Threat pattern / Contrast
DeepSeek Harness CVE-2026-82533 (CVSS 9.4): a sandboxed agent could flip its own session to 'danger-full-access' through the unauthenticated local UI
The Hacker News
Stack layer / Threat pattern
Claude Code 2.1.269 Ships a Plugin Eval Runner and a Knob to Raise the Workflow Tool's Concurrent Agent Cap to 256
Anthropic (claude-code CHANGELOG)
Stack layer / Threat pattern
Salesforce frames its agent stack as an 'Enterprise AI Harness' with a separate AI Control Plane
Salesforce
Stack layer / Threat pattern
GreyNoise Traced One Attacker Running OpenAI's Codex Harness With a DeepSeek Model Through 395 Organizations in 48 Countries
Help Net Security
Stack layer / Threat pattern
100 LLM agents running a simulated town economy for 26 weeks froze the money supply: 0.3% of prices ever changed, and memory deletion made no difference
arXiv
Stack layer / Threat pattern
easyspecs.ai Sells "Spec Review" as a Standalone Product Category, #6 on Product Hunt With 152 Votes
Product Hunt leaderboard for 2026-09-11 (single source; the product site was not independently verified)
Stack layer / Threat pattern
Tailscale gates customer model access by tailnet identity and issues no API keys to agents at all
Vercel Blog
Policy dependency / Threat pattern
Thousand Launches Git-Backed Markdown Docs Where Teammates, Outsiders and Agents Each See Only Their Folders
Thousand (via the Product Hunt leaderboard for 2026-09-10; single source)