Recent photos from NASA / ESA’s Hubble Space Telescope show magnificent crimson plasma and dazzling blue stars in all their glory. The image in question is from LH 95, a star-forming region in the Large Magellanic Cloud, a dwarf galaxy that orbits the Milky Way.

The light from hundreds of young stars is burning the surrounding hydrogen, creating a bright red glow across the entire image. This type of emission, known as H-alpha, is a good indicator that star formation is actually taking place, and it’s similar to a flashing neon sign that draws your attention. You can also detect darker wisps of dust breaking through the light gas, indicating that denser material has withstood the relentless battering of star winds and radiation.

LEGO Icons Project Hail Mary Collectible Building Set for Adults – Room or Office Decor for Men & Women…

• PROJECT HAIL MARY SCI-FI MODEL – Launch into the world of cinematic science fiction with the LEGO Icons Project Hail Mary (11389) microscale…
• WHAT’S IN THE BOX – Includes everything you need to build a microscale replica of the Hail Mary spacecraft, a functional display stand, plus a…
• FUNCTIONAL SPACE MODEL – Turn the crank on the display stand to send the crew module into orbit, simulating the spaceship’s centrifugal gravity…

Blue and white stars sparkle vividly against the blazing red backdrop. Many of these are extremely hot, massive objects that have only been around for a few million years, and if you look closely, there’s one that really stands out, located just above the middle and to the left, as this bloke has a mass between 60 and 70 times that of the sun, and he’s probably a million years younger than the ones around him, the majority of which are around 4 million years old.

So far, astronomers have identified nearly 2,500 stars in LH 95, which have all of the mass they’ll ever need. They are still in the pre-main-sequence phase, essentially hanging out in their disks while nuclear fusion takes place and transforms them into stable, hydrogen-burning stars. However, the results reveal that this growth stage can last several million years, which is longer than some researchers had previously anticipated. However, accretion rates do slow down over time, but material continues to pour in for quite some time.

Hubble has given us a magnificent view of the hottest gas and brightest stars in all of their glory. The hues in this photo are fairly typical, with shorter wavelengths displaying blue and longer visible light, as well as that magnificent red hydrogen emission glowing brightly red. It all helps to accentuate the dynamic dance between stars and their surroundings, and the dark dust lanes stand out since they are the only areas that conceal some of the backdrop glow, adding depth to the entire impression.

AI is changing how vulnerability research gets done, but most of the conversation is still theoretical: what a model might eventually be capable of, rather than what it can actually find today.

We wanted to answer a more practical question: using the models already available to us right now, how far can AI take us in finding real, exploitable vulnerabilities in production software?

This piece details how the team at Intruder is using LLMs to find novel vulnerabilities using code scanning frameworks alongside current, pre-Mythos models.

We walk through a remote, multi-stage SQL injection zero-day we discovered in a WordPress plugin with over 300,000 users — fully automated from discovery through exploitation, with no human in the loop.

The focus problem: why pointing AI at a whole codebase doesn’t work

The big problem when pairing AI with a code scanner is focus. LLMs are excellent at taking small segments of code, or a description of a specific problem, and finding an interesting solution. But point one at a large codebase and ask it to find security issues, and it will try to ingest every file in the repo.

That’s expensive in tokens, and worse for accuracy: by the time the model is halfway through, its context is full of irrelevant code, and the bug you actually want is buried in noise.

For more complex bugs that require chaining several steps together, you’re then relying on the framework to keep the right context in memory, or retrieve it intelligently when needed. In our experience, that produces poor output rather than real and interesting bugs.

Traditional code scanning frameworks already solve this. We use a technique we’re calling a program slice, which is similar to when an IDE or LSP tool uses features like “find implementation” or a call graph to find all functions called by the current function. These are mature, well-tested tools, and they sidestep the diluted-context problem entirely.

Our pipeline: from codebase to working exploit

We built a pipeline that takes a codebase, runs it through a code scanning engine (we use Joern), generates slices of code relevant to each finding, and uses an LLM to triage and exploit the issue. The design was inspired by nooperator’s work on Slice, though we use Joern rather than CodeQL and designed the slicing algorithm quite differently to handle the specific vulnerability classes we’re looking for. 

We pointed it at the top 200 WordPress plugins — code that’s already heavily picked over by bug bounty researchers, so finding something real there would mean the process can compete with skilled humans.

First, Joern runs against the codebase with rules designed to flag broadly “interesting” patterns — this is deliberately loose to avoid creating rules that are too specific and might miss bugs. Since we have the triage agent filtering later anyway, we can err on the side of false positives.

For this experiment we were after unauthenticated WordPress plugin attack surface, so we had Joern identify every place a script can be affected by user input: REST routes, template hooks, nopriv AJAX calls, and so on.

For each WordPress hook, Joern generates a slice: the function the hook calls, every method that function calls, and so on down the chain. Basic taint tracking rules out obviously safe functions, such as SQL and XSS inputs that go through a known-safe sanitizer. Where we can verify statically that the code is safe to run, we drop those passing onto an LLM.

Each slice goes to a lightweight triage model (Sonnet, in our tests) to filter out the obviously uninteresting: hooks that are meant to be public and have no side effects, for example.

What’s left goes to a heavier model (Opus) to assess exploitability, with the full relevant call context in memory so it isn’t hunting through unrelated source.

Anything judged exploitable goes to a final exploitation agent to try and write an exploit. This agent has access to full source again (if needed) since it can now use targeted searches to find relevant code, and it will also spin up a Docker container running the software to test while developing.

The first vulnerability: a blind SQL injection in a popular WordPress plugin

The first bug the pipeline vended was CVE-2026-3985, a SQL injection vulnerability in the Creative Mail plugin. It stood out to us for a few reasons: 

• It’s high impact, giving an attacker read access to the database (including admin hashes and secret tokens!)

Advertisement

• It requires multiple chained requests to exploit, making it less likely to be detected by traditional tooling

• The root cause was hidden from the developer’s own static analysis tooling by a mistake in their code

Exploitation does require WooCommerce to be installed alongside Creative Mail, but since WooCommerce is a common reason people run WordPress (over 7 million active installs), the combination is common.

The exploitation agent one-shotted a working proof-of-concept, producing a check to confirm the issue existed and a full extraction method capable of pulling password hashes from the database.

This vulnerability was also found independently by Dmitrii Ignatyev of CleanTalk Inc., who reported it to Wordfence.

The plugin has been pulled from the WordPress store pending review; if you’re running Creative Mail alongside WooCommerce, disable it until a patch is available.

For the full technical details, see our write-up.

Discovery is getting faster. Detection has to keep up

This is just the first vulnerability the pipeline has vended. We’re already finding more and reporting them to affected vendors (those are still under disclosure). 

AI clearly has a growing role to play in vulnerability research, and the work now is building the frameworks to get the most out of current models. Attackers are already using similar tooling to feed AI high-signal input, which means the same speed advantage we’ve demonstrated here isn’t unique to defenders.

Vulnerabilities surfaced by our vending machine become detection checks in the Intruder platform, so your next scan finds and reports them.

Get started on Intruder for free.

Author: 

Sam Pizzey, Security Engineer, Intruder

Sam Pizzey is a Security Engineer at Intruder. Previously a pentester a little too obsessed with reverse engineering, currently focused on ways to detect application vulnerabilities remotely at scale.

Sponsored and written by Intruder.

Sometimes the sun throws a temper tantrum, and that tantrum hits Earth. For some people, that means a chance to see the beautiful aurora borealis light up the night sky. But solar storms can cause damage to Earth and the various people and items orbiting in space. That’s because, according to a new study, the risks from solar storms might be worse than originally thought. 

The study, authored by NASA’s Nithin Sivadas and Maria Walach of Lancaster University, posits that science’s understanding of solar storms and the electrical currents generated in Earth’s upper atmosphere may have been misunderstood, and that solar storms may be much riskier, especially for satellites and astronauts in orbit. 

To understand the problem, understanding what the science currently says is key. When solar winds hit the upper atmosphere, they create all sorts of interactions, resulting in effects such as the auroras and electrical current. The current scientific consensus is that there’s a maximum amount of electrical current that can exist in the upper atmosphere because of factors such as solar wind energy, atmospheric limits and current saturation. Once that threshold is reached, Earth’s magnetosphere naturally dissipates the excess.

But this new research suggests that the limit doesn’t actually exist and that earlier assumptions about it were based on “uncertainties in solar wind measurements.” Those uncertainties likely arose because most solar wind measurements are taken by spacecraft about a million miles closer to the sun than Earth, at a location known as Lagrange Point 1.

Measurements taken closer to Earth’s surface by NASA spacecraft show a direct correlation between solar wind strength and electrical current in the atmosphere, and suggest that there is no upper limit as previously assumed. That means the atmosphere can produce as much electricity as there is solar wind to generate it. 

How much risk are we talking?

It’s not well understood how much riskier solar storms are versus what science already knew. Solar storms have caused a lot of damage in the past, with instances such as the Carrington Event in 1859 setting telegraph machines on fire or the 12,350 BC solar storm, which researchers said was “orders of magnitude stronger than everything directly observed.” Further study is still needed to understand the potential risks. 

“If there is no upper limit to our planet’s response to the solar wind, modeling for extreme cases needs to take this into account, and we should be vigilant of space weather effects,” Walach said in a statement. “Fortunately, these very extreme cases are rare, but this also means we have limited data to work with and only time will tell what happens at the very extreme one-in-a-thousand-year kind of event.”

Current simulations that use the limits explained above already paint a pretty grim picture for our electronic-heavy modern existence. A solar storm of the magnitude of the Carrington Event would level a significant number of Earth’s satellites and cause untold damage to systems here on Earth as well. And it wouldn’t be the first time: During the Halloween solar storms in 2003, Earth lost contact with 59% of its satellites at the time. While modern technology is better at resisting these risks, it’s not immune. 

But there is some good news. First, such a catastrophic solar storm wouldn’t spell the end for humanity. Walach says that the Earth’s magnetic field “does a really great job of protecting us against many space weather effects,” and that most of the time, all humans will notice is the occasional glitch or a beautiful aurora. She says satellites would fare more poorly in such extreme space weather conditions. 

The other good news is that the sun is nearing the end of its solar maximum for its current 11-year cycle (assuming it hasn’t already ended), so the odds of extreme space weather are much lower than they were in 2024, when Earth was treated to an extreme solar weather event that pushed the aurora borealis down to Texas for an entire week.

Apple is in early talks with PrismML, a startup that shrinks large AI models to run directly on a phone. The company’s on-device AI pitch could help Apple keep more of Siri’s work off the cloud.

PrismML chief executive Babak Hassibi told CNBC that Apple and other companies were evaluating its technology. He called the discussions very early and said it was unclear where they would lead, but that “things are progressing nicely.” Apple did not comment. The Information first reported Apple’s interest last week.

The startup is a Khosla Ventures-backed spinout from the California Institute of Technology. Caltech owns the underlying patents and licenses them exclusively to PrismML. The company raised a $16.25 million seed round in March.

What PrismML built

On Tuesday, PrismML released Bonsai 27B. It is a compressed build of Alibaba’s open-source Qwen model, not a new one trained from scratch. The company shrank it from roughly 54GB to as little as 3.9GB.

The 💜 of EU tech

The latest rumblings from the EU tech scene, a story from our wise ol’ founder Boris, and some questionable AI art. It’s free, every week, in your inbox. Sign up now!

PrismML ships two versions under a free licence. A ternary build runs on a laptop. A smaller 1-bit build, about 3.9GB, is designed to fit within the memory budget of an iPhone 17 Pro. PrismML says it is the first model of that size to run on a phone.

The trick is how the model stores its internal values. PrismML reduces each one from 16 bits to just one or three possible values. It says this cuts memory use by 10 to 15 times, speeds up responses by six to eight times, and lowers energy use by three to six times.

There is a cost. Hassibi said the compressed models lose a few percentage points of performance. Factual recall weakens first, he said, before skills such as reasoning, maths, and coding. PrismML says its builds keep about 95% of full performance in the ternary version and 90% in the 1-bit one.

Why Apple cares

The timing is not an accident. PrismML released the model a day after Apple opened the public beta of iOS 27, which carries its long-delayed Siri overhaul. Apple is trying to make Siri competitive with assistants from OpenAI and Anthropic.

Running more AI on the device would help. Apple already sends complex requests to cloud models. Keeping more work local would cut delay, lower cloud costs, and support the company’s privacy pitch. Some features would also work offline.

There is a cost angle too. Morgan Stanley estimates Apple’s memory costs could climb sharply in its 2027 financial year. The bank expects the company to raise iPhone prices to protect margins. Smaller models help Apple fit capable AI into tight hardware without paying for more memory.

Claims still to be proven

Analysts urged caution. Tarun Pathak of Counterpoint Research said the real test would be millions of queries across thousands of devices. Phil Solis of IDC said power use was the biggest open question, since a model that runs often could still drain a battery.

The release also feeds a debate over whether efficiency gains will cut demand for memory and data-centre chips. Gil Luria, an analyst at D.A. Davidson, said shrinking models would not remove the need for processors. It would simply move some of them from data centres onto phones, part of a broader shift toward edge AI.

Hassibi said Google’s open-source Gemma model is next in the pipeline, followed by larger frontier models. “It’s very important that the intelligence be local and that it can run fast,” he said.