1 of 14

WHAT MESSI, LONDON UNDERGROUND AND FORMULA ONE CAN TEACH US ABOUT TRUST IN AI

Trust, Transformed

Anthony Widdop

27 June 2026

LLM × Law Conference

Judge Business School, Cambridge

2 of 14

Who wants AI you �can fully trust?

Who fully trusts the processes we have today?

3 of 14

We are asking the wrong question

We hold AI to a higher standard than we hold ourselves

“Can we trust AI?” is too narrow a question

It is more nuanced than an either / or consideration

It is humans OR AI or humans AND AI depending on the task and context

The jagged frontier means that trust in AI will continue to be task-specific until a step change in performance + ease of verification

4 of 14

Trust in AI is starting from a low point in developed countries…

Source: Edelman Trust Barometer Flash Poll: Trust and Artificial Intelligence at a Crossroads (November 2025)

5 of 14

…and AI is inheriting a trust deficit

Source: 2026 Edelman Trust Barometer: Trust Index

6 of 14

Mind the gap

As AI advances, will we continue to see a gap between what it can do and what we trust it to do?

Moves at the pace of tech

Advancing faster than our ability to fully understand, govern or trust it

THE GAP

What AI can produce vs. what we can verify

ORGANISATIONS

Move at the pace of people

Every move up the s-curve resets the trust clock

7 of 14

But can we fully trust humans?

HUMAN FAILURE

High tolerance

MACHINE FAILURE

Near-zero tolerance

We are forgiving of legacy processes long considered part of the fabric of how work gets done

Rather than comparing against perfection, we must (re)calibrate our risk tolerance for this new world of work

We measure AI (e.g., via evals); we rarely measure human performance in the same way

8 of 14

Start from a blank page

Don’t integrate AI on top of legacy processes

Re-evaluate the performance of those processes

This is the opposite of incremental improvement and a marginal-gains mindset

The real question: do we still need that process in 2026?

Artificial Lawyer. May 2024

9 of 14

https://trust-transformed.netlify.app/

10 of 14

Everyone has, or will have, access to similar AI models

The winners will be those who know where and where not to deploy them

In this way trust becomes a key asset

To a regulator or academic, trust is a public good and not just relevant to private organisations

Trust can become a moat…

11 of 14

...and a core business capability to drive transformation

“Organizations that treat AI trust as a core business capability, rather than as a compliance requirement, are better positioned to scale AI adoption to its full potential.”

McKinsey & Company: State of AI trust in 2026, Shifting to the agentic era (March 2026).

McKinsey: State of AI trust in 2026

12 of 14

Whose job is it to build trust in AI? Everyone's.

Law firms

Communicate where you use AI, where you don't, and why

Consider the full cost of AI: creation, oversight, review, rework and �sign-off

Invest in differentiated training and upskilling: building universal AI fluency lifts trust over time

Clients

Legal tech vendors

Learn from experience: run your chosen AI solutions against a selection of completed matters

Be clear on where AI is welcome, where it isn’t, and what you want disclosed when working with outside counsel

Build for ease of verification, not just generation

Design for the average user, not the superuser: build guardrails and checkpoints around usage

Engineer traceability by design to support audit trails

13 of 14

Remember to always trust your brakes!

It’s easy to assume that an F1 car can drive �fast because of its superior engine

But counter-intuitively, it's all in the brakes

An F1 car can commit to the straight at maximum speed because the driver has complete trust in their ability to brake late