Infoblox researchers found that the Chinese DCloud Uni-App framework underlies more than 236,000 scam domains, from fake crypto exchanges to pig‑butchering operations. The shared code lets disparate fraudsters quickly clone phishing and investment‑scam sites, amplifying a global scam economy that can even spill into real‑world schemes.
Saxe shows GLM-5.2 can be run locally on eight H200 GPUs, bypassing API logging, and matches GPT-5.5‑Cyber in code and terminal tasks, giving attackers long‑horizon autonomous capabilities. This removes the biggest friction for malicious use of frontier AI, likely spawning a dark‑economy of inference services.
Tracebit ran 951 attack simulations with ten frontier AI models in an AWS cyber range. The models achieved admin privilege escalation in 162 cases, typically within minutes, and 95.9% of those trips triggered a canary an average of 8 minutes before the critical action.