GAZA

Famine descends, the images break through, and Western opinion finally turns

After 21 months of war and blockade, mass starvation arrived in Gaza. The images of emaciated children, at last, pierced the world's exhaustion, and Western governments that had stood by began, haltingly, to move: France said it would recognise a Palestinian state.

The catastrophe in Gaza reached its most harrowing phase this fortnight as mass starvation took hold across the territory. After 21 months of war, blockade, displacement and the collapse of the aid system, malnutrition spread and deaths from hunger mounted, and the images, of skeletal children, of parents with nothing to feed them, of crowds crushed and shot at the few, chaotic aid-distribution points, finally broke through the world's long exhaustion with the conflict in a way that months of casualty statistics had not. Aid agencies and much of the international press described a famine, or its imminent arrival; Israel disputed the characterisation and blamed Hamas and the aid agencies, while facing the sharpest criticism from its allies of the entire war.

And the political dam, held for so long, began to crack. France announced on July 24 that it would formally recognise a Palestinian state, becoming the most significant Western power to do so and breaking with the American position, as pressure mounted on Britain and others to follow. Governments that had stood by Israel through 21 months of war found the images of starving children impossible to defend against, and the ground of Western opinion, which had shifted slowly throughout the conflict, moved sharply this fortnight. It was, in a terrible way, a demonstration of something this paper has noted about how the world attends to suffering: that statistics and warnings, however grave, move people far less than a single unbearable image, and that the famine had been forecast and warned of for many months by aid agencies whose reports went unheeded, until the children were visibly starving and the pictures could no longer be looked away from. The warning had been there for months. It took the images to make it real.

THE ECONOMY

Trump strikes trade deals with Europe and Japan, on his terms

The tariff war lurched toward a resolution that ratified the president's coercive method. Facing the August 1 deadline for his threatened punitive tariffs, major economies chose to settle: Japan agreed a deal on July 22, and the European Union on July 27, each accepting a baseline 15 percent tariff on their exports to the United States, along with pledges of vast investment, in exchange for avoiding the far steeper levies Trump had menaced. The deals were widely seen as capitulations, the world's other economic giants submitting to a tariff regime they had called unacceptable, because the alternative Trump offered was worse. The coercion had worked, at least in the narrow sense: the threats had extracted concessions, teaching every other nation that the way to deal with this administration was to pay.

IN BRIEF

An AI plan; a border war; a rocker's farewell

The administration released its "AI Action Plan" on July 23, a sweeping strategy to win the AI race against China through aggressive deregulation, the fast-tracking of data centres and the power plants to feed them on federal land, and the export of American AI abroad, stripping away the guardrails of the previous administration in favour of speed and dominance. Fighting broke out along the disputed border between Thailand and Cambodia on July 24, the deadliest clashes between the two neighbours in over a decade. And the heavy-metal pioneer Ozzy Osbourne died on July 22, weeks after a farewell concert, at 76.

The Column

Fast enough to lose control of

The administration's AI strategy this fortnight chose, explicitly, speed over safety: deregulate, build, win the race with China. The choice is usually argued as a values question. It is better understood as a trap, one that every powerful technology sets, and that this one sets more sharply than any before.

By Daniel Reeve

The administration unveiled an artificial-intelligence strategy this fortnight that made its priorities explicit: deregulate, build the data centres, remove the guardrails, and win the race with China, speed chosen over caution not by accident but as declared policy. The debate has framed this as a values question, safety versus progress, caution versus ambition, and argued about which value should win. But the deeper structure of the choice is not a clash of values at all; it is a trap, a specific and well-mapped dilemma that every powerful technology sets for the society developing it, and that artificial intelligence sets more sharply than any technology before it.

The trap has a name, given by the scholar David Collingridge, and once you see it you cannot unsee it in any debate about governing a new technology. The Collingridge dilemma runs as follows: in the early life of a technology, when it would still be easy to control, to shape, to regulate, to steer, we do not yet know enough about its effects to know how it should be controlled, because those effects have not yet unfolded; and by the time the effects are clear, by the time we know exactly what needs controlling, the technology has become so entrenched, so woven into the economy and infrastructure and daily life, that it is no longer feasible to control it at all. Early on, we have the power to steer but not the knowledge; later, we have the knowledge but not the power. The window in which we possess both is narrow and closes fast, and the whole difficulty of governing any powerful technology is the mismatch between when control is possible and when the need for it is understood.

Now the counterargument, which the strategy's defenders press with real force and which I will not caricature. The race with China is real; a powerful technology developed first by an authoritarian rival is a genuine danger; caution that cedes the lead is not safety but its own kind of recklessness; and there is no version of this in which the West simply pauses to get the governance right while its competitors race ahead, because the competitor's progress is not hypothetical. This is the strongest case, and it is not wrong about the dilemma it names: unilateral caution in a race can indeed be a form of losing. But notice that this argument, valid as far as it goes, is precisely the logic that drives every party through the Collingridge window before anyone has learned to steer, because every party faces the same race and reasons the same way, so that the competitive structure guarantees that the technology is built out and entrenched during exactly the phase when control was still possible, foreclosing it for everyone. The race does not resolve the dilemma; it weaponises it, ensuring that no one steers while steering is possible because everyone fears that to steer is to lose.

And this is why the strategy's explicit choice of speed is graver than a mere values error, and why artificial intelligence makes the Collingridge trap sharper than any technology before. The dilemma has always been hard, but with most technologies the entrenchment was gradual, leaving some time to learn and adjust; artificial intelligence entrenches at a speed and scale that collapses the window almost to nothing, building itself into the economy's foundations faster than its effects can be understood, so that the phase in which control is possible and the phase in which the need for control is clear may barely overlap at all. To choose speed explicitly, to deregulate and build and race, is to choose to sprint through the closing window, to entrench the technology as fast as possible during exactly the period when we understand it least and could still shape it most, and to arrive at the far side, where we finally understand what needed controlling, having made control impossible. The strategy calls this winning. Collingridge's framework calls it the oldest trap in the governance of technology, entered at full speed, with the door already closing, in the name of not being the one who slowed down.

The Ledger

The AI Action Plan formalises the year's regulatory reversal: from the Biden guardrails and the Bletchley spirit to explicit deregulation in the name of beating China. The gap between AI's power and its governance, which this paper has tracked since 2023, is now being widened as deliberate policy.

Data centers & power

The plan's centrepiece is the power question this paper has pressed for three years, now embraced as strategy: fast-track data centres and the power plants, including fossil and nuclear, to feed them, on federal land. The build-out's energy hunger is now official national priority, environmental cost be damned.

Rates

The trade deals with Japan and Europe, ratifying a 15 percent tariff baseline, remove some uncertainty while institutionalising higher costs. The inflationary effect is now structural; the Fed adjusts to a permanently more protectionist world.

Real estate

US 30-year mortgage 6.74 percent (July 24). Steady; the structural rise in tariffs and debt keeps a floor under yields, and thus under mortgage rates, for the foreseeable future.

India tech

The AI Action Plan's push to export American AI and dominate the global market is a direct competitive challenge to India's own AI ambitions; the deregulatory, speed-over-safety approach also sets a global tone that India's frugal, pragmatic AI sector will have to navigate.

What we called wrong

Nothing to retract. The AI Action Plan is the culmination of the regulatory arc this paper tracked from Italy's ChatGPT ban through the EU AI Act to this: the pendulum swinging hard from caution to speed, widening the governance gap this paper has warned about since the beginning.

The Back Page

The last note

Ozzy Osbourne died this fortnight at 76, just weeks after hauling his broken body onto a stage in his native Birmingham for a farewell concert, seated on a black throne because Parkinson's disease and years of hard living had robbed him of the ability to stand, singing to fifty thousand people who had come to say goodbye to the man who, more than almost anyone, invented heavy metal. It was a fitting exit for a genuinely strange and improbable life: the working-class Birmingham kid who could barely read, who founded Black Sabbath and gave dark, heavy, doom-laden rock its template, who bit the head off a bat, who by every reasonable expectation should have died decades ago from the sheer quantity of substances he consumed, and who instead survived into a bewildered elder-statesman role, reinvented as a lovable, mumbling reality-television grandfather to a generation that had never heard his music. Beneath the caricature was a real artist and, by the accounts of those who knew him, a genuinely kind and gentle man, terrified and bemused by the demonic persona he had created and ridden to fame. He got his farewell, which few performers do, one last night on the throne in his home city, singing to the faithful, and then he was gone, weeks later, having somehow beaten the odds long enough to say goodbye on his own terms. The prince of darkness, it turned out, was a sweet man from Birmingham who loved his family and could not quite believe his luck, and he took his final bow, seated but singing, in front of the people who loved him, which is more than most of us will get, and exactly what he had earned.