Also: Simon Willison · The Register Sec · Schneier
MosaicLeaks
Hugging Face · 2026-06-18 · single-source · vendor advisory
Hugging Face warns of LLM data leakage risk.
Fable Prompt Injection
Simon Willison · 2026-06-19 · corroborated · primary research
Simple prompts can cause security issues in AI models.
Also: Simon Willison
US Gov AI Use
Schneier · 2026-06-17 · single-source · analysis
Massive AI use by US government raises concerns over security and control.
Also: Schneier
Orphaned AI Agents
The Hacker News · 2026-06-18 · single-source · reporting
Untracked AI agents pose hidden access risks
AI Chat Abuse
The Hacker News · 2026-06-18 · single-source · reporting
Malware delivered via AI chat links