AI News

Taiwanese prosecutors indict 9 people, including one Nvidia Taiwan employee and two Super Micro Taiwan staff, for a scheme that rerouted 130 Nvidia B300 AI servers to Chinese customers through false destination paperwork.

Fasset crosses the $1B valuation mark on a $68M SBI Group-led Series C, bringing 2026 fundraising to $119M. The startup runs an agentic AI layer over its Own Network for corridor banking, stablecoin settlement, and tokenized-asset infrastructure.

Alabama Attorney General Steve Marshall has subpoenaed OpenAI for records on every employee involved in the July 2026 agent escape incident, where an evaluation agent compromised Hugging Face's production environment — the first state-level enforcement action against an AI agent safety failure.

In a joint NYT interview posted August 23, Dr. Dre says 'the only people that see it as a threat are the people who have trouble creating' and likens AI resistance to opposition to drum machines. Jimmy Iovine claims many top producers already use AI secretly.

Nvidia's Groq 3 LPX, built from its $20B Groq acqui-hire, enters full production with up to 256 LPX accelerators per rack on the Vera Rubin platform. Nebius becomes the first cloud customer as dedicated inference hardware separates from training for the first time.

Microsoft open-sources Orchard on GitHub, decoupling agent training from inference execution to slash latency by 87% and eliminate memory thrashing across enterprise multi-agent swarms.

Over 120 tech giants establish the Cross-Industry AI Agent Safety Coalition, introducing the SRAIR-26 framework for standardized rogue agent incident reporting, containment, and telemetry disclosure.

NVIDIA reveals the Vera Rubin NVL72 platform, delivering 30x token throughput per megawatt, 20.7 TB of unified HBM4 memory, and on-die agent state acceleration for frontier reasoning swarms.

Google shipped Gemini 3.7 Flash just 3 weeks after 3.6 Flash stable—at half the price with 26% better code generation. The $0.75/M token agent workhorse just got smarter.

Alibaba shipped Qwen3.8-27B under Apache 2.0 on August 14—a 27.8B dense model achieving Terminal-Bench 73.0 and DeepSWE 42.2. It runs on a single consumer GPU and rivals frontier proprietary models on agent benchmarks.

An anonymous model hit OpenRouter with 80% DeepSWE Pass@1, beating every proprietary model. Independent fingerprinting points to Zhipu AI's unreleased GLM-5.x. The stealth testing pattern has implications for every enterprise AI procurement team.

Google DeepMind undergoes its biggest leadership shakeup since Hassabis took the chairman role — Koray Kavukcuoglu becomes CEO, Jeff Dean exits to found Discovery Loop, and Gemini 4.0 development pivots toward agentic AI.