> ## Content Index
> Fetch the complete content index at: https://www.ginc.org/llms.txt
> Use this file to discover other available public pages before exploring further.

# The Shift #1. Political Stability, Hormuz Infrastructure & Cyber Acceleration
- URL: https://www.ginc.org/the-shift-1-political-stability-hormuz-infrastructure-cyber-acceleration/
- Published: 2026-11-08T10:02:54.000Z
- Updated: 2026-11-08T10:02:54.000Z
- Description: The Shift #1: US midterms and persuadable voters, oil below US$100, the NCSC’s 2028 AI cyber warning and China’s new AI export controls.
- Author: Patrick Brothers
- Tags: Newsletter

Welcome to very first edition of 'The Shift'. Every Sunday, I'm sharing my personal notes on the major shifts playing out around the world and how they are shaping national capability and long-term power, prosperity and resilience.

Alongside the major developments of the week, I'll be drawing on the emerging datasets we're developing, most notably the [GINC Trends Index](https://trends.ginc.org/?ref=ginc.org) which offers weekly insight into the 250 trends we've identified and the weekly changes in trend attention and expectations around the world (in 14 languages). We'll bring more data-driven insights together each week from a broader range of datasets and indices as time goes by.

## Week in Brief  

### Information and Influence

US voters handed the Democrats [both chambers of Congress](https://www.nytimes.com/live/2026/11/05/us/midterm-elections-democrats-trump?ref=ginc.org) this week, flipping seven Senate seats including Texas and Kansas. The president retained his veto and, perhaps surprisingly, accepted the result on the night, as did the Republicans who conceded their seats.

Perhaps the most interesting reflection from the midterms is on the nature of democracies. We have talked for a long time about political polarisation, and the evidence for it is substantial. Yet this week is also a reminder that in this case, the US has a broad, persuadable electorate rather than a permanently sorted one. Biden won convincingly against Trump; Trump won convincingly against Biden; and in the midterms the Democrats pulled off an enormous blue wave. The US midterms prompts a rethink of how permanently organised voters really are, as opposed to how readily they respond to the issues that matter to them most. Capital (in this case political funding) tells the same story: in most races, picking the candidate without the funding would have picked the winner. That is an uncomfortable finding for the assumption that funding powers politics, and for the assumption that voters were sorted for good.

A second major reflection from the US midterms is one of policy volatility. The pattern is becoming increasingly familiar: a president drives major change through executive power for two years, only to see it disrupted two years later and what we think of as a two term limited, 8 year potential for strategic continuity is increasingly becoming a biennial micro-cycle. This also presents the kind of sovereign risk that partners and investors find difficult to plan for, while eroding any long-term planning assumptions. 

### Strategic Infrastructure

One immediate consequence of the midterms was oil falling nearly 20 per cent to [below US$100 a barrel](https://ca.investing.com/news/economy-news/markets-fret-over-growth-and-inflation-despite-easing-oil-prices-4937307?ref=ginc.org), on the result itself and on Chuck Schumer’s declaration that the war must end. In the same week came [further ship attacks](https://www.cbsnews.com/live-updates/iran-war-us-trump-midterms-strait-of-hormuz/?ref=ginc.org) and an Omani proposal for joint Iran–Oman control of Hormuz traffic.

Glass half full, we see a cash-starved regime without a Supreme Leader that has been pushed towards a real nuclear freeze, and a midterm result that may give Trump the exit ramp he has been looking for. Glass half empty, nothing has really changed for Iran since September, and stability on Oman’s terms, with a toll booth on the strait, may prove quite attractive to Tehran. 

### Critical Technology

The UK's National Cyber Security Centre, part of GCHQ, released its [tenth annual review](https://www.ncsc.gov.uk/collection/annual-review-2026?ref=ginc.org), counting more than 1,800 attacks in the first ten months of 2026, 80 of them nationally significant. Most interestingly, is its [assessment of AI and the cyber threat](https://www.ncsc.gov.uk/collection/annual-review-2026/understanding-threat/ai-impact-cyber-threat?ref=ginc.org), which projects [2028](https://www.ncsc.gov.uk/collection/annual-review-2026/understanding-threat/autonomous-attack?ref=ginc.org) as the year in which AI coordinates attacks reaching robots and cloud-controlled infrastructure. Meanwhile a [commercial intrusion market](https://www.ncsc.gov.uk/collection/annual-review-2026/understanding-threat/cyber-threat-uk?ref=ginc.org) lets states buy capability off the shelf rather than build it. AI is making attacks faster and more common rather than new in kind. Defenders increasingly know what to do, and frontier models now ship to defenders first. 

## GINC Trends Index. Week 35, 2026

[GINC Trends Index. Week 36, 2026Week 36 of the GINC Trends Index: midterms and the AI sell-off drove attention, European security followed, and expectations of political volatility fell.![](https://storage.ghost.io/c/21/72/21722704-b926-4123-9487-66da96e944be/content/images/icon/Global-Institute-of-National-Capability--GINC--logo-62a07e55-7dc7-418e-812e-9526ece0c6c2.png)Global Institute for National CapabilityPatrick Brothers![](https://storage.ghost.io/c/21/72/21722704-b926-4123-9487-66da96e944be/content/images/thumbnail/ginc-trends-bento-wave-header@2x-eca7d613-ef97-4f82-aa1c-cd46242cb07f.png)](https://www.ginc.org/ginc-trends-index-week-36-2026/)

### Attention Index

[**Commercial spyware**](https://trends.ginc.org/trends/commercial-spyware/?ref=ginc.org) **was the largest driver, up 30%.** Apple warned iPhone users in 99 countries of mercenary spyware ([Techcrunch](https://techcrunch.com/2026/11/02/germany-says-lawmakers-iphones-may-have-been-hacked-with-spyware/?ref=ginc.org)), Germany confirmed that Bundestag members and ministers were among them; the attacker unknown ([Euronews](https://www.euronews.com/2026/11/02/german-mps-and-ministers-warned-iphones-have-been-targeted-by-spyware?ref=ginc.org)). The rise came from the [UK](https://trends.ginc.org/region/united-kingdom/?ref=ginc.org), up 138%, and the US, up 40%, but not from [Germany](https://trends.ginc.org/region/germany/?ref=ginc.org), where the trend already stands at eleven times typical trend levels.

[**Wealth concentration**](https://trends.ginc.org/trends/wealth-concentration/?ref=ginc.org)**, up 26% rising from 212 to 246**; in the US it nearly doubled**.** California voted on Proposition 40, the first US billionaire tax. It was previously failing 54 to 46 on partial returns after billionaires gave $333 million against it ([The Hill](https://thehill.com/business/6185092-california-voters-reject-wealth-tax/?ref=ginc.org)). Elon Musk lost about $113 billion in two days, and Bill Gates backed a wealth tax ([Fortune](https://fortune.com/2026/11/06/bill-gates-tax-wealthy-billionaires-ai-inequality?ref=ginc.org)).

[**AI capital concentration**](https://trends.ginc.org/trends/capital-concentration-in-ai/?ref=ginc.org)**, up 20%, after 67% last week.** Anthropic's IPO valuation expected to fall to $1.1-1.3 trillion from \~$1.5 trillion after the sell-off ([Bloomberg](https://www.bloomberg.com/news/articles/2026-11-06/anthropic-ipo-valuation-expectations-pressured-after-ai-rout?ref=ginc.org)). OpenAI lowered its own to \~$1.25 trillion, and Ramp data showed company AI spending flattening ([TechCrunch](https://techcrunch.com/2026/11/06/ramp-data-shows-ai-spending-stalling-just-as-anthropic-nears-its-ipo?ref=ginc.org)). The index rose from 344 to 407, now the 25th highest of the 250\. [India](https://trends.ginc.org/region/india/?ref=ginc.org) rose 33 per cent, the UK 25, [France](https://trends.ginc.org/region/france/?ref=ginc.org) 24.

### Expectations Index

**Wealth and emergency powers: a lot of attention, but not expected to grow.** The expectation that [wealth concentration](https://trends.ginc.org/trends/wealth-concentration/?ref=ginc.org) gets worse has fallen from 47% to 27% in four weeks. The expectation that [executive overreach](https://trends.ginc.org/trends/executive-overreach-and-emergency-powers/?ref=ginc.org) grows has fallen from 46% to 28%. Both were in the news all week, yet the index expects California's vote and the midterms to hold both back.

**The stock market fall was not a surprise.** The expectation that [asset valuations](https://trends.ginc.org/trends/stretched-asset-valuations/?ref=ginc.org) stay stretched is 77%, one of the highest in the index, and it did not change in the week shares fell. In simple terms, the fall was expected. The expectation that [AI capital concentration](https://trends.ginc.org/trends/capital-concentration-in-ai/?ref=ginc.org) grows is 58%, down from 66% a month ago, after Anthropic's valuation has likely fallen in line.

**AI data centres and science: yes. AI agents: no.** The expectation that [AI for science](https://trends.ginc.org/trends/ai-accelerated-science/?ref=ginc.org) grows is 80%, the highest of any trend. [Open-weight AI models](https://trends.ginc.org/trends/open-weight-ai-diffusion/?ref=ginc.org) are also at 80% and [AI data centre build-out](https://trends.ginc.org/trends/compute-scaling-and-ai-infrastructure/?ref=ginc.org) at 73%. The expectation that [AI agents](https://trends.ginc.org/trends/ai-agents/?ref=ginc.org) take off has fallen from 44% to 17% in four weeks. [AI job displacement](https://trends.ginc.org/trends/ai-and-knowledge-work-automation/?ref=ginc.org) is also at 17%: the expectation index does not expect AI to take many jobs soon.

## China's National AI Strategy

Quite recently we've seen China expand export controls on frontier AI. On 16 October Beijing proposed adding 23 AI technologies to its export-control list, including large language models, agents, chip designs and computing clusters, with model weights and source code above set performance thresholds restricted and models trained on China’s ‘core data’ banned from export altogether. A temporary registration requirement took effect the same day. It is a reversal for a country whose global position was built on open weights, and a mirror of US chip controls. A week later DeepSeek and MiniMax were cleared as the first vendors licensed to open-source and sell models overseas, suggesting China will likely run as managed export rather than blanket prohibition.

And just this week, the NDRC said it was accelerating legislation to govern AI across the entire model life cycle, from data to compute, covering developers from DeepSeek to Moonshot and Alibaba. The move follows a SemiAnalysis finding that only 3.6 per cent of 857 Chinese model releases since 2021 shipped safety evaluations, and extends the CAC’s content-and-licensing regime into compute, data and loss-of-control risk. Both policies land against a sharp correction: the STAR50 index of AI and semiconductor leaders has fallen 40 per cent in under four months, erasing some US$480 billion in value.

We're preparing our annual update of the most popular GINC analysis by far, [China's National AI Strategy](https://www.ginc.org/chinas-national-ai-strategy/).

[China's National AI Strategy | GINCChina’s AI push has shifted from catching up at home to setting the rules abroad, melding chip self-reliance, state capital, and Digital-Silk-Road diplomacy into a single strategic engine.![](https://storage.ghost.io/c/21/72/21722704-b926-4123-9487-66da96e944be/content/images/icon/Global-Institute-of-National-Capability--GINC--logo-bb6f0bb8-e0fd-44e8-abea-50056ed46a2b.png)Global Institute for National CapabilityGINC![](https://storage.ghost.io/c/21/72/21722704-b926-4123-9487-66da96e944be/content/images/thumbnail/iStock-683984956-4ae7d9e0-a47b-49b3-b813-fb41b8e89def.jpg)](https://www.ginc.org/chinas-national-ai-strategy/)

A brief summary of the strategy as at January 2026 is below and the note above has the full analysis.  
  
**Sovereign, state-led, 2030.** The State Council’s 2017 New Generation AI Development Plan set three phases: catch up by 2020, global competitiveness in core technologies by 2025, and primary global AI innovation centre by 2030\. The target has been restated in the 14th Five-Year Plan and in Politburo study sessions. AI is treated as a mission of state rather than a technology left to the market. Capital follows: a ¥60 billion National AI Industry Investment Fund launched in 2024, alongside the semiconductor ‘Big Fund’, provincial guidance funds and compute vouchers. The stack is being rebuilt from the ground up, from Huawei’s Ascend and Baidu’s Kunlun chips to the PaddlePaddle and MindSpore frameworks and more than 20 CAC-cleared language models. Fabrication remains the weak link, since SMIC lacks EUV lithography.

**Centralised design, decentralised execution.** Strategy is set by the State Council and divided among ministries: MOST for research, MIIT for industry and standards, the CAC for algorithms, content and data. Between the ministries and the firms sit the 200-member AI Industry Alliance and the Open Innovation Platform programme, which assigns national champions by domain: Baidu for autonomous driving, Tencent for medical imaging, Alibaba for smart cities. Below that, more than 20 pilot zones in cities such as Beijing, Shanghai, Shenzhen and Hangzhou operate as regulatory sandboxes, and their results feed back into national rules. This generally delivers regulation ahead of deployment: algorithm registration in 2022, deep synthesis rules in 2023, generative AI licensing within months of ChatGPT and mandatory labelling of synthetic content from September 2025.

**Resilience, oversight and unity.** US export controls turned ‘indigenous controllability’ into a central theme, directing state funds to ‘stranglehold technologies’, accelerating domestic chip designers and spreading compute hubs across cheaper western provinces. Ethical oversight runs through MOST’s Ethical Norms, BAAI’s Beijing AI Principles and emerging audit standards, with the firm requirement that generative models align with ‘core socialist values’ and watermark their outputs. AI is pushed into public services by policy: City Brain traffic management in Hangzhou, diagnostic tools in township clinics, AI tutors in low-income schools. Multilingual tools trained on Uyghur, Tibetan and Mongolian serve courts and hospitals in ethnic regions, framed as inclusion while extending state reach. Public trust is cultivated through state media, visible regulatory interventions and draft rules published for comment: guided participation that responds to public concern without ceding control of the direction.