# News: September 16, 2026

> 13 curated links on AI agents, Python, security, Google Cloud, and engineering, selected by Médéric Hurier.

- Author: [Médéric Hurier (Fmind)](https://www.fmind.dev/)
- Published: 2026-09-16
- URL: https://www.fmind.dev/news/2026-09-16/

- [Introducing Gemini 3.8 Live and 3.8 Live Extended Thinking](<https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-3-8-live-gemini-3-8-live-extended-thinking>): Google unveils Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking, introducing low-latency dialogue models with real-time reasoning and proactive audio capabilities.
- [Agent Substrate available on GKE](<https://cloud.google.com/blog/products/containers-kubernetes/agent-substrate-available-on-gke>): Google open-sources Agent Substrate on GKE, a high-density agent runtime engineered to scale millions of sandboxes on a single Kubernetes cluster.
- [Filestore agent volumes](<https://cloud.google.com/blog/products/storage-data-transfer/filestore-agent-volumes>): Google Cloud introduces Filestore agent volumes, providing secure, fully managed, elastic file storage purpose-built for AI agent workspaces.
- [Build real-time voice applications with Gemini 3.8 Live and 3.5 Transcribe](<https://blog.google/innovation-and-ai/technology/developers-tools/build-real-time-voice-applications-gemini-audio>): Google announces developer availability for Gemini 3.8 Live, Gemini 3.8 Live Extended Thinking, and Gemini 3.5 Transcribe to power real-time voice applications.
- [What's new in the Gemini Live API](<https://www.youtube.com/watch?v=3CyW24Pkz4o>): Google DeepMind demonstrates new Gemini Live API capabilities, including async function calling, proactive audio, context injection, and background extended thinking.
- [Build a real-time voice AI agent with Google ADK and Gemini Live API](<https://www.youtube.com/watch?v=yQEKMsCtsmE>): Google Cloud Tech demonstrates how to build bidirectional, real-time voice agents using the Google Agent Development Kit (ADK) and Gemini Live API.
- [Build zero-trust AI agents that judge intent, not just syntax](<https://developers.googleblog.com/build-zero-trust-ai-agents-that-judge-intent-not-just-syntax>): Google outlines runtime security architectures for zero-trust AI agents using Model Armor, semantic governance, and anomaly detection on the Gemini Enterprise Platform.
- [Humanist AI Code of Conduct](<https://microsoft.ai/code-of-conduct>): Microsoft AI releases its Humanist AI Code of Conduct, defining 15 fundamental behaviors, safety constraints, and evaluation methodologies to align frontier MAI models with human flourishing.
- [OpenAI's malicious bot swarm attacked RubyGems](<https://www.theregister.com/security/2026/09/14/openais-malicious-bot-swarm-attacked-rubygems/5296356>): Security researchers discover that an internal OpenAI agent swarm uploaded thousands of malicious packages to RubyGems to execute remote code and scrape data during training runs.
- [AI Makes Familiar Cyber Attacks Cheaper to Run](<https://jasondoyle.ie/whitepapers/ai-makes-familiar-cyber-attacks-cheaper-to-run>): A security whitepaper details how AI drastically reduces attacker labor and coordination costs through faster retries, dynamic adaptation, and parallel operations.
- [Sharpen your study routine with new Gemini Notebook tools](<https://blog.google/innovation-and-ai/products/gemini-notebook/new-study-tools-september-2026>): Google adds new study and learning tools to Gemini Notebook to help students organize materials and synthesize course content.
- [Anthropic researchers are quitting... and now we know why](<https://www.youtube.com/watch?v=7r4ikZHm9AI>): Fireship reviews a 154-page safety report revealing how external attackers and rogue models compromised Claude, sparking internal resignations at Anthropic.
- [The Risk Dario’s AI Warning Leaves Out](<https://www.youtube.com/watch?v=_6PR25pkHKI>): Matthew Berman examines Anthropic CEO Dario Amodei's essay "We Must Pace the Frontier," weighing the dangers of rogue AI against the threat of regulatory capture.