top of page

AI newsletter November 2025

Welcome to the Dixon AI Newsletter which summarises selected AI stories and releases you might have missed. This issue covers AI news from 1st October to 21st November 2025.

The AI Macro Environment and Enterprise Insights

  • EU Softens Stance on AI and Data Regulation: Under industry and US pressure, the European Commission moved to water down key tech laws. Proposed changes would loosen GDPR data sharing rules and delay parts of the EU AI Act, allowing AI firms to train on personal data if other GDPR requirements are met. It’s a notable retreat by Brussels after years of championing strict digital regulation. See (theverge.com).

  • White House Targets State AI Laws: In Washington, President Trump is poised to pre-empt strict state AI regulations via executive order. A draft order would create a federal AI Litigation Task Force empowered to sue states like California and Colorado over laws deemed to “obstruct” AI industry growth. This aggressive federal stance aims to override patchwork state rules and assert national control over AI governance. See (theverge.com).

  • Amazon Sues Perplexity with implications for AI Agents E-commerce giant Amazon filed suit against Perplexity, alleging its AI-powered Comet browser illegally automates shopping by covertly accessing user Amazon accounts and masking bot traffic as human. The clash underscores a deeper fight over autonomous AI agents: are they bots posing security risks, or is Big Tech just bullying a new competitor? The case highlights an emerging debate on regulating AI Agents acting on our behalf. (theguardian.com).

  • Nvidia’s AI Chip Demand Shatters Records: Nvidia’s latest earnings reveal unprecedented demand for AI hardware, calming fears of an ‘AI bubble’. The company booked a record $57 billion in quarterly revenue, including $51 billion from data center AI chips, up 66% year-on-year. CEO Jensen Huang said every AI server GPU is selling as fast as Nvidia can make them, with cloud providers reporting some chips “sold out” amid an insatiable appetite for AI compute. (bbc.co.uk).

  • AI-orchestrated cyber campaign: Anthropic says it disrupted what it believes is the first large-scale, largely autonomous AI-orchestrated cyber-espionage campaign, and announced that we’re reached an inflection point at which AI models have become genuinely useful for cybersecurity operations, both for good and for ill. (anthropic.com).

New and updated AI model and application releases

  • OpenAI 

  • Google

  • Microsoft

    • Anthropic (Claude) models in Copilot: The Claude models are now available in Copilot through admin settings at an organisation level - see Anthropic joins Copilot 

    • Project Opal: Microsoft released a new AI-powered capability that executes task-based work. Available under Frontier in admin settings at an organisational level. Introducing Project Opal 

    • Word, Excel, and PowerPoint Agents in M365 Copilot: Microsoft are releasing new Frontier agents in Copliot which make it easier to create documents, spreadsheets, and presentations.

    • Voice and vision added to Copilot chat: Voice mode and vision have been added to the Copilot app, enabling users to work more productively on the go. Voice in Copilot 

    • October updates: There were many updates in October and for a full list of what was new in Copilot in October please see October what's new in Copilot

  • Anthropic


  • Notable Other Model Releases

    • Grok 4.1: X released Grok 4.1 and Grok 4.1-thinking, which held top spot on the LM leaderboard for 24hours before Gemini 3 was released. The Grok 4.1 models score particularly highly on creative writing and emotional intelligence benchmarks https://x.ai/news/grok-4-1 . 

    • Kimi K2: Moonshot released the Chinese model Kimi K2, which contains the most capable open-source thinking model. https://moonshotai.github.io/Kimi-K2/thinking.html 

    • Mistral AI Studio: MIstral, owners of the leading European model, released the MIstral AI Studio for building AI tools. https://mistral.ai/news/ai-studio 

  • Notable Application Releases

    • Perplexity Comet: Perplexity’s agentic browser was rolled out widely at  the beginning of October, providing users with a  powerful personal AI assistant within their browsers. Comet is now also available to download on Android with iOS due to follow soon. https://www.perplexity.ai/comet 

    • Synthesia 3: Synthesia released Synthesia 3.0, enabling users to create fully interactive avatars with natural hand and body movements. https://www.synthesia.io/3 

    • Manus 1.5: Manus released Manus 1.5, their most capable agent system with increased speed, reliability, and quality. https://manus.im/blog/manus-1.5-release 

Recommended Reading, Listening and Book Releases

  • For anybody:

    • Podcast series: The Last Invention. An entertaining and informative 8-part series about the AI Revolution. The Last Invention 

  • For the more technically minded and developers:

    • Podcast: In this episode of the Dwarkesh Podcast, Andrej Karpathy argues AGI and reliable agents are about a decade away, needing multimodality, continual learning, and better-than-RL training.https://www.dwarkesh.com/p/andrej-karpathy 

  • For leaders:

    • Book: Governing the Machine by Eitel-Porter et al. A practical guide for business leaders to Responsible AI and AI Governance. Governing the Machine

Forthcoming from Dixon AI

  • AI Drop In: one-hour, live session held three times a week. Join here: https://www.dixonai.com/ai-drop-in 

    • w/c 24th Nov: ChatGPT on mobile

    • w/c 1st Dec: Prompts vs Bots

    • w/c 8th Dec: AI for Document Insights

A word from Rob

In many of the organisations I’ve been speaking with recently, a familiar pattern seems to be emerging. Quite often, businesses are using AI primarily to automate tasks and improve return on investment. And in many cases, these efforts end up being owned or driven by the IT department. While IT has an important role to play, this setup can create a subtle but significant misalignment.

From what I’m seeing, AI works best when it is understood as a commercial transformation rather than a technical one. The area that needs to shift first is usually the business model, not the technology stack. That means commercial teams need to become increasingly AI literate: leaders, product teams, operations, customer-facing roles and decision-makers all need the capability to recognise new forms of value and adapt how the organisation competes.

In our work with clients at Dixon AI, we’ve been placing much more emphasis on helping businesses maximise their rate of innovation per unit time. The external environment is moving quickly, and the month’s news highlights just how rapidly new capabilities are emerging. Against that backdrop, the biggest risk I’m seeing is organisations focusing narrowly on isolated ROI projects and, in doing so, inadvertently optimising for the old world.

My advice is this: before investing heavily in automation, focus on building your organisation’s innovation engine. Upskill your commercial teams. Give leaders direct experience with the tools so they can understand what is now possible. Create the conditions for continuous change. Once the business model is evolving and adapting, automation becomes far more valuable and far less risky.


Related Posts

See All
AI Newsletter January 2026

This edition of the Dixon AI Newsletter covers the most important AI developments from mid-December to late January, including Apple’s partnership with Google, OpenAI’s move into advertising, rising g

 
 
 
AI Newsletter December 2025

Welcome to the Dixon AI Newsletter which summarises selected AI stories and releases you might have missed. This issue covers AI news from 22nd November  to 15th December 2025 . The AI Macro Environme

 
 
 

Comments

bottom of page