AI news, 9 November 2026

Apple pushes Siri upgrades for iOS 27.3 and Spring third‑party integration

Apple plans fine‑grained control of first‑party apps (e.g., rotate/crop Photos, organize Safari tabs) in iOS 27.3 and third‑party app actions like Grubhub orders in a Spring release, aiming to match OpenAI’s Dots and Meta’s Muse.

Google launches Elephant‑Goldfish plugin for spec‑first coding with five zero‑context reviews

Google’s Elephant‑Goldfish Model plugin makes coding agents draft specs, then runs five zero‑context review stages before any code is generated.

Brava Security raises $15M, launches multi-agent AI platform

Brava Security announced $15 million seed funding as it emerges from stealth, offering a multi‑agent AI platform that generates attack telemetry to identify enterprise defense gaps.

Open‑weight DeepSeek V4.1 Flash achieves 72.8% cyber‑attack success at one‑tenth cost

In the Vals AI benchmark, OpenAI's GPT‑6 Sol Daybreak hit 80.8% of attack objectives, while open‑weight DeepSeek V4.1 Flash reached 72.8% at roughly one‑tenth the cost. Longer runtimes were critical for open models to demonstrate capability.

Alibaba's Co-Design Bench shows AI coding agents fully solve only 26.4% of tasks

Alibaba's Co-Design Bench benchmark tests AI agents on real user change requests. GPT-6 Astra, the top model, meets about 70% of task requirements on average but fully solves only 26.4% of tasks, whereas older benchmarks report >99% scores.

Microsoft Research shuts AI Frontiers group

AI agents collaborating in corporations raise goal‑alignment and ethics concerns

Andrej Karpathy builds HackerTube in a weekend using Claude Opus 5.5

Nous Research releases Arete-27B, cuts Qwen3.8-27B propaganda to 0.9%

MarkTechPost compares routing methods of 10 open‑weight MoE models

GitHub repo solo-to-team-skills gains 291 stars

Open-HumanScale releases 50,000‑hour egocentric video corpus for robot policy training

Agent‑Native Primitives achieve 51.8% on Terminal‑Bench 4.0, narrowing real‑world task gap

Niantic pivots to world‑model AI, targeting home robotics and driverless cars by 2027

Anthropic study finds GPT-5 detects 98% of reward‑hacking but triggers false alarms

Benchmark shows nine AI agents fail to meet expert standards on US lab research tasks

Amazon forms new home robotics team, hires for new-to-world robot product

NYT Cooking adds iOS-only AI recipe modifier, sparks subscriber backlash

OpenAI raises $11B, valuation hits $1.25T

EY says AI alone won't transform customer service, calls for operating model reset

S&P Global says AI will boost industrial robotics growth but manufacturers may benefit less

Recruiters say AI speeds screening, threatens percentage‑based placement fees

Workato says GCCs and Indian IT firms will partner to drive AI enterprise transformation