← The Constellation
The Morning Brief · the morning report

Off the front page

Sunday · July 19, 2026 · Off the front page: a seafloor caught forming, a cross-border evidence ruling, and a real standard for ‘may contain’.
I · Earth science

Watching the ocean floor tear open and remake itself

Most of the planet's surface is manufactured at the bottom of the sea, along the mid-ocean ridges where tectonic plates pull apart and magma rises to fill the gap. The process normally hides under kilometres of water and unfolds too slowly to watch. For about two weeks in the spring of 2024, an array of instruments on the Southeast Indian Ridge recorded one of these episodes as it happened, in the closest detail yet obtained.[1]

The ridge runs through empty ocean between the southern tips of Africa and Australia, near the remote Saint-Paul and Amsterdam islands, where the Australian and Antarctic plates separate at roughly six centimetres a year.[1] That yearly average turns out to be a smooth fiction. The new record shows the spreading arriving in a single violent pulse: on 26 April 2024 a swarm of earthquakes began marching along the valley floor as a blade of magma, what geologists call a dyke, forced the crust apart at two to three metres a second.[1] A larger shock, magnitude 5.9, struck on an adjoining fault.[1]

Over the sixteen days that followed, about 160 million cubic metres of lava surfaced onto the seabed, and the floor of the valley sank 4.2 metres, most of it inside the first sixteen hours.[1] In those two weeks the ridge let out roughly four decades' worth of its ordinary spreading.[1]

~160M m³
lava erupted over about 16 days
4.2 m
the valley floor dropped, most within 16 hours
~40 yrs
of normal spreading, released in two weeks
~75%
of the fault motion was silent, with no quake

What sets the record apart is the accounting, not the eruption. The seafloor observatory paired five autonomous hydrophones with fifteen acoustic ranging beacons and a pressure gauge on the seabed, then checked their readings against repeated ship surveys of the valley.[2] That combination caught something a network of earthquake sensors alone would have missed. About three-quarters of the fault movement was aseismic, sliding without generating any earthquake at all, so a seismometer would have registered only a fraction of what the plates actually did.[1]

It is the case the authors make for instrumenting the deep seabed directly, the way faults on land have long been wired: the ridges resurface most of the Earth, and they do it out of sight, much of it without a tremor to announce it.[1]

II · European law

A grave, a paternity case, and a limit on national vetoes

A court in one European Union country was trying to settle a question of paternity, and the evidence it needed lay in a grave in another. Establishing the parent required a DNA sample from a man buried in France, so the court used the bloc's rules on cross-border evidence to ask a French court to order the exhumation. French law forbids taking genetic samples from the dead without consent they gave in life. On 16 July the EU's top court ruled that the French court cannot refuse the request on that ground.[1]

The case turns on a piece of legal plumbing most people never see. That plumbing is Regulation 2020/1783, which lets national courts ask one another directly to gather evidence across borders, and it splits the work into two roles.[2] The court that asks, the requesting court, decides whether the evidence should be taken at all and carries the duty of checking that doing so respects fundamental rights. The court that receives the request, the requested court, is responsible for carrying it out under its own procedures.[1]

The heart of the ruling is that the grounds on which a requested court may refuse are listed exhaustively in the Regulation, and a clash with that country's own substantive law is not among them.[1] The permitted refusals are narrow and procedural: a person's recognised right to decline to give evidence, a request that falls outside the Regulation, a task that is not judicial in nature, missing information, or an unpaid deposit for an expert.[2] A blanket national rule that a kind of evidence may not be gathered does not qualify. If it did, cross-border cooperation would bend to the most restrictive law in any single member state.

The case carries real ethical weight, since it hinges on whether the dead may be tested for their DNA without having agreed to it in life. But the Court placed that judgment with the requesting court that ordered the measure, not the French court asked to execute it.[1] Read narrowly, the decision is about who checks what, and it protects the mechanism that makes cross-border justice work at all: cooperation that one country's prohibitions cannot quietly override.

III · Food standards

A global standard to make ‘may contain’ mean something

The words ‘may contain’ on a packet of food are meant to warn allergic shoppers, but they have quietly stopped meaning much. Manufacturers add them defensively, to almost everything, so a person with a peanut allergy has no way to separate a real risk from a lawyer's reflex. This month the body that writes the world's food-safety benchmarks adopted its first global rule for those warnings, built on a decade of work to put a number under the guesswork.[1]

The body is the Codex Alimentarius Commission, run jointly by the Food and Agriculture Organization and the World Health Organization, and its texts carry unusual weight. Codex standards are the international benchmark that national regulators and trade bodies lean on, so a decision taken in a Geneva conference room reaches the labels and rules of governments that had no one in it. The Commission has 189 members, 188 countries and the European Union.[4]

At its session in early July the Commission adopted harmonised guidelines on what it calls precautionary allergen labelling, to be attached to the global standard for labelling packaged food.[2] The principle inverts current habit. A ‘may contain’ statement should appear only where a risk assessment shows that unintended traces of an allergen genuinely cannot be controlled by good manufacturing practice, rather than as a blanket disclaimer sprayed across the shelf.[5]

The rule leans on a set of reference doses, worked out by an expert panel between 2020 and 2023: the quantity of an allergen protein below which a reaction is very unlikely for nearly everyone who reacts to it.[3] The doses vary widely by food, which is the whole point. They run from around a milligram of protein for walnuts and cashews to roughly two hundred times that for crustaceans, so a warning calibrated to one allergen is meaningless for another.[5]

Protein dose tolerated before a reaction is likely, in milligramsWalnut, cashew~1Peanut, egg~2Hazelnut~3Wheat, fish~5Crustaceans~200 mg: a warning tuned to walnuts is off by two orders of magnitude for shellfish
Why one warning cannot fit all. The protein dose a sensitive person tolerates before a reaction becomes likely spans roughly two-hundred-fold across foods, the science the new rule rests on. Source: FAO/WHO expert reference doses (2023), as reported by Food Safety Magazine.

The same session set the first global standards for baker's yeast, vanilla and large cardamom, and tightened hygiene guidance for chicken and for listeria.[2] None of it changes a label overnight; Codex sets the benchmark and national regulators follow at their own pace. But it turns a warning that had decayed into noise back into a signal, which for someone reading a packet in a supermarket aisle is the difference that matters.