AI News, 15 November 2026

Sunday 15 November

Google to disable AI tools by default on school‑managed accounts

Source-page capture accompanying Google to disable AI tools by default on school‑managed accounts

AI-generated summary Google announced that it will disable AI-driven features in its services for school-managed student accounts by default, starting in the United States before expanding worldwide. School administrators can re-enable the tools, and the policy does not apply to third-party AI tools or personal student accounts.

Earlier:

Google launches Gemini Orchestrator

SI Agent News. SI is the new AI. Two AI podcast hosts on a newspaper front page

AI-generated summary Google has launched the Gemini Orchestrator within Google Antigravity to help developers switch dynamically between different AI models during a single development session. Data from the Terminal-Bench 4.0 shows the Orchestrator achieved nearly the same task success rates as the Carbon model while using 24% less quota.

Earlier:

Leading Open-Weight Chinese AI Models Trained On Restricted Nvidia Chips, A New Research Claim

AI-generated summary Research led by Will DePue found evidence that Chinese artificial intelligence developers may have trained models on Nvidia Blackwell processors. The analysis examined models from Moonshot AI, DeepSeek, MiniMax and Ant Group, raising fresh questions about export controls.

Also covered bySI Agent NewsCrazycodersLabelseif
Earlier:

Australia workplace AI guide: consultation opens

Australia workplace AI guide: consultation opens

AI-generated summary Australia released a draft workplace artificial intelligence adoption guide for public consultation on 12 November 2026, setting out a voluntary approach to managing technological change with workers. Employment and Workplace Relations Minister Amanda Rishworth announced the draft Best Practice Guide following consultations with unions, employer representatives, academics, and businesses.

Earlier:

MACaQ: Lower WikiText2 perplexity than MR-GPTQ and DuQuant++ on Llama-3-8B, faster quantization and a reported 64 minutes for 122B expert layers

devotd. — Clear, source-backed reporting on IT, AI, LLMs and local AI.

AI-generated summary Huawei CSL released MACaQ to create deployable 4-bit models by processing each linear layer independently to reduce quantization time and required GPU memory. In paper experiments on Llama-3-8B using WikiText2, MACaQ achieved lower perplexity than MR-GPTQ and the DuQuant++ family.

Microsoft Opens Kuwait Digital Innovation Center for AI, Cloud and Security Training

Microsoft Opens Kuwait Digital Innovation Center for AI, Cloud and Security Training

AI-generated summary Microsoft opened a Digital Innovation Center in Kuwait on November 15 to provide public sector staff, startups, developers and students with access to AI, cloud and cybersecurity tools. Microsoft has not disclosed the cost, size, partner list or programme schedule for the new facility, which was set up with the Kuwaiti government.

Also covered byAI Ping

Gemini Glasses vs. Apple's Reported Glasses: What You Can Buy Now, and What's Only Rumored

AI-generated summary Google is currently accepting preorders for its Gemini glasses, while Apple has not yet confirmed any details about its reported glasses. Google says its glasses are priced from $395 at Warby Parker and feature the Gemini assistant, a 12-megapixel camera, and open-ear speakers.

Also covered byMacworldHIMIAO Horizon
Earlier:

LangChain releases version 1.2.0 with Groq integration updates and LangSmith routing

Source-page capture accompanying LangChain releases version 1.2.0 with Groq integration updates and LangSmith routing

AI-generated summary LangChain announced version 1.2.0 of its Groq integration package with updates including routing ChatGroq calls through the LangSmith gateway. The update also introduces a performance service tier, maps context-length errors to a ContextOverflowError, and refreshes model-profile metadata.

Earlier:

Compromised MCP server can hijack AI agents by exploiting weak authentication

Source-provided image accompanying Compromised MCP server can hijack AI agents by exploiting weak authentication

AI-generated summary VentureBeat reports that the Model Context Protocol has historically lacked robust authentication controls, leaving AI agents vulnerable when connecting to external tools. Palo Alto Networks’ Unit 42 tested a scenario where a compromised server achieved a 78.3% attack success rate by injecting malicious instructions into an AI agent.

Earlier:

Hedge funds fold on the AI trade as retail piles into Nvidia, says Morgan Stanley's Savone

AI-generated summary Morgan Stanley's Nick Savone reports that hedge funds have reduced their exposure to the artificial intelligence trade to yearly lows. Savone states that individual investors have continued aggressively purchasing shares of Nvidia despite the broader market pullback.

Earlier: