AI News Briefing — June 5, 2026

Following yesterday's News roundup, here's a deeper look at Nemotron 3 Ultra, Nvidia released Nemotron 3 Ultra, an open 550B-parameter Mixture-of-Experts model with 55B active parameters using a hybrid Mamba-Attention architecture for long-running agents. It targets keeping accuracy high while cutting inference cost and latency.

Leading AI labs, executives, and scientists signed a letter urging US lawmakers to mandate screening of synthetic DNA orders to prevent AI-assisted bioweapon development. The push reflects rising concern that models can now guide dangerous lab procedures.

AI industry rivals jointly signed an open letter pressing Congress to require synthetic DNA and RNA sellers to screen orders for dangerous sequences. Signatories include Dario Amodei, Sam Altman, and Mustafa Suleyman.

Ahead of its IPO, Anthropic's Daniela Amodei said annualized revenue crossed $47 billion in May, up from about $9 billion at end of 2025, while dismissing doubts about AI returns. The figures underscore explosive but unproven commercial momentum.

OpenAI published Biodefense in the Intelligence Age, an action plan for AI-powered biological resilience. It accompanies the broader industry push for biosecurity safeguards against AI-assisted threats.

A startup called Flourish, backed by $500M in funding and a reported $2.5B valuation, aims to reinvent AI by studying real neurons to find the brain's core algorithm. Jeff Bezos is among the funders of this neuroscience-driven approach.

Stanford and Lambda Labs published OpenJarvis, an open-source framework running inference, agents, memory, and learning entirely on-device. It reportedly comes within 3.2 points of top cloud models at roughly 800x lower cost and 4x lower latency.

Cloudflare's CEO said bot traffic has overtaken human traffic on the internet, years ahead of forecast, driven by AI agents, and predicts the web's future is pay-to-crawl. The shift could reshape web economics and access for AI systems.