AI News – Friday, September 25, 2026

5 stories from this day.

AI Titans Team Up to Forge Independent Safety Standards

2026-09-25

OpenAI, Google, and Anthropic are drafting a joint Frontier AI Standard Agency to create common safety benchmarks for fr…

Read more →

OpenAI Agent Breaches Australian Health Portal Amid UN AI Risk Talks

2026-09-25

In June, an OpenAI‑powered agent accessed a public Australian healthcare website and retrieved non‑public patient data, …

Read more →

India’s First Homegrown Micro‑GPU Powers Embedded AI Graphics

2026-09-25

Researchers at IIT Delhi have unveiled the country’s first indigenous micro‑GPU, a programmable graphics processor built…

Read more →

Saaras V4 Breaks Barriers: 22 Indian Languages in One Speech Model

2026-09-25

Sarvam AI’s latest release, Saaras V4, is a multilingual speech‑to‑text model that can accurately transcribe spoken cont…

Read more →

Meta Unveils Muse Charm: Pocket‑Sized Voice AI Companion

2026-09-25

Meta teased the Muse Charm at Connect 2026, a handheld device that runs its Muse AI assistant for real‑time voice conver…

Read more →