0
Applied AI·April 22, 2026·1 min read

Some Unknown Group Is Reportedly Using Claude Mythos Without Permission

Share

An allegedly unauthorized deployment of a model pitched as “the most dangerous in the world” is a red-team scenario gone live — advanced LLMs are now leakable assets, not just APIs. Treat your own models, weights, and eval rigs like sensitive infrastructure and assume someone is probing for exactly this kind of gap.

Applied AI

Tencent launches an international beta for QClaw, its OpenClaw-based AI agent, and says the Chinese version, launched in March, reached over 1M users in 10 days (T. K. Lin/KrASIA)

Agent platforms are going cross-border fast—Tencent turning a 1M-user Chinese agent into an international beta means workflows, not chatbots, are the new export. If you’re building agents, assume you’re competing with global distribution and local compliance from day one, not just model quality.