I am a very happy personal user of AI. I have aligned myself, for now, with Anthropic. In early June I was enjoying using Fable 5, until it was abruptly and I mean very abruptly, disabled. (This was sharp - what was there the night before was gone the next morning).
The US government, citing national security authorities, has issued an export control directive to suspend all access to Fable 5 and Mythos 5 by any foreign national, whether inside or outside the United States, including foreign national Anthropic employees. The net effect of this order is that we must abruptly disable Fable 5 and Mythos 5 for all our customers to ensure compliance.
https://www.anthropic.com/news/fable-mythos-access
Anthropic was not alone, OpenAI soon after imposed similar restrictions.
The Economist was quick to pick up the thread of sovereign risk in AI models.
(sorry, for non-subscribers, this link is behind a pay wall:
https://www.economist.com/international/2026/07/16/sovereign-ai-independent-of-america-and-china-is-a-pipe-dream)
The European Union currently relies on non-EU countries for over 80% of key digital products, services, infrastructure, and intellectual property. The EU has also seen the need to foster home grown AI models as part of the EU Digital Strategy.
(https://digital-strategy.ec.europa.eu/en/policies/eu-tech-sovereignty)
When looking at sovereignty, you can begin to unpack the AI layers.
Hardware: Who makes the chips? (e.g. Nvidia)
Software: Who provides open source LLM's?
Infrastructure: Who powers and connects the grid to the data centres?
Operators: Where are the engineers to implement and operate the systems?
At issue, is that a sovereign intervention at any of these layers (e.g. trade sanction, executive order etc.) can act as a 'kill switch'. For the avoidance of doubt, this is a kill switch, with the same intent as MiFID Article 48....but with a completely different interpretation and application.
As Sarah Breeden, BoE Deputy Governor for Financial Stability observed..."whether guardrails are needed, analogous to circuit breakers or kill switches that would limit or stop trading market-wide if faulty AI models cause market meltdown."
https://www.bankofengland.co.uk/speech/2026/june/sarah-breeden-panel-at-the-european-central-bank-forum-on-central-banking-2026
Sovereignty doesn't require reinventing every layer of the stack from scratch. A common denominator — be it shared standards, shared hardware even — can sit underneath, while each region retains its own implementation, data residency, and control over the AI model (and indeed kill switch).
Hopefully my Venn diagram makes sense.
a) There are many models. In my personal experience, the models require training, and saving (so the lessons are not lost / forgotten / purged). Models can hallucinate, report to please, act selfishly or maybe even act in a herd mentality (requiring a kill switch).
b) These models all rely on the data sets they are fed. As it ever was, poor data in will lead to bad results out. Getting you data in good order is a problem that is not going away.
c) AI makes very fast and at times very convincing arguments. But, fluent and confident when the response is wrong or fabricated is dangerous. Some things, like sentimental value, unless taught to a model, may lose all meaning.
Your portfolio may have an element of physical gold, part of an inheritance, that is a non-negotiable hold. AI, unless informed otherwise, could make a very rational position to re-balance that position.
We need AI. We also need intuition, sentiment, process and control. A human in the loop.
No comments:
Post a Comment