Computing Machinery and Intelligence
Turing replaced “can machines think?” with a test of behaviour. The board version of his question is still open: when a system behaves convincingly, what have we actually learnt about whether to trust it?
Foundations
Ten documents on artificial intelligence that every board director should know. The test for inclusion is simple: would a director be worse at the job for not having read it? For each, there is a suggestion of where to start, because few boards have time to read them all in full.
These are foundations, not authorities to agree with. Their disagreements are part of their value.
How it works
Attention Is All You Need
The one technical paper on the list. It introduced the Transformer, the architecture behind today’s large language models. A board needs to know that it exists and why it mattered, not how it works.
On the Opportunities and Risks of Foundation Models
Explains how a few very large models came to sit underneath many different products. The best starting point on concentration and dependency: if one model fails or changes, everything built on it is affected.
What it means
On the Dangers of Stochastic Parrots: Can Language Models Be Too Big?
The essential sceptic. Fluent language is not the same as understanding, and people readily attribute meaning to machine output. A useful corrective for any board being shown an impressive demonstration.
AI as Normal Technology
Argues that AI’s effects will arrive through institutions, adoption and diffusion, like earlier general-purpose technologies, rather than as a sudden break. Closest to how a board should actually think about it.
Magnifica Humanitas: On Safeguarding the Human Person in the Time of Artificial Intelligence
It opens with a choice between building “a new Tower of Babel” and building a shared city. Readers need not share its theology to take its central point: progress needs an account of the people it is meant to serve, and of who holds the power it creates.
How to govern it
Artificial Intelligence Risk Management Framework (AI RMF 1.0)
A practical structure (govern, map, measure, manage) that turns principles into decisions. Useful as a checklist of what a board should expect management to be able to show.
Artificial Intelligence Act, Regulation (EU) 2024/1689
Where regulators have drawn the lines. Even for firms outside the EU, it is becoming the reference point for what counts as high-risk use.
The Financial Stability Implications of Artificial Intelligence
The financial-services view: third-party dependency and concentration, correlated behaviour, cyber risk and model risk. For a financial services board, the most directly relevant document on this list.
International AI Safety Report 2026
The broadest evidence-based assessment of what general-purpose AI can now do and where the risks lie. It names a dilemma boards share: needing to act before the evidence is complete.
Further reading
Worth knowing about, but background, specialist or covered by the ten above.
- Deep Learning The approach that preceded today’s language models. Nature may ask for a subscription.
- A Proposal for the Dartmouth Summer Research Project on Artificial Intelligence The document that named the field, and an early lesson in confident forecasting.
- Language Models are Few-Shot Learners The GPT-3 paper: one model, many tasks, from instructions alone.
- Scaling Laws for Neural Language Models Why bigger became the strategy, and the investment case that followed.
- Training Language Models to Follow Instructions with Human Feedback How human judgements shape a model’s behaviour, and whose preferences get encoded.
- Model Cards for Model Reporting What documentation a board should expect for a model it relies on.
- AI Agents That Matter Whether agent benchmarks measure the reliability that matters in real use.
- AI Principles The compact international reference point for trustworthy AI.
- Recommendation on the Ethics of Artificial Intelligence The international statement of the values at stake, beyond those of the companies building the technology.
- Antiqua et Nova The Vatican’s detailed note on artificial and human intelligence, and the natural companion to Magnifica Humanitas.