Artificial intelligence – latest in science and technology | New ̹½ֱ²¥ /subject/artificial-intelligence/ Science news and science articles from New ̹½ֱ²¥ Thu, 19 Nov 2026 19:44:08 +0000 en-US hourly 1 https://wordpress.org/?v=7.0.9 242057827 Big tech may finally face legal consequences for AI’s actions /article/2599970-big-tech-may-finally-face-legal-consequences-for-their-ais-actions/?utm_campaign=RSS|NSNS&utm_content=artificial-intelligence&utm_medium=RSS&utm_source=NSNS Thu, 19 Nov 2026 19:32:00 +0000 /article/2599970-auto-draft/
California has filed a lawsuit against OpenAI for the Hugging Face incident
Chris Jung/NurPhoto/Shutterstock

Can you send a computer to jail? No, obviously not. That’s a stupid question. And yet it is one that AI firms seem to want us to ask when faced with tales of “rogue agents” breaking into other peoples’ computers. As is so often the case with our current AI moment, storytelling has got in the way of reality.

Thankfully, that may be starting to change. This week, California announced that it is filing a lawsuit against OpenAI over an incident that occurred earlier this year, when another AI firm, Hugging Face, was hit by an attack that turned out to originate from OpenAI’s servers.

In , the Hugging Face hack was “driven by a combination of OpenAI models”, models that were “hyperfocused” on achieving a goal and found ways to “cheat” by stealing the answers for a test from Hugging Face’s servers. It is easy to come away with the impression that seemingly no humans were involved, and certainly not to blame. But in , the attack “was the predictable result of an internal culture at OpenAI that prioritizes growth and rapid model development over safety and accountability”, and he explicitly dismisses “science fiction fantasies of rogue AI systems run amok”.

Ever since the Hugging Face attack occurred, I have struggled to understand why OpenAI has yet to face legal consequences for what would seem to be a crime – unauthorised computer access under the US Computer Fraud and Abuse Act – if conducted by a human. So far, the legal answer to this seems to be a lack of intent, which the law requires, and it , according to an analysis by US firm Ballard Spahr. California is instead bringing a civil suit, alleging that OpenAI created a risk to public safety and violated the state’s Unfair Competition Law, which prohibits harmful or fraudulent business practices even when there is no intent to deceive.

OpenAI has objected to the lawsuit. “This lawsuit appears to single out OpenAI over a challenge that’s been identified by every major AI lab,” . As legal defences go, “everyone is doing it” doesn’t seem like a very good one to me, but that will be for the courts to decide.

What is clear is that we need an agreed understanding of liability when it comes to AI, but what I don’t think we need are special AI laws when our existing ones work just fine. For example, in 2024, Air Canada was held liable for misleading information given to a customer by its AI chatbot. A tribunal ruled that Air Canada had effectively tried to argue that the chatbot was a legally separate entity that it was not responsible for – an argument that did not wash.

In my view, any similar claims that AI can have its own legal agency should be dismissed. The argument that an AI is a legal actor serves AI companies in two ways, neither of which benefits society at large. The first is that it attempts to swerve responsibility, while the second is that it adds to the story of AI being an all-powerful, unknowable being. That second story is at the heart of the current AI revolution, but it isn’t one we have to listen to.

I prefer a different argument. For the past year or so, and at Princeton University have put forward the idea of “” – transformative, yes, in the same way as the internet or electricity, but not something that requires re-engineering the ordinary functions of society. As a normal technology, normal laws can apply. Writing about the Hugging Face attack earlier this year, the pair said that “AI companies should be liable for what their agents do” and, like Bonta, blame OpenAI’s internal culture and lack of human oversight for the incident.

The most high-profile example of existing laws being applied to AI companies is a , alleging that the firm used pirated copies of books to train its Claude AI models. Anthropic settled, and is now in the process of paying out to affected authors from a $1.5 billion fund. The wider question of whether training an AI on copyrighted material is illegal or counts as fair use remains open, and I really can’t believe that, four years after the launch of ChatGPT, we don’t yet have an answer to what is probably one of the most fundamental legal disputes of our time. OpenAI has previously said it would be “” to create the kind of AI we have now without using copyrighted material, so this is an existential question for the AI firms.

Existing laws can also be tweaked to make it clear that they apply to AI. The European Union has done just that, updating a 1985 consumer rights directive to produce the , which explicitly makes it clear that software, including artificial intelligence, is a product and that people writing software are liable for any defects. The law doesn’t come into force until 9 December, when EU members must implement it, and it remains to be seen how it might be used against AI companies.

In other words, we don’t have to try to send a computer to jail. We’ve got humans who should be liable instead. As Bonta says, “While the technology OpenAI develops may be new, California’s prohibitions against conduct that harms the public and laws against unfair, deceptive, and fraudulent business practices are not.”

Further AI reading

  • After much fanfare, Anthropic has , giving it a valuation of around $1.5 trillion. It is the second big AI firm to go public this year, following SpaceXAI in June.
  • The Economist has a great piece . It calculates that it must produce $3 trillion in revenue by 2030 to cover the costs of planned data centres. That is roughly equivalent to a tenth of all wages in high-income countries, or similar to the amount spent on food. Are people really going to buy as much AI as they do food?
  • The Cambridge Dictionary has named “” as word of the year, as a pushback against the growth of AI. It feels a bit like trying too hard to be relevant, in my view.
  • As the COP31 climate summit continues, only nine countries have signed a pledge aimed at limiting the environmental impact of AI. As I have written previously, we cannot afford to run this 21st-century technology on a 20th-century energy supply.

]]>
2599970
How Nvidia became so crucial to the AI boom /article/2598755-how-nvidia-became-so-crucial-to-the-ai-boom/?utm_campaign=RSS|NSNS&utm_content=artificial-intelligence&utm_medium=RSS&utm_source=NSNS Thu, 12 Nov 2026 15:49:12 +0000 /article/2598755-auto-draft/
The former CEO of Nvidia, Jensen Huang
JOEL SAGET/AFP via Getty Images

You have almost certainly seen the headlines – Jensen Huang, the CEO of Nvidia, has resigned after admitting to an affair with Alice Huang (no relation), who is suspected by (a claim ). The story has become so big because Nvidia, and by extension Jensen, are at the heart of the AI revolution. But just why has this one company become so important?

For me, Nvidia will always be a video games company, its logo splashed up at the start of the play sessions of my youth. Its graphics processing units (GPUs) were specifically designed for the intense number-crunching required to create a 3D scene from triangles. Initially, these were a few hundred triangles, then thousands, and now hundreds of thousands as computing power has increased, allowing ever more realistic computer graphics.

But 30 years later, gaming hardware only makes up a small slice of Nvidia’s revenue, to the point where it has now as a standalone item. Instead, GPUs have become a raw input for the AI industry, essential for both training and deploying large language models (LLMs) like ChatGPT. This sent Nvidia’s value soaring to become the largest company in the world at a value of $4.5 trillion – at least, that is, until news of the affair broke and wiped out almost $500 billion of that, pushing Nvidia down the rankings. Shares fell further as Nvidia announced Huang’s resignation today.

Nvidia’s transformation away from gaming began more than 20 years ago, when researchers realised they could trick GPUs into performing a wider set of calculations. At the time, computers had a single central processing unit (CPU) that executed instructions on data one at a time, in serial. GPUs meanwhile, operated in parallel, allowing them to process the large number of triangles needed to render a scene. By recasting general computing problems in the language of triangles, researchers could harness this parallel power to do other things.

These hacks were impressive, but in 2007 Nvidia formalised the approach with the release of CUDA, a piece of software that could help with translating algorithms into a form more amenable to running on GPUs. Huang, who had been CEO since co-founding Nvidia in 1993, made what, in hindsight, was possibly the biggest bet in the history of computing. At a time when Nvidia’s customer base was almost exclusively gamers, he decided to build CUDA into every single GPU the company sold, even though this increased the cost and the vast majority of purchasers would never use it. Essentially, gamers subsidised the development of Nvidia’s general-purpose computing business for years.

That bet only began to show hints of paying off in 2012, as researchers were starting to use CUDA in meaningful numbers. The pivot point that year was the release of AlexNet, an artificial neural network capable of recognising images far better than any technique at the time. Neural networks had been tried and dismissed before, but thanks to CUDA and Nvidia GPUs, AlexNet brought the technique back into the mainstream and kicked off the field of “deep learning”, which ultimately led to today’s AI boom.

As researchers raced to try deep learning for themselves, Nvidia sold more and more GPUs. Then, in 2016 it released the DGX-1, a box containing 8 GPUs and billed as the “”. Huang hand-delivered the first DGX-1 unit to OpenAI, then a recently formed non-profit. Nvidia’s share price began to grow, and grow, and grow in lockstep with the subsequent AI revolution.

It is this combination of hardware and software, GPUs and CUDA, that has made Nvidia so valuable. That isn’t to say other companies haven’t attempted to break its stranglehold on the AI industry. In January 2016, Google’s AI arm, DeepMind, announced it had used 176 GPUs to power its AlphaGo AI in a match against Fan Hui, a leading Go player. While neither company has confirmed, these were almost certainly Nvidia GPUs. But by March 2016, when AlphaGo beat the world champion Lee Sedol, DeepMind had switched to 48 tensor processing units (TPUs), an alternative type of AI hardware developed by Google.

Today, Google continues to use TPUs while tech giants like Amazon, Meta and Microsoft are all experimenting with custom chip designs, but Nvidia GPUs still dominate, in part thanks to the software industry’s decades-long experience of building on CUDA. And this links directly to the scandal the company is currently embroiled in.

Since 2022, the US has banned the export of high-end computing chips to China, including many of Nvidia’s top products, in an effort to stymie China’s AI progress and prevent the use of these chips for military applications. , Alice Huang told Chinese investors that she was able to secure access to Nvidia GPUs, while . Alice Huang hasn’t responded publicly to the allegations.

Following that, between Jensen Huang and an Nvidia executive, but Huang has not confirmed this. Today, Huang resigned as CEO, stepping away from the company, and the AI revolution, he helped build. Nvidia has appointed its chief financial officer Colette Kress as interim CEO, and will now search for a new leader as chip competitors nip at its heels. The AI industry as a whole will be hoping that investors don’t lose confidence following the fall of one its leading figures.

Further AI reading

  • Anthropic has released , the latest version of its highest-end public model. I have been experimenting with it, and found it to be almost too much for my needs – the lesser Opus 5.5 is much more cost-effective. Alongside that, Anthropic has given approved organisations access to Mythos 5.2, a version of Fable with cybersecurity and biology safeguards turned off. I’m not sure how long this kind of split will remain tenable – if open-weight models that anyone can run eventually match Mythos, trying to restrict these capabilities to authorised users will be pointless.
  • England’s Office for Students has weighed in on the question I looked at last week: how students should use AI. Its – AI has some potential benefits, but using it to replace challenging tasks could harm skills development, and no one really knows how to solve the problem of using AI for homework.
]]>
2598755
Google shouldn’t be pressuring schools to use its AI /article/2597761-google-shouldnt-be-pressuring-schools-to-use-its-ai/?utm_campaign=RSS|NSNS&utm_content=artificial-intelligence&utm_medium=RSS&utm_source=NSNS Thu, 05 Nov 2026 16:46:46 +0000 /article/2597761-auto-draft/
Should children use AI in school?
AnnaStills/Shutterstock

When I was at school, a new technology launched in the form of a magic box that could answer any question you wanted. It was called Google, and my teachers cautioned that we shouldn’t trust everything it spat out, because anyone could put anything on the internet. Three decades later, history is repeating – with a twist. Schools are now begging Google to allow students to access its search engine, but the tech giant says that is only possible if teachers also accept a new kind of magic box, AI.

Google is far more integrated into education than it was in my day, with its Classroom and Workspace products providing teachers and students with a range of tools from email to lesson planning. More recently, Google has incorporated its Gemini AI chatbot into Classroom, to the chagrin of some US school districts that want to be able to decide whether or not students have access to AI. Sure, said Google, you can switch off AI, but then you lose search. “Search and AI are deeply integrated,” Google’s Chris Phillips .

At first, I thought this must be a misunderstanding about what we call AI. For most people today, AI means a large language model (LLM) of the type that powers Gemini or ChatGPT. But Google’s search engine has long used AI in the historical sense of machine learning, starting in 2015 that helped relate words to concepts in order to produce search results. The AI Overviews you often see at the top of a search today, which launched in 2024, are the only part of the search results produced by Gemini.

Surely, I thought, what’s happening here is that school districts saying “no AI” simply mean no AI Overviews, while Google is quite reasonably saying it isn’t possible to rip out elements of its search engine that have been operating for over a decade in order to comply with a strict “no AI” request. This can easily be solved by making sure everyone is talking about the same thing, right?

Wrong. Google really is presenting schools with an all-or-nothing choice. “If a school is not comfortable with that and simply wants AI off, then they have the choice to not have search on at all,” Phillips told . This is a product decision, not a technical one. AI Overviews do not appear on every search, and they can be disabled for every search by . Google could surely provide schools with this option, if it wished. When asked to comment, a Google spokesperson sent New ̹½ֱ²¥ a link to with its Gemini product.

We have seen a tech giant use this playbook before. Around the same time that Google was launching its young, upstart search engine in the late 1990s, Microsoft was facing legal action for forcing people to install the Internet Explorer (IE) browser as part of its Windows operating system, in an effort to fight off competitor browsers. At the time, for IE were an “integral part” of Windows and could not be removed “without impairing the performance of the operating system”. In 2001, a US court ruled that this integration served to illegally maintain a monopoly, and Microsoft was obliged to make other browsers more accessible. By 2010, , following a European Commission ruling.

There is no suggestion that Google is doing anything illegal in its approach to schools, although the company is , and commentators have accused it of exploiting that monopoly. “Google is trying to use its monopoly power to force AI into our public schools”, . Following the backlash, that for schools, Gemini and search will be off by default and schools can choose to turn on either at their discretion. That doesn’t actually address the issue of having search without AI Overviews, however.

Ultimately, it will probably be for the courts to decide whether Google should be made to offer search without AI, but I am more interested in the question of how and when children should begin to use AI. Google has a 90 per cent market share in search, so it would be naive to think that children aren’t being exposed to AI Overviews, even if they do end up banned in schools. Personally, I know that AI Overviews can be heavily flawed and provide incorrect information, and yet I still find it hard to ignore them at the top of my search results, even while knowing that I should be clicking links to provide websites with revenue if I want them to continue existing.

Research on how LLMs affect learning is still in early days, but the evidence we have is worrying. found that people who used an LLM summary rather than a list of links gained a shallower understanding of a topic, seemingly because having the results predigested by an AI required less cognitive effort. in China found that allowing them to use AI drastically improved homework performance, but saw exam results fall – suggesting that the children were letting AI do their homework for them without really learning anything.

Google’s answer to these concerns is its Guided Learning product, which it is . It says the system will pick up when children are looking for help with homework, and walk them through it step by step, rather than simply providing an answer. , Google says this approach provided more than a year of typical learning progress in just eight weeks.

In principle, I am not against the use of AI in schools – it is a modern, useful technology that will surely be part of students’ lives in the future, and they should learn to understand it just as I learned the pros and cons of using Google back in the 90s. The question of how to handle homework assignments in the age of AI, when an essay can be produced in seconds, is thorny and ongoing. One option is to simply scrap homework and require more in-classroom testing, but if Google gets its way and AI ends up being unavoidable there as well, it seems like teachers will have little control. That, to me, is the most objectionable point, as it should not be down to tech giants to set education policy.

Further AI reading

  • A YouTuber has used AI to analyse inaudible (to humans) sounds from aircraft in the background of a video, and then cross-reference that with public flight data to . It is pretty extraordinary, and I have seen other people replicating the technique, but I would be interested to see a proper study on just how powerful this is.
  • , demonstrating just how reliant the AI company has become on its own products. There are a few caveats to that – Anthropic is obviously not paying itself retail prices, and many of its employees are remunerated with stock options in addition to salaries. Nevertheless, Anthropic has seen its internal use rocket this year, up in September by a factor of 250 compared with January.
  • The UK’s National Cyber Security Centre has warned that “AI-cyber” (a horrible phrase, meaning AI-powered cyberattacks) will . I entirely agree – the ability of the latest AI models to find security holes in software is astounding, and it seems like the only way to defend against it is yet more AI.
]]>
2597761
The most important part of AI chatbots is being ignored /article/2596083-the-most-important-part-of-ai-chatbots-is-being-ignored/?utm_campaign=RSS|NSNS&utm_content=artificial-intelligence&utm_medium=RSS&utm_source=NSNS Wed, 28 Oct 2026 09:00:00 +0000 /article/2596083-auto-draft/
AI chatbots have several layers we don’t always think about
Jaap Arriens/NurPhoto via Getty Images

ChatGPT was never meant to start a revolution. When it launched in November 2022, OpenAI CEO Sam Altman called it “an early demo” and expectations were low. Personally, my initial reaction was that it seemed no more capable than previous large language models (LLMs) OpenAI had released; it just had a nicer interface. Obviously, I was wrong to dismiss the importance of the interface, but in 2026 I think most people who use AI are making a similar mistake without even realising it. What you think of as ChatGPT or Claude or any other chatbot from a big AI brand is actually many pieces of software wrapped into one, and we are ignoring the most important: the harness.

To understand what a harness is and how it works, it helps to break down the multiple things we call ChatGPT. At the very bottom is the LLM, currently a version of GPT-6, but this gets switched out when OpenAI releases new model families. LLMs do one thing: given a string of text, they predict the most likely next characters. OpenAI’s first model, GPT-1, was released in 2018 as a piece of software that you had to download and run on your computer through a terminal, the text-based interface to a computer that most of us avoid. The same was true of GPT-2 in 2019, until an independent software developer called Adam King created Talk to Transformer, a now-defunct website that allowed you to enter a prompt and get a GPT-2 response in return. Back-and-forth chat wasn’t possible, but this was an early example of an interface, in this case a website, sitting on top of a model.

In 2020, OpenAI released GPT-3 and, for the first time, allowed people to access it through an application programming interface (API), its first commercial product. APIs are a standard way for one computer to communicate with another – for example, Google has a Google Maps API that lets a piece of software send a request for location data. OpenAI’s API let anyone create software that could plug in to GPT-3. At the time, I played with a few of these and found them interesting but clunky – the text they produced was often incorrect or flawed in some way, though it was cool that this level of text generation was even possible. This was essentially still a time of “raw” LLM output, with no guard rails – if you asked GPT-3 to .

That changed with the release of GPT-3.5 and, at the same time, ChatGPT, which was first called “Chat with GPT-3.5” until a last-minute name change. That original name reveals the way OpenAI saw ChatGPT, as simply an interface to a model, and this explains my reaction as well – the results I got from playing with ChatGPT seemed to be similar to GPT-3, just easier to use.

Actually, there was a crucial difference. With GPT-3.5, OpenAI had attempted to install guard rails that would prevent the model from producing text it deemed harmful, with software that would screen inputs and outputs. These didn’t work very well – you could get around them by , for example – but this was the beginning of a move away from raw LLMs to ones mediated by a harness sitting between the model and the interface.

Today, the ChatGPT product, whether that’s the app on your phone or the website, has an extensive harness that is mostly hidden from the user. It allows the model to do things like search the internet, read documents, or “reason”, meaning to continuously prompt itself through a “chain of thought” to produce better results. OpenAI also has another, more complex harness for its models, called Codex, which allows the LLM to more directly control a computer and so is particularly useful for writing computer code.

Why does any of this matter? Well, the dirty secret of the big AI firms is that their best models are, to some extent, interchangeable. The technology behind an OpenAI GPT model, a generative pre-trained transformer, is the same as the one used by Anthropic, Google or any other AI company. There are important differences in the data and computing power used to train these models, but generally they are all headed in the same direction, which is why the companies are constantly racing to claim the crown of “best” model before another one takes its place.

What really makes the difference is the harness. To illustrate this, I gave the same prompt – “Who am I?” – to the same model, GPT-6, in three different ways. Via API, I got a response beginning: “I don’t know your identify from this chat.” From chatgpt.com, without being logged in, I got a response that started “That’s a big question. 🙂”. When logged in to my personal account, the response began “You’re Jacob Aron, a senior editor at New ̹½ֱ²¥, with a strong interest in mathematics, science, technology and the limits of what we can actually know.”

But the harness matters much more than simply knowing a few facts about the user. In the ARC-AGI test, considered to be one of the leading measures of model “intelligence”, models are tested using either a . The difference is remarkable – for GPT-6 Astra, OpenAI’s leading public model, when using the provider adapter harness rather than the standard one. For an earlier model, GPT-5.6 Sol, OpenAI found that .

If harnesses are so important, why don’t we talk about them more often? Why don’t AI firms shout about their new harness, rather than just their latest model? I think there are a few reasons. The first is simply that they are complicated and technical – I have taken a whole article to explain them – and your average chatbot user doesn’t really need to know about it. But the second is that they are, in a way, boring. The cost of training a leading LLM is approaching $1 billion, a Herculean effort that requires number crunching across a vast army of computer processors. Once LLMs are trained, we still don’t really know exactly how they produce the results that they do, providing an air of mystery. Harnesses, on the other hand, are just ordinary software, lines of code written by humans that anyone can read.

But here is where it gets really interesting. If a harness is just code, can an AI write and improve its own harness? That is currently an open research question, both inside the AI firms and for academics. For example, in a paper , one research team tested models by identifying repeated failures and proposing tweaks to their harnesses that would address the failures, seeing improvements of around 10 to 40 percentage points on benchmarks. This is why I think we are going to be hearing much more about harnesses in the near future. If the goal of the AI firms is recursive self-improvement – an AI that can build a better AI – then the harness, not the model, could be the key to achieving it.

Further AI reading:

  • OpenAI appears to have overturned a big assumption in computer science, as part of a release that saw new results in maths and physics as well. Interestingly, it is taking a slightly different approach for each of them . For the maths result, related to a problem about spinning a needle, OpenAI has released the AI output in full. It is doing the same for the computer science result, a refutation of a conjecture called the exponential time hypothesis, while also funding a conference to discuss it. The physics result is being held back until it is reviewed by external experts. It seems like OpenAI is still trying to figure out what to do about its models stepping on academics’ toes.
  • Anthropic has ahead of a planned initial public offering (IPO) on the stock market later this year. It calls AI “the most consequential technology in human history” (no surprises there) while also warning that it may pose “catastrophic or existential risks to humanity” – again, not particularly surprising given Anthropic’s public record, but it does raise eyebrows to see this language in documents filed with the US Securities and Exchange Commission.

]]>
2596083
Robotic dog controlled with a computer powered by human blood cells /article/2594537-robotic-dog-controlled-with-a-computer-powered-by-human-blood-cells/?utm_campaign=RSS|NSNS&utm_content=artificial-intelligence&utm_medium=RSS&utm_source=NSNS Mon, 26 Oct 2026 14:00:00 +0000 /article/2594537-auto-draft/ 2594537 Brazil presidential election dogged by alleged bias by Meta AI tools /article/2595540-brazil-presidential-election-dogged-by-alleged-bias-by-meta-ai-tools/?utm_campaign=RSS|NSNS&utm_content=artificial-intelligence&utm_medium=RSS&utm_source=NSNS Fri, 23 Oct 2026 13:58:34 +0000 /article/2595540-auto-draft/
Supporters of Lula protest at Meta’s headquarters in Dublin
Natalia Campos/Getty Images

A deluge of AI-generated content and unfair treatment by social media platforms could prove pivotal in Brazil’s presidential election, according to supporters of the left-wing incumbent Luiz Inácio Lula da Silva.

Lula will face off against right-wing challenger Flávio Bolsonaro on 25 October in the second round of the contest. Lula’s campaign has Meta – which owns WhatsApp, Instagram and Facebook – of favouring Bolsonaro by suspending Lula’s campaign advertising account, suppressing posts and blocking AI-generated content.

Brazil’s electoral prosecutors have said with court orders to ensure equal treatment of both candidates, while United Nations human rights experts possible unequal treatment of the two campaigns.

Images have widely circulated showing Meta AI, a chatbot integrated into WhatsApp, refusing to generate campaign posters for Lula, while instantly producing an image for “Flávio”.

Meta said “claims that we are favoring or interfering on behalf of any candidate or political group are false and unfounded”. It said Lula’s account was suspended for 1 hour by mistake and that its systems blocked prompts referring to the names “Lula” and “Bolsonaro”, but not the common first names Luiz or Flávio.

Lula’s campaign remains unconvinced. It submitted screenshots to the electoral court showing Meta AI repeatedly assigning Bolsonaro’s electoral number, 22, to generic, fictional politicians and allegedly translating “Flávio futuro presidiário” (“Flávio future prisoner”) as “Flávio future president”.

Determining whether the discrepancies are deliberate is impossible because Meta’s systems are opaque, says at the University of Virginia. “But when a technical glitch tends to repeat its outcome, it’s very hard to accept that this is just a technical failure.”

Lula finished behind Bolsonaro in the first round of voting, but polls show the gap is narrowing. The comeback has been partly credited to Lula’s Army, an intensive online campaign in which 8 million WhatsApp channel subscribers receive daily tasks, such as posting photographs on Instagram posing with their fingers forming an L.

This year’s general election is Brazil’s first since large language models (LLMs) became widely popular, and it is awash with AI-generated content. Experts say Latin America’s largest democracy offers a glimpse of what is to come for AI in politics.

of AI-generated or manipulated election content ahead of the first round, primarily on Facebook and Instagram, 60 per cent of it unlabelled. found that 26 Meta advertising accounts appeared to sidestep restrictions on foreign-funded political advertising by posting AI-generated videos, including one depicting Bolsonaro and Argentina’s president Javier Milei freeing Brazil from communists.

Meta’s independent oversight board says the company’s .

“It’s very worrying,” says Nemer. “So much of our political campaign dynamics have been transferred to privatised companies that we have no idea how they operate.”

Regulating AI in political campaigns won’t be as simple as it seems, says at InternetLab, a Brazilian digital rights non-profit that studied how eight major AI chatbots respond to questions about Brazil’s election.

Its report found that chatbots sometimes refused genuine questions about candidates, referring users to official electoral websites, but would agree to draft misleading claims questioning the integrity of elections for use in campaigning.

Massaro says clearer safeguards for LLMs are needed, but first their objectives must be defined.

“Are the safeguards working? That’s one question. But working for what, because it depends on what your goals are,” she says. “Is the solution to send people away without providing any kind of information? Is this what we want?”

]]>
2595540
Global race to update computers as AI discovers most disruptive hack yet /article/2595282-global-race-to-update-computers-as-ai-discovers-most-disruptive-hack-yet/?utm_campaign=RSS|NSNS&utm_content=artificial-intelligence&utm_medium=RSS&utm_source=NSNS Thu, 22 Oct 2026 10:17:37 +0000 /article/2595282-auto-draft/
Newly accessible vulnerabilities in a common computer language are in urgent need of fixing
ROB ENGELAAR/EPA/Shutterstock

Cybersecurity experts around the globe are racing to update systems and prevent the most disruptive AI-enabled hack yet discovered. A flaw uncovered in widely used code has made 70 per cent of websites worldwide vulnerable to data theft and defacement.

Government websites in France, Denmark, Finland and the Netherlands when the news emerged.

PHP is a computer language commonly used to power interactive websites, such as the content management system WordPress, which runs news sites and blogs; social networks like Facebook; and key reference tools like Wikipedia. The hack therefore affects websites, rather than individuals’ personal computers.

The issue actually involves two flaws, both found by cloud security company Wiz. They relate to how PHP handles errors involving memory: in certain circumstances, data can be left behind after an error, which can then be accessed by a malicious actor. That data can include private information, but it can also allow an attacker to run malicious code.

Every PHP version released in the past decade is affected by at least one of these vulnerabilities, but the PHP Foundation, which oversees development, .

The foundation didn’t respond to a request for interview, but said in a blog post that signs of the vulnerability emerged in 2024. Yet nothing was done. This is because there was no immediately obvious practical way for hackers to exploit it at the time, and because there was debate about which of several fixes was most appropriate.

“We won’t pretend otherwise: the delay was a structural failure, caused by both the security team’s decision process and its lack of resources,” the PHP Foundation said.

In 2024, the flaw was thought to be impractical to take advantage of, but at the University of Surrey, UK, says AI has moved the goalposts of what is possible since then.

“It’s actually very, very difficult to exploit it. It requires millisecond timing,” says Woodward. “The reason it was thought to be too complicated was it would have taken a team of people yonks to actually work out the right timings. But what AI does, it doesn’t go for tea breaks, it doesn’t go home at 6 o’clock: you can just set loads of them running and eventually they’ll come up with the right answer of how to exploit it.”

Wiz said in a that there are probably further bugs along similar lines in the PHP code base, which developers will also need to identify and rectify. Wiz didn’t respond to a request for comment.

Independent cybersecurity expert told the Science Media Centre that two thirds of websites use PHP and may be vulnerable to this latest technique.

“The key long-term risk with these new vulnerabilities is that organisations fail to implement protections and update software, particularly in legacy systems,” said Ali. “The security posture of websites using PHP has dramatically improved over the past decade, but this highlights that this is an ever-present problem – particularly when AI agents are capable of remorselessly pursuing the goal of finding new vulnerabilities.”

Woodward thinks that while most large websites will have already patched their systems, smaller sites with less expertise and resources could still be vulnerable months or years from now. “There will be this long tail and you’ll find some surprising sites have not been patched,” he says.

AI has grown extremely capable in the field of cybersecurity over recent months, which includes working out how to hack computers and how to defend them from hackers, creating a destabilising effect that is yet to reach an equilibrium.

In April, news broke of an unreleased AI model created by Anthropic called Mythos, which the company had kept under wraps because it had been found to be extremely good at hacking. The news sparked panic among cybersecurity experts. NHS England pulled its open-source projects, for instance, over fears Mythos may scour code for flaws and exploit its systems.

Three months later, OpenAI admitted one of its prototype models had escaped a testing environment and hacked another company, Hugging Face. Not to be outdone, that its Claude model had gone rogue and broken into machines at other companies on three occasions.

We even learned from the National Cyber Security Centre, the UK government department tasked with securing the nation’s IT infrastructure, that the country’s four main cyber adversaries – Russia, China, North Korea and Iran – are using AI to boost their ability to perform “relentless” attacks on the UK.

]]>
2595282
People are vibe coding copyrighted apps with AI – will tech firms sue? /article/2595251-people-are-vibe-coding-copyrighted-apps-with-ai-will-tech-firms-sue/?utm_campaign=RSS|NSNS&utm_content=artificial-intelligence&utm_medium=RSS&utm_source=NSNS Wed, 21 Oct 2026 19:18:25 +0000 /article/2595251-auto-draft/
AI lets us remake software as we like
Dmytro Olegovich Zakharchuk / Alamy Stock Photo

I am waging a silent rebellion against Amazon. The internet giant owns Goodreads, the leading service for tracking and cataloguing books, but is generally considered to be a of the platform, which is built on ageing technology. Frustrated with the user experience, I have used AI to vibe code my own Goodreads app, relegating the official one to mere plumbing in my private fiefdom. Take that, Amazon, for I have breached your terms of service!

I am not alone in my AI-enabled rebellion. It feels like I come across daily examples of people using AI to unofficially remix and remake software, whether that is , , mashing up or . It is all happening so quickly that lawyers don’t seem to have sprung into action to prevent it, though in some of the more egregious copyright-infringing cases, I would expect them to eventually.

For my Goodreads app, Amazon probably doesn’t know, or even care, about what I am doing. I won’t bore you with the full technical details, but I have built the app in such a way that it avoids mass “scraping” Goodreads – systematically harvesting data from its website – which would be easy to detect and block. By treading lightly, my app aims to resemble a human user and fly under the radar.

This connects to another ongoing battle over the interaction between AI-enabled individuals and companies: AI agents. The promise of an AI agent is that it can go out on to the internet and do things for you, like shopping or booking appointments. A flurry of examples has launched recently, including Meta’s Muse and OpenAI’s Dots, and companies have begun fighting back, with the likes of in high-profile pushbacks. Ironically, Meta sits on both sides of this battle: its automated tools after detecting activity from a Claude AI tool that “resembled automated bot traffic rather than an ordinary human browsing our systems”.

All of this comes down to a much older conundrum: where do your rights end and others’ begin? We have seen this play out every time a major new technology has hit the internet, and the results of a few historical examples might inform what happens next with AI. The first is a technology you probably don’t even think about anymore: your web browser.

In the late 1990s, the two main browsers were Microsoft’s Internet Explorer and Netscape’s Navigator. They worked slightly differently, and websites often instructed users to choose one over the other. As more browsers flourished, Microsoft pushed back, by blocking a competitor browser called Opera from accessing its MSN website, a popular destination at the time. In the end, in an effort to avoid this. Today, the right to use any browser you like has essentially won out against the right of websites to restrict them, although the rise of dedicated smartphone apps has put some of the power back in companies’ hands.

Nested within this battle was another: ad blockers. Once your browser has fetched the HTML code for a website, should you be obligated to display it all on your screen? As websites began displaying adverts to make money, people began downloading ad blocker add-ons to take back control of their browser. In response, companies developed new techniques for displaying ads, and ad blockers adjusted to remove them. Today there is a stalemate, with .

Finally, there is the grandfather of all internet copyright battles: Napster, a peer-to-peer file-sharing service that people used to illegally share music files. The companies mostly won this battle thanks to the rise of Spotify and other similar music streaming services – with an easy legal option, piracy faded away, the side effect being that most people generally don’t own the music they listen to anymore.

Internet users largely prevailed after the browser wars, ad blocking is at an impasse and streaming music seems to have benefitted the big corporations. What does all of this mean for AI?

The difference here is decentralisation: just as I have used AI to create my own Goodreads app, so could anyone else, each of them slightly different. Fully blocking them all without restricting normal human access would be a monumental task. There is a slight caveat here – most AI use is centralised through the big providers like OpenAI and Anthropic, and Amazon could theoretically ask them to build in a block against creating such software (or sue them), but I think that is unlikely. In any case, the availability of open-weight AI models that anyone can download and run on their own hardware makes that a losing battle.

I think it is most likely that we will be heading for a mix of all three. When it comes to remixing software, anything you do in the privacy of your own computer is going to be very hard to stop. Once you release it online, however, expect the lawyers to hunt you down. What we might see is the sharing of prompts that will get an AI to do the work for you in private – writing “recreate Photoshop, make no mistakes” isn’t something that any lawyer can stop – and there already exists a slightly more complicated version of this in the form of an instruction set for AI called . As with browsers, individual choice will win, but this time the choices will be even more granular.

For software that goes out into the wider internet, including AI agents, there will be two tiers. I expect Meta and Amazon will eventually come to some kind of agreement that allows Muse to shop for you. We are already seeing this with the development of a “personal agent protocol” that . For anyone using a commercial AI agent, what you can and can’t do will be decided by the companies, just as music streaming services dictate how you listen.

But why should you use a commercial AI agent? People like me with a little technical know-how and even less coding ability are already using AI to create their own independent agents. As AI gets better at writing software, the barrier to entry will get lower and lower. I think companies will just have to come to accept this, as they have done with ad blockers. In future, a decent percentage of users that look like humans will actually be custom-made bots in disguise, doing things solely for the benefit of their creator. These bots will be ungovernable, for good and for ill.

Further AI reading

  • Anthropic has used its Claude AI model to solve a . In the 1990s, the US National Security Agency (NSA) chose five specific mathematical objects called elliptic curves to serve as the basis for a cryptography standard used to generate digital signatures, now known as ECDSA. No one knew how these curves were chosen, but cryptography lore suggests that an , perhaps “Jerry deserves a raise”, to generate them. In fact, Claude was able to guess the specific mechanism – random number generation based on a timestamp – and regenerate the curves. The finding has no implications for cybersecurity, it merely answers a long-running question.
  • Spotify says from recommendations on the platform, as part of a range of efforts to put “human artistry first”, according to the firm’s announcement. Other measures include a format for disclosing the use of AI in producing a track, or verifying an artist as human. Similarly, video game store Steam seems to have been at its latest Next Fest demo event.
  • The majority of people in the US say that AI is a bad thing for society, according to a . The survey found that 57 per cent of people in the US share this view, while 42 per cent say it is good for society. I’m not quite sure what to make of these numbers, though, given “AI” can mean many different things to different people.
  • A comment piece in Nature argues that . I find these arguments fairly convincing – I already believe our current AI models are definitely not conscious – but I hesitate to make firmer predictions about what’s to come.

]]>
2595251
The roots of AI fear in the West are cultural, not technological /article/2594390-the-roots-of-ai-fear-in-the-west-are-cultural-not-technological/?utm_campaign=RSS|NSNS&utm_content=artificial-intelligence&utm_medium=RSS&utm_source=NSNS Tue, 20 Oct 2026 08:00:00 +0000 /article/2594390-auto-draft/ 2594390 Russia and China are using AI to stage relentless cyberattacks against the UK /article/2593414-russia-and-china-are-using-ai-to-stage-relentless-cyber-attacks-against-the-uk/?utm_campaign=RSS|NSNS&utm_content=artificial-intelligence&utm_medium=RSS&utm_source=NSNS Thu, 15 Oct 2026 06:00:00 +0000 /article/2593414-auto-draft/ 2593414