日本語
2026-07-24 Evening edition
Evening edition
AI News Daily

AI News Daily
2026-07-24

OpenAI moved on product and compute at once, publishing a live agent resolution rate alongside a multi-gigawatt data centre build, while Washington took frontier-model leakage to government level.

Today

Today’s highlights

Story 01 · Corporate developments

OpenAI unveils “OpenAI Presence,” an enterprise AI agent platform

Published: 2026-07-22
  • An enterprise agent platform handling customer service and internal work by voice and chat.
  • OpenAI’s own phone support desk resolves 75% of enquiries with no human involved.
  • Built for production use, with permission settings and monitoring.
Why it matters

A resolution rate from real operation, not a benchmark, gives companies weighing AI for call centres and internal help desks a concrete yardstick for their adoption decision.

75%
of enquiries resolved without human involvement
Source: Tier 1, official website — OpenAI https://openai.com/index/introducing-openai-presence/
Story 02 · Corporate developments

OpenAI announces “Project Camellia” in Effingham County, Georgia

Published: 2026-07-22
  • A new data centre site OpenAI will design and build itself.
  • Power ramps in phases from 2028 to 2032, reaching 3.2 GW; investment above $30 billion.
  • An $80 million community fund plus technology for educational institutions.
Why it matters

An AI company taking on construction and local negotiation itself shows the scale investment has reached, and confirms that securing compute is now the core of competitive position.

3.2 GW
supplied in phases, 2028 to 2032
$30 billion
total investment, exceeding
Source: Tier 1, official website — OpenAI https://openai.com/index/building-ai-infrastructure-with-the-effingham-county-community/
Story 03 · Regulation and policy

U.S. accuses Moonshot AI of improperly distilling an Anthropic model for “Kimi K3”

Published: 2026-07-23
  • Michael Kratsios, Director of the White House Office of Science and Technology Policy, said Moonshot AI improperly distilled Anthropic’s Fable model when developing Kimi K3.
  • He also raised suspected illicit access to Nvidia chips under export controls, and referred to possible sanctions.
  • Reporting is split: one Tier 2 outlet leads with experts arguing that exploiting Fable is not how Kimi K3 got so good.
Why it matters

U.S. and Chinese friction over model technology leakage and export-control breaches has surfaced in a senior official’s remarks, adding compliance considerations for organisations using or procuring overseas open models.

Sources: Tier 2 — CNBC https://www.cnbc.com/2026/07/23/moonshot-kimi-nvidia-ai-chips-export-ban.html · TechCrunch https://techcrunch.com/2026/07/23/experts-say-exploiting-anthropics-fable-isnt-how-kimi-k3-got-so-good/
Overview

Two trends running through the day

Compute and product investment at once

OpenAI disclosed agent operating results and announced a self-led giant data centre build on the same day, accelerating advance investment on both the compute and the product side.

Frontier-model leakage becomes policy

Exchanges over model distillation and chip export controls have surfaced at government level between the United States and China, making prevention of technology leakage a policy question.

Where they meet

The more a company gains by owning its own stack, the more valuable the resulting capability becomes, and the more its transfer turns into a matter for governments rather than contracts.

Wrap-up

Three things to remember

What to watch next

Whether other vendors follow OpenAI in publishing real-operation resolution rates, and whether the sanctions referred to in the Moonshot AI allegations move from statement to action.