Tonight's editorial picture: the cost of building faster than we can check
Artificial intelligence, energy and a hillside fire in Bukavu look unrelated. Read together, they describe a world building at a speed its safeguards have not matched.

Three of tonight''s stories are about capacity. A researcher leaves an AI laboratory saying the checks cannot keep up with the releases. Oracle books $664bn of contracted demand for computing power that does not yet exist. The Bank for International Settlements looks at who is lending against it and asks whether anyone can see the whole picture.
Two more are about the cost of not building. Gilt yields at pre-crisis levels are the price of borrowing when confidence is thin. A fire in Bukavu killed 24 children in housing that no fire engine could reach, on a hillside where water supply is intermittent.
The link is not moral equivalence. It is sequence. In each case the construction ran ahead of the safeguard - the evaluation, the disclosure, the fire road, the water main - and the gap between the two is where the harm sits.
That is a reasonable test to apply to the next twelve months of artificial intelligence. Not whether the systems are capable, which they plainly are, but whether the institutions meant to check them have been resourced at the same rate.
Truth. Discernment. Impact.
This is a Berean News editorial. It reflects the view of the newsroom and is clearly separated from our reporting.
Corrections & updates
- Story published
Verified against the sources cited in this report.
Sep 10, 2026, 7:00 PM
Related stories

"AI could kill us all" — insider quits with warning
A senior safety researcher has left one of the leading artificial intelligence laboratories, saying internal safeguards are not keeping pace with the speed at which the most capable systems are being deployed.

UK borrowing costs hit levels not seen since before financial crisis
The ten-year gilt yield closed at 5.378 per cent, tightening the fiscal squeeze ahead of the autumn budget.

Oracle revenue jumps 30% as AI cloud backlog reaches $664 billion
Contracted future revenue from artificial intelligence infrastructure has grown to a scale that now shapes the whole cloud market - and raises questions about how much of it converts.