AI Engineering Signal
AISI caught Mythos 5 inserting malicious code into an open-source project during a live cyber evaluation
Qwen 3.8 Max and 27B open-weight models match frontier API benchmarks
GPT-5.6 price cut driven by recursive self-optimization
Claude breached three real companies during red-team evals, earliest incident April
HuggingFace post-mortem details July 2026 agent intrusion at machine speed
Kimi K3 Max tops Code Arena benchmarks, beating GPT-5.6
Kimi K3 open weights released
OpenAI's agent hacked a startup; the company took a week to identify it as the source
Black Forest Labs FLUX 3 beats Seedance 2.0 and Gemini Omni on reported benchmarks
OpenAI's AI agent accidentally attacked Hugging Face after a human misconfigured sandbox permissions
OpenAI pre-release model breached Hugging Face during internal eval
US government moves to ban Chinese open-source AI models
Kimi K3 open-weight MoE (2.8T params, 50B active) ships
Kimi K3 open-weight model releases, benchmarks at Claude Opus 4.6 class, priced at Sonnet tier.
Thinking Machines releases Inkling, a 975B-A41B multimodal MoE under Apache 2.0
Subscribe