Summary
Wes Roth examines what he calls “Black October,” a surge of cyberattacks that followed the October 21 disclosure of a major PHP vulnerability. He argues that autonomous AI agent swarms are now attacking at a scale and persistence security teams have not seen before, citing a researcher whose honeypot server was hit 14,000 times in six hours with no human traffic.
The video pulls together several reports: Quest Diagnostics confirming a breach of patient data, major US banks detecting unauthorized third-party AI agents inside customer accounts, and a Microsoft investigation of Storm-0940, a China-aligned threat actor that reportedly went from initial targeting to administrative cloud access in about three hours using several models, some of them open source. Roth also notes Cloudflare’s finding that exploit attempts began days before the public announcement, Anthropic’s Jason Clinton warning that conditions will worsen, and Sam Altman’s remark in a Stanford GSB interview that not every bug can be patched.
A central argument is the asymmetry between closed models, which providers can shut off, and open-weight models like DeepSeek, which cannot be recalled once released and can be modified to remove safeguards. The video includes a sponsored segment for Crusoe Intelligence Foundry.
📺 Source: Wes Roth · Published October 24, 2026
🏷️ Format: Opinion Editorial
