Skip to content
Source · Daily Brief

AI Daily Brief — 4 August 2026

Safety and education framed OpenAI’s output, with published third-party cyber evaluations, a criminal-scam disruption report, and new ChatGPT Work and Codex learning tools. NVIDIA released its Alpamayo 2 Super driving model for commercial use, Google recapped its July releases, and Texas moved to gate data-center connections to its power grid.

Top stories

  • OpenAI details a criminal scam takedown. The lab described disrupting malicious uses of its models tied to a scam operation. via OpenAI
  • OpenAI publishes third-party cyber evaluations. External assessments of its models’ cyber capabilities were released. via OpenAI
  • Anthropic names a Chief Global Affairs Officer. Mariano-Florentino (Tino) Cuéllar joins to lead the company’s global policy work. via Anthropic
  • NVIDIA opens its Alpamayo 2 Super driving model. The frontier open model for robotaxis and autonomous vehicles is now available for commercial use. via NVIDIA Blog
  • Google recaps its July releases. A roundup collected the AI updates the company announced in July 2026. via Google AI Blog
  • Google publishes a unified model-routing API. The developer post outlines one API for routing across AI models. via Google AI / Gemma
  • Sakana AI expands its Daiwa Securities project. The joint work on wealth-management support AI moved to a full-scale phase. via Sakana AI

Who shipped

OpenAI led the day with new ChatGPT Work and Codex education tools alongside two safety writeups. Google summarised its July work and detailed a unified model-routing API. NVIDIA released Alpamayo 2 Super as an open model for autonomous driving and posted developer guides on generating trajectories and reasoning traces with it. Anthropic focused on leadership rather than models, adding a global-affairs chief.

Open-source pulse

NVIDIA‘s Alpamayo 2 Super arrived as a frontier open model cleared for commercial use, the clearest open release of the day. llama.cpp continued its steady nightly cadence with a long run of tagged builds, and a DiffusionGemma technical report surfaced among the day’s papers.

Quiet corners

Meta, xAI, DeepSeek and Alibaba‘s Qwen team were absent from the primary feeds. Mistral appeared only through an older scraped search-toolkit page rather than a fresh announcement, and Microsoft stayed quiet aside from GitHub Copilot housekeeping notes.

Money, infra & hardware

AMD reported its datacenter segment climbing while gaming took a smaller share, per The Verge – AI. Texas moved to halt new data-center connections to its grid and require an audit before hookup, via Ars Technica – AI. NVIDIA also flagged storage and memory demand rising as AI workloads grow.

By the numbers

  • 130 stories across 20 sources
  • Most-mentioned model: GPT
  • Most-mentioned lab: OpenAI
  • Notable absences: Meta, xAI

Compiled by AI Feed’s editor from all 130 headlines published on 4 August 2026.