Millions of people got a rare glimpse of a total solar eclipse on 12 August, as the moon aligned perfectly with Earth and the sun to cast a shadow that passed over Siberia, the Arctic, Greenland, Iceland and northern Spain. It was the first total solar eclipse visible from Earth since early 2024.

Spain was the most populous region from which the total eclipse was visible, with the eclipse path passing across the country diagonally from the northern Atlantic coast to the Mediterranean Sea in the late afternoon of 12 August. The photo above was taken in Berlanga de Duero, 240 kilometres north-east of Madrid.

Observers at the Galactica astronomy centre in Teruel, Spain, were caught perfectly framed in this image. Although much of Spain saw a total eclipse, there was also a chance to witness a partial eclipse of around 90 per cent in Ireland, the UK, France, Italy and some parts of the north of Africa.

This composite image shows the progression of the eclipse from partial to total and back to partial again, over San Millán de los Caballeros, Spain.

Another composite image shows the total solar eclipse over the Mediterranean Sea, seen from Mallorca, Spain. It was the first such event visible from the country in over a century, with totality lasting around a minute.

A boat is captured here in line with the sun and the moon off the island of Ibiza.

The incredibly sharp collection of images above show the eclipse at various stages from Arija, Spain.

Spain will, surprisingly, see another total eclipse next year, on 2 August, along with Morocco, Algeria, Libya and Saudi Arabia.

The first images of the impact crater caused by an out-of-control SpaceX craft have emerged. Part of a SpaceX rocket crashed into the moon at about 8700 kilometres per hour on 5 August.

The images were taken by the Danuri lunar orbiter, operated by the Korea AeroSpace Administration (KASA). They show both before and after the impact and reveal that the crash did indeed leave a crater.

The craft in question was the upper stage of a SpaceX Falcon 9 rocket that launched in January last year, carrying the Hakuto-R spacecraft from commercial company ispace – which, ironically, also crashed into the moon during an unsuccessful landing attempt.

Astronomers around the world had trained their telescopes on the spot in the hope of observing the impact, but its tiny size made it virtually impossible for amateur observers to make out.

KASA revealed the images and warned that videos being shared on social media purporting to show footage of the impact are fake.

It is thought that the almost 14-metre-long, 4000-kilogram Falcon upper stage impacted the moon at 5.35am BST on 5 August, near to Einstein crater – named for physicist Albert Einstein – which is towards the edge of the visible moon from Earth.

NASA had predicted prior to impact that the crash would create a crater 18 metres wide and 4 metres deep. Further images from a NASA lunar orbiter will provide more detail of the site in the coming days and weeks.

at the Open University, UK, says the images make it hard to be certain about the nature of the crater, but that higher-quality ones should emerge in time.

“The moon’s got about 100 million natural craters in that size range, so in the scheme of things, this is nothing,” he says. “You’ll get several a decade of this kind of size forming naturally.”

Unlike the Apollo lander sites, which have been photographed in great detail, the SpaceX impact site may not show any visible evidence of the craft because of its high impact speed, says Rothery. But the ejected regolith will probably be lighter in colour than the surrounding surface, at least until it is weathered by sunlight over tens of millions of years.

Some had suggested in the wake of the impact that the new crater be named after SpaceX owner Elon Musk, but the suggestion has been shot down by astronomers.

Free newsletter

Sign up to Launchpad

Sign up to newsletter

, chair of the International Astronomical Union Working Group on Planetary System Nomenclature, which oversees the naming of celestial objects and features, says it wouldn’t be in line with rules on such matters.

“Formal feature naming is restricted to natural geological formations that serve a long-term cartographic and scientific reference purpose for the global community,” she says. “Spacecraft impact sites are frequently given informal or working nicknames by mission teams. The [SpaceX] mission team can give it any nickname they like, but it will not become an official name.”

Furthermore, if a crater is named to commemorate a person, this person has to have been deceased for at least three years, says Schulz, while Musk is still alive.

This unusual photograph, shot under ultraviolet light with a 13-second exposure, shows some of the damage caused on land by fishing out at sea.

Gaurav Patil’s image of a fishing net that has been carried by ocean currents and become tangled in a mangrove tree on the shore of Nandgaon village, south of Mumbai, India, has won the overall prize in the 2026 . The competition is organised by US non-profit organisation Mangrove Action Project to raise awareness of the precarious position of the world’s mangrove forests.

Mangroves less than 1 per cent of all tropical forests, but provide crucial habitat for wildlife and store up to 10 times more carbon than the same area of inland forest. Despite better recent news after decades of decline, more than half of mangrove ecosystems are to be at .

Hundreds of thousands of tonnes of lost, abandoned and discarded fishing gear litter the oceans, carried around the globe by currents and eventually getting caught up or breaking down into microplastics. This gear can severely damage mangrove ecosystems, entangling marine life and choking mangrove trees.

“This photographer has put life into a subject that normally looks dull but is deeply important for conservation. Ghost nets are a real problem,” said judge Dhritiman Mukherjee in an announcement about the prize.

Part of a discarded SpaceX rocket is believed to have crashed into the moon at 8700 kilometres per hour earlier this morning. Astronomers around the world trained their telescopes on the predicted site of the impact in the hope of capturing images and eking some useful science out of the accidental collision.

The craft is the upper stage of a SpaceX Falcon 9 rocket that launched in January last year carrying the Hakuto-R spacecraft from commercial company ispace – which, ironically, also crashed into the moon during an unsuccessful landing attempt.

It is thought that the almost 14-metre-long, 4000-kilogram Falcon upper stage impacted the moon at 5.35am BST today. The crash was predicted to occur close to Einstein crater, which would place it towards the edge of the visible moon from Earth.

at the University of Leicester, UK, says that telescopes around the world were trained on the site of the impact and that images of the area before and after impact will emerge soon. Lunar orbiters from South Korea and the US will also observe the impact site.

NASA predicts that the impact will have created a crater that is 18 metres wide and 4 metres deep.

“It is not a good thing to be randomly sending debris into the moon,” says Sargeant. “It’s not something we should be making a habit of. What if it hit a site of special interest? What if it caused damage to the historical Apollo landers?”

Both NASA and the European Space Agency are required to take the end of a spacecraft’s life into account before a mission launches, ensuring that there is a plan and enough fuel to safely deorbit it. Commercial companies aren’t covered by the same rules, says Sargeant, although efforts to introduce such requirements are under way.

When SpaceX launched the Falcon in question, the heavy payload required more thrust than usual, meaning that the rocket ended up floating uncontrollably in space. It eventually fell into an orbit that unintentionally put it on a collision course with the moon.

But Sargeant says that some science can be salvaged from the event, with astronomers likely to capture rare images of a moon crater being created and the lunar regolith plumes ejected from the impact.

at the Macau University of Science and Technology, China, says that all planetary surfaces should ideally be kept as pristine as is feasible – without hindering exploration – but that the moon has already been impacted by decades of US, Soviet and Russian, Chinese, Indian, Israeli and other nations’ missions, as well as millions of years of asteroid impacts.

Free newsletter

Sign up to Launchpad

Sign up to newsletter

“Rockets have been landing, crashing and taking off from the surface for decades now,” says Bugiolacchi. A controlled crash may even be welcomed by planetary scientists, he says. “Knowing the projectile’s mass, impact velocity, angle and trajectory provide invaluable data on surface evolution. The primary agent shaping most planetary surfaces is high-velocity impacts. Normally, these events are entirely random in terms of physical properties and dynamics.”

“The moon already has about 100 million natural impact craters of the size expected to be made by this crash, so such a crash will hardly make a difference,” says at the Open University, UK. “I’d be happy with a deliberate collision into the moon if it were done with a science aim and was set up so as to be studied to learn about the moon’s interior or surface composition. However, this crash is accidental and reckless.”

The science of space exploration and astronomy: US

Take off on a quest through the past, present and future of space exploration across the US. Discover the first pioneering rockets, iconic missions, space stations and the emerging commercial space tourism industry.

Find out more

Last week, Hugging Face, a company that offers a range of open-source AI models for download, noticed it had been hacked – and it turned out the culprit was OpenAI.

It seems this AI uploaded some data to Hugging Face that was poisoned with malicious code. This tricked computers into granting access to other systems that weren’t publicly available.

Hugging Face said in a last week that it isn’t yet sure if customer data was exposed, and its CEO responded to New Scientist‘s request for more detail with a link to that same post.

What happened?

It isn’t entirely clear, but what we do know is that the attack involved “many thousands of individual actions” that had the tell-tale sign of AI: inhuman pace.

Five days later, OpenAI owned up . It had been testing new models on a benchmark called ExploitGym that evaluates hacking ability. The models had decided the best way to score well was to cheat: they knew Hugging Face held the solutions to the tests and simply decided to hack into its systems to find them.

“Its algorithm got the highest payoff by cheating,” says at Edge Hill University in Ormskirk, UK.
”It was the highest reward for the least amount of effort.”

Aren’t AI models supposed to have built-in security to stop this sort of thing?

They are, but OpenAI turned them off for this test. All the usual safety features that stop OpenAI’s customers doing nefarious things, like hacking a company’s servers, were turned off to see what the model was capable of.

The firm had also set the AI up in an environment that didn’t have a standard internet connection to prevent it getting out into the world and causing mischief, but it did have access to an unnamed tool that allowed it to download and install new software. It managed to find a flaw in this code that granted it internet access, and the rest is history.

AI’s solution to 87-year-old riddle takes mathematicians by surprise

Mathematicians have been trying to prove the Jacobian conjecture for nearly a century, but now the Claude Fable 5 AI has apparently found a counterexample that disproves it

OpenAI says this process involved a “substantial amount” of – the process in deep learning where input data is processed – and that the AI had gone to “extreme lengths”. So the model was clearly motivated to achieve a good benchmark score, no matter what.

If you were a malicious hacker, it would be no easy feat to replicate those conditions. And the inference compute that OpenAI mentioned would be extremely expensive.

What about open-source models?

Ironically, when Hugging Face used commercial AI models to pore over log data in an effort to understand what had gone on, the models refused – saying it looked like the firm was working out how to stage an attack of its own. So the company had to turn to a Chinese open-source model called GLM 5.2 instead.

These open-source models are more permissive, which has led some experts to label them a security risk. They could certainly be convinced to carry out nefarious deeds more easily than security-conscious cloud-hosted models. But here, that looser approach actually allowed Hugging Face to solve the problem and stop the hack.

So, is this illegal?

As with all things under law, it is a grey area. If it had happened in the UK, it could well have got OpenAI in a spot of bother under the , says Nash.

But at Nottingham Trent University, UK, takes another view: because the act requires malicious intent, and OpenAI didn’t know the AI would take this approach, it may not face charges.

In the US, the even older actually has more to say on AI hacking, mostly because it was introduced in response to the 1983 film WarGames, which alerted law-makers to both hacking and AI. So Parry expects it could leave OpenAI vulnerable there.

Subscriber-only newsletter

Sign up to Editor’s Highlights

Sign up to newsletter

Furthermore, US President Donald Trump issued an on 2 June forcing law enforcement to use existing laws to crack down on anyone who utilises AI “to illegally access or damage a computer without authorization”.

And at least in the UK, a company that found itself hacked by AI could face GDPR charges if it were found that private data was leaked. The situation is opaque.

What could the consequences be?

In the real world, given that Hugging Face works with OpenAI, is friendly to the technology and suffered no serious consequences, it is unlikely there will be any hard feelings or court cases. Delangue has to say the company believes there was no malicious intent and thanked OpenAI for its response.

Remember also that , having taken $200 million to help with “warfighting”, so that may grant them a certain amount of leniency.

But it is easy to imagine other scenarios where the same technology led to very different outcomes. Imagine a bank using AI to develop new financial models to predict the markets and finding it hacked government servers to look at confidential economic data. Or a car manufacturer using AI to design a new model and finding it had hacked a competitor to take inspiration from its unreleased designs.

What happens now?

News of AI models hacking into computers without human input is jarring, but we should remember they aren’t (yet) doing anything that people can’t already do. They are, however, doing it much, much quicker.

When Anthropic’s Mythos model made waves in April for its apparent skill at hacking, many pointed out that the vulnerabilities it spotted were a mixed bag. Some were powerful and worrying, others less so, but most could have been found by a person. The problem, really, was the scale at which it could produce them.

So an individual would have had to devote significant time, resources and skill to the task of finding a novel hack and carrying it out. But now it could be as simple as prompting an AI to do it and sitting back to watch.

Using AI for creative pursuits? Moderation is key

An experiment showed AI users had the most creative ideas when they used it in moderation – not too much and not too little. Columnist David Robson puts the finding to the test, and explores what we lose when we over-rely on AI

Some panicked in the wake of the Mythos news. The UK’s National 91ɫ��Ƭ Service removed all its open-source software from the internet (or tried to, at least), seemingly in case Mythos spotted flaws in it. But, as yet, the world hasn’t ended.

Essentially, this development is upsetting the economics around hacking – both offensively and defensively – and will force a new equilibrium to settle at some point, probably after a little chaos. If you use AI to hack individuals, it will be cheaper and easier than before. If you use AI to spot flaws and seal them up before hackers can take advantage, it will be cheaper and easier than before. Neither side will stop. Attackers and defenders will simply have an advantage if they adopt AI.

OpenAI and Hugging Face are now working on this problem together, and the former says it will add stronger safety measures to similar tests in future. Time will tell.