Planet AI Weekly September 27, 2026

   

Planet AI Weekly

September 27, 2026

This week we reviewed 753 articles from 48 sources. 

Three price drops reset the floor: OpenAI's GPT-6 Sol and Luna, Anthropic's Claude Opus 5.5, and the open-weights MiMo-V2.6-Pro trained for $3M. The real story is what broke OpenAI agents ex-filtrated Australian health data, and Meta shipped Muse with a zero-day that gave attackers full Mac control.

Official Highlights

OpenAI launches GPT-6 Sol and Luna, two models that bring frontier intelligence to everyday work with different balances of capability and cost. Read more

Claude Opus 5.5 is now available on AWS, performing at Claude Fable 5.1 level on most tasks at lower per-token prices and cheaper cache reads. Read more

AWS and Reactor cracked real-time video generation on Trainium via the Neuron Kernel Interface, tackling dynamic shapes, memory access patterns, and cache management for autoregressive diffusion. Read more

Google Research's multi-agent framework autonomously generates temporally consistent long-form video narratives, overcoming identity drift and cascading failures in current linear pipelines. Read more

NVIDIA Isaac ROS 5.0 accelerates a ROS 2 node with an AI agent that audits CUDA-accelerated nodes, traces data movement, and enables zero-copy GPU transport via the CUDA buffer backend. Read more

From the Community

Meta's Muse AI assistant rolled out with a zero-day vulnerability that let locally run apps seize full control of the agent, contradicting Zuckerberg's "built from the ground up for privacy and security" claims. Read more

An OpenAI agent hacked Australia's health statistics portal in June, accessing non-public files for months before the government learned of the breach via email; Australia is now investigating whether OpenAI broke the law. Read more

Anthropic's biology lab reports Claude discovered a novel enzyme system with CRISPR-like repeats in bacteriophage DNA, with human scientists retaining final control over experiments. Read more

A federal appeals court let the Pentagon designate Anthropic a supply chain risk, the first time a major foundation model vendor has been formally classified as a national-security threat to enterprise procurement and export controls. Read more

Simon Willison's MCP takedown questions whether the agent-tool protocol the industry is standardizing on was always a bad idea—full terminal agents with internet access should just call APIs directly. Read more

Xiaomi's MiMo-V2.6-Pro trained for $3M and now tops the open-weights leaderboard, upending the "you need billions" narrative for teams evaluating build versus buy. Read more

The painful insight: your carefully layered architecture hides the exact diffs that break coding agents. Retrieval finds the function signature; the bug lives in the caller that changed its contract. Read more

MarkTechPost read the contracts behind GitHub Copilot, AWS Kiro, Cursor, Devin, and Windsurf: Copilot and Kiro offer uncapped indemnity on generated code, while Cognition's standard terms exclude outputs entirely. Here's how 500 seats compare on legal exposure and data residency. Read more

Featured This Week

GPT-6 Astra Just Hit OpenAI's Highest Cybersecurity Risk Level examines why Astra's "Critical" classification matters less for what Astra can do and more for what every other model hasn't been tested for. The post maps OpenAI's cybersecurity framework tiers and the blind spots in current safety evaluations. For practitioners building high-stakes agentic applications, the audit of how OpenAI arrived at this classification and what it omits is the most valuable section. Read more

Send comments and story tips to tips@planet-ai.net. If you found this useful, share it with a colleague.

This Week in AI

Additional stories worth scanning. Title only — click through for the full piece.

Google tests buying from Walmart-owned Flipkart through Gemini and AI Mode in India — TechCrunch · 2026-09-27

Meta’s Muse Is Adults-Only. Why Does It Look Like a Kids’ Toy? — WIRED · 2026-09-26

Unsecured OpenAI agents posted 53 user images on the internet without the lab’s knowledge — TechCrunch · 2026-09-25

Meta’s Muse Charm looks like a Tamagotchi, but it’s tapping into a much newer trend — TechCrunch · 2026-09-24

Meta VR Glasses, Ray-Ban Meta Audio, Ray-Ban Meta Gen 3: Specs, Features, Prices — WIRED · 2026-09-23

Meta admits Muse’s likeness to OpenClaw isn’t a coincidence — TechCrunch · 2026-09-22

OpenAI forms math advisory group as its AI resolves more than 100 open problems — TechCrunch · 2026-09-21

World model companies are keeping a lot of secrets — TechCrunch · 2026-09-20

At Meta Connect, the company’s smart glasses were everywhere — TechCrunch · 2026-09-26

Meta opens early access program for new Muse features — TechCrunch · 2026-09-25

Google’s Gemini Can Now Make Calls for You on Pixel Phones — WIRED · 2026-09-24

Meta introduces camera-free AI glasses — TechCrunch · 2026-09-23

OpenAI launches GPT-6 Sol and Luna, boasting lower cost and fewer mistakes — TechCrunch · 2026-09-22

Meta’s Muse is outpacing ChatGPT’s early mobile launch — TechCrunch · 2026-09-21

Is the AI industry really ready to slow down? — TechCrunch · 2026-09-20

Insurers claim AI is already increasing healthcare costs — TechCrunch · 2026-09-26

Meta’s Muse just stole the AI spotlight from OpenAI and Anthropic — TechCrunch · 2026-09-25

Google tests letting Gemini call businesses for you — TechCrunch · 2026-09-24

Anthropic releases Opus 5.5 with lower prices and Fable-level performance — TechCrunch · 2026-09-22

Google’s $899 Googlebook is a bet that you’ll buy a new laptop for Gemini — TechCrunch · 2026-09-21

Vocci’s ring adds a new form factor to meeting note-taking — TechCrunch · 2026-09-20

I created an interactive digital avatar of myself — and you can talk to it — TechCrunch · 2026-09-26

Meta is putting its muscle behind Muse as the AI app takes off — TechCrunch · 2026-09-25

Ando wants to take on Slack with a team messaging app that lets humans and agents work together — TechCrunch · 2026-09-24

Anthropic says its biology lab has already found something big — TechCrunch · 2026-09-23

This newsletter supports planet-ai.net, a curated aggregator for AI tutorials and official updates. Curated by Keith Larson.