AI News Briefing — June 7, 2026

At Computex 2026, Nvidia unveiled RTX Spark, a Windows-PC version of its Blackwell GB10 superchip, with Microsoft launching Surface Laptop Ultra and a Dev Box and OEMs including Asus, Dell, Lenovo, HP, and MSI shipping systems. It positions powerful local AI hardware against earlier Arm-based Copilot+ efforts.

OpenAI and the Trump administration are reportedly negotiating a direct government stake in the company, framed as a Public Wealth Fund that could pay American citizens directly. Senator Sanders is separately pushing a 50 percent tax on AI shares, while critics warn the deal could create too-big-to-fail dynamics.

President Trump indicated the administration is discussing arrangements where the US government and public could share in AI's financial upside, potentially including an equity stake in OpenAI. The comments point to an unusual blending of state ownership and frontier AI.

An SEC filing reveals SpaceX is leasing roughly 110,000 Nvidia AI chips to Google for about $920 million per month to support its Gemini Enterprise platform. The arrangement highlights how scarce AI compute has become and how interdependent major tech players now are.

OpenAI introduced Lockdown Mode for ChatGPT, designed to reduce the risk of sensitive data being exfiltrated through prompt injection attacks. The company acknowledges it does not fully eliminate the vulnerability but aims to lower the likelihood of data leakage.

A new open-source voice model called Audio Interaction processes audio in a continuous stream, deciding roughly every 0.4 seconds whether to speak or stay silent rather than waiting for a recording to end. It handles translation, transcription, chat, and ambient sounds, with weights and code released under Apache 2.0.

New York's legislature approved a one-year moratorium on hyperscale datacenters above 20MW, which would make it the first US state to enact such a ban if Governor Hochul signs it. The vote signals intensifying political resistance to the energy and land demands of AI infrastructure.

Reporting alleges xAI trained its coding models on Anthropic's Claude outputs for months, continuing via private accounts and a third-party service even after Anthropic cut off access. The piece also describes xAI's pretraining team shrinking to a handful of people and surplus compute being rented to rivals.

Meta is developing a paid AI agent called Hatch, reportedly priced up to $200 per month, that builds working tools, schedules appointments, and sends emails from plain-language requests. It would be Meta's first paid AI product as Zuckerberg seeks revenue beyond advertising to fund massive AI spending.

NVIDIA released Nemotron 3.5 ASR, a 600M-parameter cache-aware streaming speech recognition model that transcribes 40 language-locales in real time from a single checkpoint with built-in punctuation and capitalization. The open-weights model uses a FastConformer-RNNT architecture with prompt-based language conditioning and is available on Hugging Face.