Read aloud by our house voice. If it is not available, your device's voice takes over.
Beta · Answers only from Globalance Intelligence texts, with source. No investment advice. Questions are stored for 90 days without personal data to check the answers.
What would you like to understand?
All 0
Type at least two characters. The search covers only our own pages.
↵ Open↑↓ NavigateESC Close
HOW WE WORK
How Globalance Intelligence works
Artificial intelligence is changing how knowledge is produced. It compresses in seconds what used to take hours. But it can also be convincingly wrong. For a platform that informs decisions on the future and on investment, reliability is therefore not a feature but the foundation.
Human and machine: clear accountability
AI is used where it is strong: scanning, structuring, and condensing large volumes of information. Responsibility for what appears under our name rests with a person. Two levels apply.
What we publish
Newsletters and lead articles pass through the full editorial chain: Source attribution, a automated fact-checking and a human sign-off. We put our name to this content.
What the platform produces through analysis
Scenarios, early indicators, and the ongoing research stream are AI-synthesised guidance. They draw on real sources and data but are treated as hypotheses and early indications, not as conclusively verified facts. This layer is monitored systematically rather than approved individually.
Why Anthropic
Anthropic was a deliberate choice. The company is among the world's leading voices on safety, ethics and foresight in AI development. Those values align with the standards expected of a responsible private bank. A model that names its own limits is worth more to a platform like this than one that has an answer to every question.
Its legal form fits this standard too. Anthropic is a public benefit corporation under Delaware law, a legal form that grew out of the B Corp movement. Its board must weigh shareholders’ interests against the well-being of those affected and a purpose set out in its charter: “the responsible development and maintenance of advanced AI for the long-term benefit of humanity”. The planned stock market listing does not change this. According to the preliminary prospectus of 26 October 2026, a trust of five trustees with no financial stake elects the majority of the board. 5% of the shares sit in a company-owned fund for the public good, and the founders have pledged 80% of their shares to philanthropy.
This is no guarantee. The legal form requires a balancing of interests, not that the public good comes first. And new shareholders receive a share of the profits but little say: through a separate class of shares, the founders hold 50.1% of the votes. We are therefore watching whether the structure delivers what it promises. The same standard applies as for the Globalance AI Footprint: evidence counts, not declarations of intent.
Scenario or hallucination?
Don't AI models hallucinate? They do. That is why two things are kept strictly separate: the Factual error (which we counter through source verification and fact-checking) and the Imagination (which every scenario exercise requires). A scenario is disciplined imagination: grounded, labelled as a hypothesis, and testable against early indicators.
TRUST & TRANSPARENCY
Why AI's apparent weakness in futures work is its greatest strength, and how we use it.
An objection we sometimes hear: "AI models hallucinate. How can I trust the scenarios at all?" It is a fair question. The answer depends on what one means by 'hallucination'. Behind that word lie two entirely different things, and the two must not be conflated.
Two kinds of "hallucination"
First: the error. The model asserts something false about the present or the past — an invented figure or a fabricated source. This is a flaw, and a dangerous one: the AI's linguistic elegance makes a wrong statement sound convincing.
Second: imagination. The model constructs coherent, plausible worlds that do not yet exist. That is not an error. That is the work of scenario-building. In this sense, a scenario is itself a kind of hallucination — only a disciplined one.
What distinguishes a scenario from a hallucination
The creative act is the same. The difference lies in three things:
Grounding. A scenario is anchored to real drivers: demographics, productivity, capital flows. A hallucination floats free.
Labelling. A scenario states openly: "This is a hypothesis about the future." The pathology of hallucination is precisely that it disguises itself as fact.
Accountability. Every scenario comes with early indicators (signposts). These are observable signals that show whether reality is moving towards it. A hallucination has none.
A hallucination is a scenario without accountability. A scenario is a hallucination with grounding, a label, and early indicators.
Why AI's imagination is an advantage
When designing futures, machine imagination is a genuine lever. In minutes, AI generates many plausible futures, including ones that are far from obvious. It combines drivers that a human rarely thinks of together. And it breaks our strongest blind spot: the tendency to see the future as merely an extension of the present. Since the pioneers of scenario planning (Shell, Pierre Wack), the aim was never to predict correctly, but to think rigorously in order to be prepared.
Where we take particular care
Facts are where things become dangerous. More subtle is the illusion of coherence: plausible-sounding chains of cause and effect that feel internally consistent but may rest on a false assumption. The guiding principle: plausible is not the same as probable.
How we address this at Globalance Intelligence
Imagination is not switched off. It is fenced in. The architecture rigorously separates the two modes:
Grounding, not invention: Facts are drawn from curated sources and real data, not from the model's unconstrained output.
Plurality over single-point certainty: always multiple scenarios with explicit probabilities, never the false certainty of a single answer.
Early indicators rather than assertion: every scenario is equipped with early indicators that we continuously track against real-world developments.
Human oversight and labelling: Contextualisation and scrutiny remain with the human; what is hypothesis is clearly identifiable as such.
The bottom line
The question is not whether AI hallucinates. It does. The question is whether imagination can be used and fenced in at the same time. With facts, hallucination is a risk to guard against. With futures, controlled imagination is precisely what good scenario work requires. The difference does not lie in the model. It lies in the architecture that separates the two.