Backcover | EYES |

Cybersecurity Eye

  • DPRK-Linked macOS Malvertising Uses Fake Updates to Deliver Crypto-Stealing Malware
    Threat actors with ties to North Korea have been attributed to a sophisticated macOS malvertising campaign that involves redirecting users to fake web pages displaying a full-screen non-existent update sequence to deliver malware as part of a new iteration of the long-running Contagious Interview campaign. […]
  • ThreatsDay: AI-Powered Hacking, 370 Chrome Flaws, SonicWall Attacks, DNS Hijacking + 22 More Stories
    A lot of security still comes down to trusting the wrong screen. This week, that screen might be a login page, an install guide, a recruiter call, or a familiar service behaving slightly wrong. Behind it: reused credentials, exposed systems, quiet loaders, abused trust, and […]
  • Azure Cosmos DB Flaw Exposed Platform-Wide Key That Could Access Any Database
    A now-patched vulnerability in Azure Cosmos DB could have let an attacker escape the service's Gremlin query sandbox and obtain full read and write access to databases across customer tenants, according to Wiz. Wiz, which codenamed the chain CosmosEscape, said the exploit chain began with […]
  • Microsoft Copilot for Word Can Copy Hidden Prompts Into New Documents
    Hidden instructions in a Word document can make Microsoft 365 Copilot rewrite figures in a report, then copy the same instructions into the finished file. Håkon Måløy disclosed the technique on July 28, 144 days after reporting it to Microsoft. In his proof of concept, […]
  • The Network Has Become the Control Plane for AI Security
    Network firewalls are the workhorses of modern cybersecurity. They are trusted to protect the network, blocking malicious traffic and preventing intrusions and breaches. And for decades, network security teams have built controls around a relatively stable model: users connect to applications, applications exchange data, and […]
Articles by the Director
(The Director Eye)

Not yet: the third answer in machine intelligence

Article 004 · 14 July 2026

The most difficult thing to build into an artificial intelligence is not accuracy. It is restraint. Contemporary systems are rewarded, structurally, for producing an answer — a label, a score, a confident sentence — because a fluent reply reads as competence and a hesitation reads as failure. Yet in the places where these systems will matter most, medicine among them, the honest reply is often neither yes nor no. It is not yet. A machine that cannot say so, and instead manufactures closure to satisfy the shape of a question, is not intelligent in any sense a physician would recognise. It is merely articulate.

This is the problem the Vectorial System was built around. Its logic is ternary rather than binary: every claim resolves to admission, refutation, or a preserved, declared indeterminacy — the value we write as U. That third state is neither an evasion nor a rounding error. It marks a structural commitment: when the evidence does not close, the system is required to say that it does not close, and to record why — which measurement is missing, which interference remains unresolved, which boundary was only partially traced. U is what separates an instrument that reports honestly from one that performs certainty on demand.

The Vectorial System has given this orientation a name and, increasingly, a body. Watson — after Conan Doyle's physician, the observer who asks the plain question and refuses to be hurried — is the humanoid the SV is developing on this foundation. It is defined not by any resemblance to a person but by declared domain, interface, transduction, residual and return. Its usefulness is meant to begin in medicine: immunology, genetics, pulmonology, haematology, oncology, differential diagnosis. What some would call its naivety — its insistence on asking again, on holding the U when a lesser system would have concluded — is precisely the point. It is resistance to premature closure, made into a design principle rather than apologised for.

The clearest way I can show what this discipline does is to take it out of the clinic and into a place where it has no permission to bluff: the determination of a chemical element. Within the extended structural domain of the periodic table, the SV framework identifies a target it designates SV-399 and asks a single, bounded question — whether a specific actinic-refractory residual, associated with tungsten and its microfrontiers in scheelite and wolframite, survives once every ordinary explanation has been subtracted. The protocol is deliberately small: a sixteen-week pilot, a closed mineralogical matrix, defined starting locations, a published budget. And it is built to end in exactly three ways. Candidate. Rejection. Or U — justified material indeterminacy, recorded with its cause. No favourable presumption is permitted at any gate.

That constraint is the whole argument. A framework that wanted to announce a discovery would design itself to reach "yes." This one is designed so that a negative result is as publishable as a positive one, and so that an unresolved result is not quietly buried but preserved as an honest open edge. It has been placed on the public record, released publicly under a Creative Commons licence, and offered to any laboratory willing to test it — because a claim you are afraid to expose to refutation was never a scientific claim to begin with. The same ternary that lets Watson withhold a diagnosis lets SV-399 withhold a conclusion. In both cases the machinery is trusted precisely because it can return nothing rather than return a fiction.

There is a lesson here for artificial intelligence beyond any single system of mine. The field has spent years optimising for the confident answer and is only now discovering the cost of it — in medicine, where a fabricated certainty can reach a patient, and in science, where a premature closure can misdirect years of work. The intelligence worth building does not always conclude. It concludes when the evidence closes, refuses when it fails, and, in the wide territory between, has the discipline to say not yet and mean it. That third answer is not the weakness of the instrument. It is the reason anyone should trust it.

Previous articles
  1. Article 003 · 11 July 2026
    External validation is not a bureaucratic detail
  2. Article 002 · 11 July 2026
    A medical algorithm must not confuse cost with health
  3. Article 001 · 11 July 2026
    When AI sounds certain but should say “I do not know”

Why did I look?

Quick view; for proper reading, click this link

Archive · European AI Alliance · 2020

From 2020 to 2026: the same problem, two documents

«In May 2020, as a member of the European AI Alliance, I took part in two European Commission debates on the use of artificial intelligence against COVID-19. I keep them sealed, with their comments from the time, because their value lies in the date: what I warned about then in a forum is today, six years on, the doctrine I uphold in these Director’s Articles».

Each document expands separately. Select the one you wish to read.

Document 1 · Diagnosis in hospitals ▸ Using AI to diagnose COVID-19 in hospitals The CT tool across ten European hospitals and the warning on clinical judgment.

Director’s Note — 2026

In May 2020, the European Commission announced within the European AI Alliance the deployment of an artificial-intelligence tool to diagnose COVID-19 from computed-tomography images across ten hospitals in nine countries. The text below is that announcement; the comment accompanying it is the one I signed at the time, as a member of the Alliance. I have not amended it: I keep it exactly as published, sealed, because its value lies precisely in its date.

I warned there of three things. That operating below a minimum sample threshold breeds singularities and dangerous identifications. That replacing clinical judgment with software leads to serious errors. And that institutional haste, no less than slowness, has a price. Six years on, those three warnings have not merely held: they have hardened into architecture. What in 2020 was an intuition written in a forum is today, in these Director’s Articles, reasoned doctrine. Medical AI must distinguish prediction from decision, assistance from authority, and uncertainty from failure; it must undergo external validation before deployment, not after harm; and it must not mistake cost for health. The Watson case —which recommended a contraindicated drug to a patient with haemorrhage, without hesitating— and the Epic Sepsis case are the bill for having ignored what was already known.

From 2020 to 2026, the problem remains the same. Only one thing has changed: we now have the cases that prove it.

Juan Antonio Lloret Egea

Author’s comment

Juan Lloret · 26 May 2020, 15:07

Thanks for sharing this information.

This subject is very topical, along with the use of data in pandemics such as COVID-19. And it is convenient to have real, highly qualified experts in the fields, or investments can be very unsuccessful for Europe. For example, using data below a minimum threshold (statistical sample universe) can lead to singularities or very dangerous identifications for citizens. Also trying to substitute qualified doctors and specialists (guided by the patient’s clinic and the experience of the healthcare provider) with software or hardware can carry serious errors. In the USA this is already done under the supervision of international standards such as the FDA and the IEEE (for example, the Computer Society and others).

If Europe walks slowly, it will be a serious problem for citizens and significant investment losses. On the contrary, if Europe walks too fast it can cause chaos in the profession of medicine: lightning technicians (X-ray technicians), radiologists, etc. Regards, Juan Antonio.

▸ View the original document (European Commission)
Document 2 · The training data ▸ COVID-19 detector using X-ray images Nine hundred images to decide in a pandemic and the warning on scarce data.

Director’s Note — 2026

This second document, also from May 2020 and from the same European AI Alliance forum, addresses the other half of the problem: not the deployment of the tool, but the data it is trained on. It concerns a COVID-19 detector based on chest X-rays, built on a set of barely some nine hundred images.

My comment at the time insisted on the essential point: in a new and poorly documented disease, knowledge of the terrain is everything, and the unknowns were —and are— many. There, already in raw form, was the thesis I have developed six years later in these Director’s Articles: a model trained on scarce or biased data does not learn to diagnose; it learns to repeat the bias of its origin, and it does so with the fluency of one who does not doubt. It is the same lesson as the Obermeyer case —the algorithm that mistook healthcare spending for health need— and the Epic Sepsis case: external validation is not a later formality; it is the prior condition that separates a clinical tool from a well-presented statistical superstition.

Nine hundred images in 2020 to decide on a pandemic. In 2026 we are still discussing the same thing, only now with the statistical casualties in plain sight.

Juan Antonio Lloret Egea

Author’s comment

Juan Lloret · 30 March 2020 · Additional comment V

COVID-19 is a novel, still poorly documented disease, and a good knowledge of the terrain is essential. The description given by Wang and Wong, of the University of Waterloo and Darwin AI, is a good starting point, though I am well aware that the scientific community is working most intensively. Nevertheless, there are still many unknowns, and this alone shows the complexity of the matter. I also attach technical bibliography for context in the references of the original thread.

▸ View the original document (European Commission)

A Different Look

(Papers with Code — Trending Research)

The most influential AI research papers with code, datasets, methods and evaluation leaderboards — ranked this month by citation count. An external reference we recommend for tracking the state of the art.

Open Papers with Code ↗

Opens on paperswithcode.co · content and rankings belong to Papers with Code.

Opening your eyes: Nature & NASA & Science

  • Starburst Galaxy Centaurus A
    NASA’s James Webb Space Telescope’s Mid-Infrared Instrument (MIRI) reveals the nearby galaxy Centaurus A, exposing the dusty structures and hidden activity that shape this unusual system.
  • Understanding How Martian Auroras Are Made
    This illustration depicts charged particles from a solar storm stripping away charged particles of Mars' atmosphere, one of the processes of Martian atmosphere loss studied by NASA's MAVEN mission.
  • NASA Astronaut Chris Williams Returns to Earth
    NASA astronaut Chris Williams is seen outside the Soyuz MS-28 spacecraft after he landed with Expedition 74 Roscosmos cosmonauts Sergey Kud-Sverchkov, and Sergei Mikaev in a remote area near the town of Zhezkazgan, Kazakhstan on Sunday, July 26, 2026. The trio returned to Earth after logging 241 days in space as a members of Expeditions […]
  • New Crew Members Welcomed to International Space Station
    From left, Expedition 74 flight engineers Anna Kikina and Pyotr Dubrov of Roscosmos, and Anil Menon of NASA pose for a portrait while holding a cake celebrating their recent arrival aboard the International Space Station.
  •  Crews Move Artemis IV Liquid Hydrogen Tank
    Crews at NASA’s Michoud Assembly Facility in New Orleans transport the 130-foot-tall liquid hydrogen tank out of a production cell inside the main factory building into a detached test building on a separate portion of the 829-acre site.
  • A New Look – and Sound – for Messier 94
    Messier 94 is a spiral galaxy with a bright inner ring around it, called a starburst ring, where new stars are forming.
Share this on: