AISOF
Home

Insights

AI adoption and risk in numbers

Figures on adoption, risk, liability and spending from published research. Each figure links to its source, and estimates carry uncertainty.

These figures are published by third parties, including McKinsey, OpenAI, Gartner and industry surveys. They are collected here to give boards and product leaders a concise view of AI adoption and risk.

560,000

ChatGPT users a week may show possible signs of psychosis or mania, about 0.07% of weekly users (estimate)

Source

OpenAI estimates, Oct 2025

0.15%

of weekly ChatGPT users have conversations with explicit indicators of suicidal planning or intent (estimate)

Source

OpenAI, Oct 2025

What the numbers imply

Adoption outruns control

Most organisations now use AI somewhere. Far fewer can inventory it, name owners, or show independent evidence of oversight. That gap is a governance problem for boards.

Boardroom reviewing AI strategy

Shadow AI is an inventory problem

When staff use personal tools with company data, the organisation has AI in use that it cannot see. AISOF-3300 starts with knowing what is in use and who is accountable.

Abstract inventory of untracked AI tools as incomplete connected grid

Absolute harm at scale

Percentages that look small become large when user bases reach hundreds of millions. Product makers in high stakes domains need more than a model card.

Healthcare context for high stakes AI

How to read this

Percentages that look small (for example 0.07%) become large in absolute terms when user bases reach hundreds of millions. Forecasts from Gartner and others are not certainties. They are signals for risk planning. Always open the original source before citing a figure in a board paper.

For narrative cases and public reporting, see News & risk briefings. For how AISOF turns awareness into assessable controls, see How it works.

Teams building competence around AI risk

From statistics to controls

Assessment packs and the AISOF frameworks turn these risks into assessable controls.