Chainalysis breach exposes all user credentials amid agentic AI attacks
Blockchain analysis firm Chainalysis shut down key services after an agentic cyber attack exposed names, usernames, password hashes, and live tokens for every user.
Daily coverage of AI, developer tools and infrastructure. Each story explains what happened and why it matters, with a link to the original source.
Blockchain analysis firm Chainalysis shut down key services after an agentic cyber attack exposed names, usernames, password hashes, and live tokens for every user.
Mesh Applied released Goethe, an open-source tool that turns Linux machines with NVIDIA GPUs into private, OpenAI-compatible API servers in three steps.
OpenAI launched GPT-6 Cyber, a model designed for security research that discovered hundreds of Linux and Firefox flaws while demonstrating stricter adherence to safety boundaries.
NVIDIA details how TensorRT LLM VisualGen accelerates FLUX 3 Image Klein generation using NVFP4 quantization, sparse attention, and multi-GPU scaling on Blackwell hardware.
The AbMaze benchmark evaluates AI agents on protein fitness landscapes. Gemini 4 ties a Bayesian optimizer, while other models lag behind and show varying failure disclosure rates.
New techniques in vLLM, SGLang, and Kubernetes aim to cut LLM server startup times by reusing weights and capturing engine state.
Thinking Machines Lab releases Inkling 2 Small, a 594B parameter Mixture-of-Experts model that excels in agentic coding, vision, and autonomous research tasks.
Attackers are exploiting CVE-2026-123158 to bypass authentication on password-disabled accounts in OpenSSH running on musl libc, with active exploits confirmed on Alpine Linux.
Apple researchers use small neural networks to approximate support functions, speeding up maximum inner product search for retrieval-augmented generation.
Nubank researchers detail a simulation workflow that screens AI customer experience agents using synthetic users, improving metrics without exposing live customers to errors.
AWS releases Web Answer, a managed tool that returns synthesized, source-cited answers to natural language questions, eliminating the need for custom RAG pipelines.
Raindrop releases a public beta for its Simulations platform, which reconstructs production environments to test AI agent behavior before deployment.
Microsoft Research introduces PDD Kit, a toolkit that applies formal verification to agent-written software, ensuring code matches user intent through mathematical proofs and targeted testing.
OpenAI’s autonomous pipeline fixed nearly 1,000 Linux kernel vulnerabilities in one week, while Moonshot released the open weights for its flagship Kimi K3.1 Pro model.
NVIDIA has released four new open-weight checkpoints in its Cosmos 3 family, enabling developers to generate interactive driving and robotics simulations with synchronized multi-sensor data.