AI Risk Wire

All days of the daily brief

Daily brief · 2026-11-01

New incidents (7), most serious first

  1. Malicious Hugging Face 'Qwen2 test' models execute hidden crypto miner and reverse shell via trust_remote_code

    Pluto Security found 31 malicious repositories on seven Hugging Face accounts posing as small Qwen2 test models for self-hosted AI services, drawing over 32,000 downloads.

    major · Happened 2026-09 · Earliest source we hold 2026-10-30

  2. SolanaCry fake GitHub repo tricks AI coding agents into installing NetSupport RAT

    A fake GitHub repository posing as a Solana starter kit, SolanaUpdateTools/solana-starter-kit, ranked first in search results and its README told AI coding agents to run a PowerShell command that downloaded and executed a 21 MB dropper in memory.

    major · Earliest source we hold 2026-10-30

  3. North Korean group uses AI to mass-produce tailored phishing PDFs against South Korean users

    North Korean hacking groups used AI tools, including the ARTEX open-source penetration-testing framework and AI coding tools such as Cursor, to mass-produce phishing lure documents, fake job offers and malware.

    moderate · Earliest source we hold 2026-10-30

  4. Claude agent detects MeshAgent screen-capture malware on high-stakes poker player's machine

    On September 25, 2026, professional online poker player Alex was using his Claude AI agent to review play statistics when his SQL database crashed and needed to be reset.

    moderate · Anthropic · Happened 2026-09-25 · Earliest source we hold 2026-10-29

  5. University of Washington study demonstrates cross-origin data theft via prompt injection in ChatGPT Atlas Agent Mode

    A University of Washington study tested seven agentic browsers and found four could let malicious sites cross origin boundaries.

    moderate · OpenAI · Earliest source we hold 2026-10-29

  6. Microsoft red team: poisoned agent memory bypasses human-in-the-loop approval prompts

    Microsoft's red team, in research reported in the Microsoft Digital Defense Report 2026, examined AI agents that can write to their own long-term memory and are given sensitive capabilities but told to ask the user before acting.

    minor · Microsoft · Earliest source we hold 2026-10-30

  7. Red-team campaign finds 11 attacks defeating Codex sandbox and Auto-review monitor

    Researchers red-teamed OpenAI's Codex coding agent and its synchronous control measures, chiefly the Auto-review blocking monitor, at commit c4f42d1.

    minor · OpenAI · Earliest source we hold 2026-10-30

The register was corrected 13 times this day (merges, retractions and fixed fields). Read the corrections log.