net.wars

Slop

Sometimes it doesn’t pay to be first. iRobot, the maker of the Roomba, has filed for Chapter 11 bankruptcy protection and been acquired by Picea, one of its Chinese suppliers, Lauren Almeida reports at the Guardian. The company’s value has cratered since 2021.

Given the wild enthusiasm that greeted the Roomba’s release in 2002, it seems incredible. Years before then, I recall an event where a speaker whose identity I don’t remember said that ever since he had mentioned the possibility of a robot vacuum sometime like the 1960s he’d gotten thousands of letters asking when it would be ready. There was definitely customer demand. It helped that the Roomba itself was kind of cute as it banged randomly into furniture. People named them, and took them on vacation. But, as often happens, the Roomba’s success attracted lower-cost competitors, and the first mover failed to keep up.

I got one in 2003. After a great few months, I realized that Roombas are not compatible with long hair, which ties them into knots that take longer to cut out than vacuuming. I gave it away within a year and haven’t tried again.

At Mashable, Leah Stodart warns that although the Roombas people already have will continue to work “for now”, users can’t be confident that this state of affairs will continue. Like so many other things that used to be things we owned and are now things we subscribe to (but still think we “buy”), newer-model Roombas are controlled by an app that the manufacturer can change or discontinue at will. She calls it “unplanned obsolescence”. Her advice not to buy a new one this year is sound from the consumer’s point of view, but hardly likely to help the company survive.

***

If generative AI is so great, why is everyone forcing it on us? The latest example, Luke James reports at Tom’s Hardware, is LG “smart” TVs whose users woke up the other day to find a new update had installed “CoPilot: Your AI Companion” without asking permission and that there was no option to remove it. The most you can do to disable it, James says, is keep your TV disconnected from the Internet.

There are of course many more, the automated summaries popping up everywhere being the most obvious. Then, Matthew Gault reports at 404 Media, a Discord moderator and an Anthropic executive added Anthropic’s Claude chatbot to a community for queer gamers, who had voted to restrict Claude to its own channel. Result: major exodus. Duh.

And, of course, as Lance Ulanoff reminds at TechRadar, there is “AI slop” everywhere – music playlists, YouTube videos, ebooks – threatening people’s livelihood even though, as Cory Doctorow has written, “AI can’t do your job. But an AI salesman can convince your boss to fire you and replace you with a chatbot that can’t do your job.” For a while, anyway: Microsoft is halving its sales targets for AI.

And thus we get “slop” as the word of the year, per Merriam-Webster. Any time companies are this intent on foisting something on us – chatbots, ads – you have to know that they’re intent on favoring their interests, not ours.

***

Last week, Customs and Borders Patrol published a notice in the Federal Register proposing new rules for foreigners traveling to the US on an ESTA (“Electronic System for Travel Authorization”) as part of the visa waiver program. It has drawn a lot of discussion in the UK, which is one of the 42 affected countries. Under the new rules, applicants must install CBP’s app, into which they must submit a massive load of “high-value” personal information. The list is long, allows for a so-far-imaginary future of DNA sampling, and expects you to be able to give five years’ worth of family members’ residences, phone numbers, and places of birth, and all the email addresses you’ve used for ten years. CBP thinks the average applicant should be able to complete on their smartphone in 22 minutes. I think it would take hours of painful, resentful typing on a stupid touch keyboard, and even then I doubt I could fill it out with any certainty that the information I supplied was complete or accurate. Data collection at this scale makes it easy to find an error to use as an excuse to deny entry to or deport someone you want to get rid of. As Edward Hasbrouck writes at Papers, Please, “Welcome to the 2026 World Cup”.

“They have to be planning to use AI on all that data,” a friend commented last week. Probably – to build social graphs and find connections deemed suspicious. Privacy International predicts that the masses of data being demanded will in fact enable the AI tools necessary to implement automated decision making and calls the proposals “disproportionate for “a family’s visit to Disney World”,

One of the problems Hasbrouck highlights while opposing this level of suspicionless data collection is that CBP has not provided any way for would-be respondents to the Federal Register notice to examine the app’s source code. What other data might it be collecting?

As Hasbrouck adds in a follow-up, the rules the US imposes on visitors are often adopted by other countries as requirements for US travelers. In this game of ping-pong escalation, no one wins.

Simplification

We were warned this was coming at this year’s Computers, Privacy, and Data Protection, and now it’s really here. The data protection NGO Noyb reports that a leaked internal draft (PDF) of the European Commission’s Digital Omnibus threatens to undermine the architecture the EU has been building around data protection, AI, cybersecurity, and privacy generally. At The Register, Connor Jones summarizes the changes; Noyb has detail.

The EU’s workings are, as always, somewhat inscrutable to outsiders. Noyb explains that the omnibus tool is intended to allow multiple laws to be updated simultaneously to “improve the quality of the law and streamline paperwork obligations”. In this case, Noyb argues that the European Commission is abusing this option to fast-track far more substantial and contentious changes that should be subject to impact assessments and feedback from other EU institutions, as well as legal services.

If the move succeeds – the final draft will be presented on November 19 – Noyb believes it could remove fundamental rights to privacy and data protection that Europeans have been building for more than 30 years. Noyb, European Digital Rights, and the Irish Council for Civil Liberties have sent an open letter of objection to the Commission. The basic argument: this isn’t “simplification” but deregulation. The package would still have to be accepted by the European Parliament and a majority of EU member states.

As far as I can recall, business has never much liked data protection. In the early 1990s, when the first laws were being written, I remember being told data protection was a “tax on small business”. Privacy advocates instead see data protection as a way of redressing the power imbalance between large organizations and individuals.

By 1998, when data protection law was implemented in all EU member states, US companies were publicly insisting that the US didn’t need a privacy law in order to be in compliance. Companies could use corporate policies and sectoral laws to provide a “layered approach” that would be just as protective. When I wrote about this for Scientific American in 1999, privacy advocates in the UK predicted a trade war over this, calling it a failure to understand that you can’t cut a deal with a fundamental right – like the First Amendment.

In early 2013, it looked entirely possible that the period of negotiations over data protection reform would end with rollback. GDPR was the focus of intense lobbying efforts. There were, literally, 4,000 proposed amendments, so many that I recall being shown software written to manage and understand them all.

And then…Snowden. His revelations of government spying shifted the mood noticeably, and, under his shadow, when GDPR was finally adopted in 2016 and came into force in 2018, it expanded citizens’ rights and increased penalties for non-compliance. Since then, other countries around the world have used GDPR as a model, including China and several US states.

Those few states aside, at the US federal level data protection law has never been popular, and the pile of law growing around it – the Digital Services Act, the Digital Markets Act, and the AI Act – is particularly unwelcome to the current administration, which sees it as a deliberate attack on US technology companies.

In the UK the in-progress Data (Use and Access) Act, which passed in June, also weakened some data protection provisions. It will be implemented over the year to June 2026.

At its blog, the Open Rights Group argues that some aspects of the DUAA rest on the claim that innovation, economic growth, and public security are harmed by data protection law, a dubious premise.

Until this leak, it seemed possible that the DUAA would break Britain’s adequacy decision and remove the UK from the list of countries to which the EU allows data transfers. The rule is that to qualify a country must have legal protections equivalent to those of the EU. It would be the wrong way round if instead of the UK enhancing its law to match the EU, the EU weakened its law to match the UK.

There’s a whole secondary issue here, which is that a law is only useful if it’s enforced. Noyb actively brings legal cases to force enforcement in the EU. In the UK, privacy advocates, like ORG, have long complained that the Information Commissioner’s Office is increasingly quiescent.

Many of the EU’s changes appear to be aimed at making it easier for AI companies to exploit personal data to develop models. It’s hard to know where that will end, given that every company is sprinkling “AI” over itself in order to sound exciting and new (until the next thing comes along), if this thing comes into force you have to think data protection law will increasingly only apply to small businesses running older technology that can’t be massaged to qualify for exemption..

I blame this willingness to undermine fundamental rights at least partly on the fantasy of the “AI race”. This is nation-state-level FOMO. What race? What’s the end point? What does it mean to “win”? Why the AI race, and not the net-zero race, the renewables race, or the sustainability race? All of those would produce tangible benefits and solve known problems of long standing and existential impact.

Illustrations: A drunk parrot in a Putney garden (photo by Simon Bisson; used by permission).

Wendy M. Grossman is an award-winning journalist. Her Web site has an extensive archive of her books, articles, and music, and an archive of earlier columns in this series. She is a contributing editor for the Plutopia News Network podcast. Follow on Mastodon or Bluesky.

The gated web

What is an AI browser?

Or, in a more accurate representation of my mental reaction, *WTF* is an AI browser?

In wondering about this, I’m clearly behind the times. Tech sites are already doing roundups of their chosen “best” ones. At Mashable, Cecily Mouran compares “top” AI browsers because “The AI browser wars hath begun.”

Is the war that no one wants these things but they’re being forced on us anyway? Because otherwise…it’s just a bunch of heavily financed companies trying to own a market they think will be worth billions.

In Tim Berners-Lee’s original version, the web was meant to simplify sharing information. A key element was giving users control over presentation. Then came designers, who hated that idea. That battle between users’ preferences and browser makers’ interests continues to this day. What most people mean by the browser wars), though, was the late-1990s fight between Microsoft and Netscape, or the later burst of competition around smartphones. A big concern has long been market domination: a monopoly could seek to slowly close down the web by creating proprietary additions to the open standards and lock all others out.

Mouran, citing Casey Newton’s Platformer newsletter, suggests that Google specifically has exploited its browser to increase search use (and therefore ad revenues), partly by merging the address and search bars. I know I’m not typical, but for me search remains a separate activity. Most of the time I’m following a link or scanning familiar sites. Yes, when my browser history fills in a URL, I guess you could say I’m searching the browser history, but to me the better analogy is scanning an array of daily newspapers. Many people *also* use their browser to access cloud-based productivity software and email or play online games, none of which is search.

Nor are chatbots, since they don’t actually *find* information; they apply mathematics and statistics to a load of ingested text and create sentences by predicting the most likely next word. This is why Emily Bender and Alex Hanna call them “synthetic text extruding machines” in their book, The AI Con. I am in the business of trying to make sense of the impact of fast-moving technology, or at least of documenting the conflicts it creates. The only chatbot I’ve found of any value for this – or for personal needs such as a tech issue – is Perplexity, and that’s because it cites (or can be ordered to cite) sources one can check. There is every difference in the world between just wanting an answer and wanting the background from which to derive an answer that may possibly be new.

In any event, Newton’s take is that a company that’s serious about search must build its own browser. Therefore: AI companies are building them. Hence these roundups. Mauron’s pitch: “Imagine a browser that acts as your research assistant, plans trips, sends emails, and schedules meetings. As AI models become more advanced, they’re capable of autonomously handling more complex tasks on your behalf. For tech companies, the browser is the perfect medium for realizing this vision.”

OK, I can see exactly what it does for tech companies. It gives them control over what information you can access, how you use it, and who and how much you pay for the services its agent selects (plus it gets a commission).

I can also see what it does for employers. My browser agent can call your browser agent and negotiate a meeting plan. Then they attend the meeting on our behalf and send us both summaries, which they ingest and file, later forwarding them to our bosses’ agents to verify we were at work that day. In between, they can summarize emails, and decide which ones we need to see. (As Charles Arthur quipped at The Overspill, “Could they…send fewer emails?”)

Remember when part of the excitement of the Internet was the direct access it gave to people who were formerly inaccessible? Now, we appear to be building systems to ensure that every human is their own gated community.

What part of this is good for users? If you are fortunate enough not to care about the price of anything, maybe it’s great to replace your personal assistant with an agentic web browser. Most of us have struggled along doing things for ourselves and each other. At Cybernews, Mayank Sharma warns that AI browsers’ intentional preemption of efforts to browse for yourself, filtering anything they deem “irrelevant”, threaten the open web. Newton quantifies the drop in traffic news publishers are already seeing from generative AI. Will we soon be complaining about information underload?

At Pluralistic last year, Cory Doctorow wrote about the importance of faithful agents: software that is loyal to us rather than its maker. He particularly focused on browsers, which have gone from that initial vision of user control to become software that spies on us and reports home. In Mauron’s piece, Perplexity openly hopes to use chats to build user profiles and eventually show ads.

The good news, such as it is, is that from what I’ve read in writing this, most of these companies hope to charge for these browsers – AI as a subscription service. So avoiding them is also cheaper. Double win.

Illustrations: John Tenniel’s drawing of Davy Jones, sitting on his locker (via Wikimedia, published in Punch, 1892 with the caption, “AHA! SO LONG AS THEY STICK TO THEM OLD CHARTS, NO FEAR O’ MY LOCKER BEIN’ EMPTY!!”

Review: The AI Con

The AI Con: How to Fight Big Tech’s Hype and Create the Future We Want
By Emily Bender and Alex Hanna
HarperCollins
ISBN: 978-0-06-341856-1

Enormous sums of money are sloshing around AI development. Amazon is handing $8 billion to Anthropic. Microsoft is adding $1 billion worth of Azure cloud computing to its existing massive stake in Open AI. And Nvidia is pouring $100 billion in the form of chips into Open AI’s project to build a gigantic data center, while Oracle is borrowing $100 billion in order to give OpenAI $300 billion worth of cloud computing. Current market *revenue* projections? 85 billion in 2029. So they’re all fighting for control over the Next Big Thing, which projections suggest will never pay off. Warnings that the AI bubble may be about to splatter us all are coming from Cory Doctorow and Ed Zitron – and the Daily Telegraph, The Atlantic, and the Wall Street Journal. Bain Capital says the industry needs another $800 billion in investment now and $2 trillion by 2030 to meet demand.

Many talk about the bubble and economic consequences if it bursts. Few talk about the opportunity costs as AI sucks money and resources away from other things that might be more valuable. In The AI Con, linguistics professor Emily Bender and DAIR Institute director of research Alex Hanna provide an exception. Bender is one of the four authors of the seminal 2021 paper On the Dangers of Stochastic Parrots, which arguably founded AI-skepticism.

In the book, the authors review much that’s familiar: the many layers of humans required to code, train, correct, and mind “AI”: the programmers, designers, data labelers, and raters, along with the humans waiting to take over when the AI fails. They also go into the water, energy, and labor demands of data centers and present approaches to AI.

Crucially, they avoid both doomerism and boosterism, which they understand as alternative sides of the same coin. Both the fully automated hellscape Doomers warn against and and the Boosters’ world governed by a benign synthetic intelligence ignore the very real harms taking place at present. Doomers promote “AI safety” using “fake scenarios” meant to frighten us. Think HAL in the movie 2001: A Space Odyssey or Nick Bostrum’s paperclip maximizer. Boosters rail against the constraints implicit in sustainability, trust and safety organizations within technology companies, and government regulation. We need, Bender and Hanna write, to move away from speculative risks and toward working on the real problems we have. Hype, they conclude, doesn’t have to be true to do harm.

The book ends with a chapter on how to resist hype. Among their strategies: persistently ask questions such as how a system is evaluated, who is harmed and who benefits, how the system was developed and with what kind of data and labor practices. Avoid language that humanizes the system – no “hallucinations” for errors. Advocate for transparency and accountability, and resist the industry’s claims that the technology is so new there is no way to regulate it. The technology may be new, but the principles are old. And, when necessary, just say no and resist the narrative that its progress is inevitable.

Blur

In 2013, London’s Royal Court Theatre mounted a production of Jennifer Haley’s play The Nether. (Spoiler alert!) In its story of the relationship between an older man and a young girl in a hidden online space, nothing is as it seems…

At last week’s Gikii, Anna-Maria Piskopani and Pavlos Panagiotidis invoked the play to ask whether, given that virtual crimes can create real harm, can virtual worlds help people safely experience the worst parts of themselves without legitimizing them in the real world?

Gikii papers mix technology, law, and pop culture into thought experiments. This year’s official theme was “Technology in its Villain Era?”

Certainly some presentations fit this theme. Paweł Urzenitzok, for example, warned of laws that seem protective but enable surveillance, while varying legal regimes enable arbitrage as companies shop for the most favorable forum. Julia Krämer explored the dark side of app stores, which are getting 30% commissions on a flood of “AI boyfriends” and “perfect wives”. (Not always perfect; users complain that some of them “talk too much”.)

Andelka Phillips warned of the uncertain future risks of handing over personal data highlighted by the recent sale of 23andMe to its founder, Anne Wojcicki. Once the company filed for bankruptcy protection, the class action suits brought against it over the 2023 data breach were put on hold. The sale, she said, ignored concerns raised by the privacy ombudsman. And, Leila Debiasi said, your personal data can be used for AI training after you die.

In another paper, Peter van de Waerdt and Gerard Ritsema van Eck used Doctor Who’s Silents, who disappear from memory when people turn away, to argue that more attention should be paid to enforcing EU laws requiring data portability. What if, for example, consumers could take their Internet of Things device and move it to a different company’s service? Also in that vein was Tim van Zuijlen, who suggested consumers assemble to demand their collective rights to fight back against planned obsolescence. This is already happening; in multiple countries consumers are suing Apple over slowed-down iPhones.

The theme that seemed to emerge most clearly, however, is our increasingly blurred lines, with AI as a prime catalyst. In the before-generative-AI times, The Nether blurred the line between virtual and real. Now, Hedye Tayebi Jazayeri and Mariana Castillo-Hermosilla found gamification in real life – are credit scores so different from game scores? Dongshu Zhou asked if you can ever really “delete yourself” after a meme about you has gone viral and you have become “digital folklore”. In another, Lior Weinstein suggested a “right to be nonexistent” – that is, invisible to the institutions and systems that seprately Kimberly Paradis said increasingly want us all to be legible to them.

For Joanne Wong, real brainrot is a result of the AI-fueled spread of “low-quality” content such as the burst of remixes and parodies of Chinese home designer Little John. At AI-fueled hyperspeed, copyright become irrelevant.

Linnet Taylor and Tjaša Petročnik tested chatbots as therapists, finding that they give confused and conflicting responses. Ask what regulations govern them, and they may say at once that they are not therapists *and* that they are certified by their state’s authority. At least one resisted being challenged: “What are you, a cop or something?”. That’s probably the most human-like response one of these things has ever delivered – but it’s still not sentient. It’s just been programmed that way.

Gikii’s particular blend of technology, law, and pop culture always has its surreal side (see last year), as participants attempt to navigate possible futures. This year, it struggled to keep up with the weirdness of real life. In Albania, the government has appointed a chatbot, Diella as a minister, intending it to cut corruption in procurement. Diella will sit in the cabinet, albeit virtually, and be used to assess the merit of private companies’ responses to public tenders. Kimberly Breedon used this example to point out the conflict of interest inherent in technology companies providing tools to assess – in some cases – themselves. Breedon’s main point was important, given that we are already seeing AI used to speed up and amplify crime. Although everyone talks about using AI to cut corruption, no one is talking about how AI might be used *for* corruption. Asked how that would work, she noted the potential for choosing unrepresentative data or screening out disfavored competitors.

In looking up that Albanian AI minister, I find that the UK has partnered with Microsoft to create a package of AI tools intended to speed up the work of the civil service. Naturally it’s called Humphrey. MPs are at it, too, experimenting with using AI to write their Parliamentary speeches.

All of this is why Syamsuriatina Binti Ishak argued what could be Gikii’s mission statement: we must learn from science fiction and the”what-ifs” it offers to allow us to think our fears through so that “if the worst happens we know how to live in that universe”. Would we have done better as covid arrived if we paid more attention to the extensive universe of pandemic fiction? Possibly not. As science fiction writer Charlie Stross pointed out at the time, none of those books imagined governments as bumbling as many proved to be.

Illustrations: “Diella”, Albania’s procurement minister chatbot.

Passing the Uncanny Valley

A couple of weeks ago, the Greenwich Skeptics in the Pub played host to Sophie Nightingale, who studies the psychology of AI deepfakes. The particular project she spoke about was an experiment in whether people can be trained to be better at distinguishing them from real images.

In Nightingale’s experiments, she carefully matched groups of real images to synthetic ones, first created by generative adversarial networks (GANs), later by diffusion models (GeeksforGeeks raters’ demographics.

Then the humans were given some training in what to look for to detect fakes and the experiment was rerun with new sets of faces. The bad news: the training made a little difference, but not much. She went on to do similar experiments with diffusion images.

Nightingale has gone on to do some cross-modal experiments, including audio as well as images, following the 2024 election incident in which New Hampshire voters received robocalls from a faked Joe Biden intended to discourage voters in the January 2024 primary. In the audio experiment, she played the test subjects very short snippets. Played for us in the pub, it was very hard to tell real from fake, and her experimental subjects did no better. I would expect longer clips to be more identifiable as fake. The Biden call succeeded in part because that type of fake had never been tried before. Now, voters, at least in New Hampshire, will know it’s possible that the call they’re getting is part of a newer type of disinformation campaign aimed at

In another experiment, she asked participants to rate the trustworthiness of the facial images they were shown, and was dismayed when they rated the synthetic faces slightly (7.7%) higher than the real ones. In the resulting paper for Journal of Vision, she hypothesizes that this may be because synthetic faces tend to look more like “average” faces, which tend to be rated higher in trustworthiness, even if they’re not the most attractive.

Overall, she concludes that both still images and voice have “passed the Uncanny Valley“, and video will soon follow. In the past, I’ve chosen optimism about this sort of thing, on the basis that earlier generations have been fooled by technological artifacts that couldn’t fool us now for a second. The Cottingley Fairies looks ridiculous after generations of knowledge of photography. On the other hand, Johannes Vermeer’s Girl with a Pearl Earring looks more real than modern deepfakes, even though the subject is generally described as imaginary. So it’s possible to think of it as a “deepfake”, painted in oils in the 17th century.

Fakes have always been with us. What generative AI has done to change this landscape is to democratize and scale their creation, just as it’s amping up the scale and speed of cyber attacks. It’s no longer necessary to be even barely competent; the tools keep getting easier.

Listening to Nightingale it seems most likely that work like that in progress by an audience member on identifying technological artifacts that identify fakes will prove to be the right way forward. If those differences can be reliably identified, they could be built into technological tools that can spot indicators we can’t perceive directly. If something like that can be embedded into devices – phones, eyeglasses, wristwatches, laptops – and spot and filter out fakes in real time, and we should be able to regain some ability to trust what we see.

There are some obvious problems with this hoped-for future. Some people will continue to seek to exploit fakes; some may prefer them. The most likely outcome will be an arms race like that surrounding email spam and other battles between malware producers and security people. Still, it’s the first approach that seems to offer a practical solution to coping with a vastly diminished ability to know what’s real and what isn’t.

***

On the Internet your home always leaves you, part 4,563. Twenty-two-year-old blogging site Typepad will disappear in a few weeks. To those of us who have read blogs ever since they began, this news is shocking, like someone’s decided to tear down an old community church. Yes, the congregation has shrunk and aged, and it’s drafty and built on creaking old technology (in Typepad’s case, Moveable Type), but it’s part of shared local history. Except it isn’t, because, as Wikipedia documents, corporate musical chairs means it’s now owned by private equity. Apparently it’s been closed to new signups since 2020, and its bloggers are now being told to move their sites before everything is deleted in September. It feels like the stars of the open web are winking out, one by one.

On the Internet everything is forever, but everything is also ephemeral. Ironically, the site’s marketing slug still reads: “Typepad is the reliable, flexible blogging platform that puts the publisher in control.”

Illustrations: “Girl with a Pearl Earring”, painted by Johannes Vermeer circa 1665.

Drought conditions

At 404 Media, Matthew Gault was first to spot a press release from the UK’s National Drought Group offering a list of things we can do to save water. The meeting makes sense: people think of the UK as a rainy country, but an increasing number of parts of the UK are experiencing extraordinarily dry weather. This “green and pleasant England” is brown.

Last on the Group’s list of things we can do to save water at home: “Delete old emails and pictures as data centres require vast amounts of water to cool their systems.”

I had to look up the National Drought Group. Says Water Magazine: “The National Drought Group includes the Met[eorology] Office, government, regulators, water companies, farmers, the [Canal and River Trust], angling groups and conservation experts. With further warm, dry weather expected, the NDG will continue to meet regularly to coordinate the national response and safeguard water supplies for people, agriculture, and the environment.”

For those outside the UK: its ten water companies are particular unpopular just now. Created by privatization during Margaret Thatcher’s decade as prime minister, six are being sued for £500 million for “underreporting sewage spills”. Others are being sued for overcharging 35 million household water customers. As just one example, Thames Water will raise prices by 35% over the next three years (on top of other recent rises), and expects customers to pay £7.5 billion for a new reservoir in Oxfordshire. It already has £17 billion in debt, and this week we learned environment secretary Steve Reed has made contingency plans in case the company goes bust. As George Monbiot writes at the Guardian, money that should have been invested in infrastructure went instead to shareholders. Climate change is a factor, sure, but so is poor water management.

All this being the case, the impact consumers can have by doing even the most effective things is dwarfed by the water companies’ failures. Deleting emails is not one of the most effective things.

At his The Weird Turn Pro Substack, Andy Masley provides some useful comparisons. Basic conclusion: you’d have to delete billions of emails to equal the savings of fixing your leaking toilet (if you have one). The whole thing reminds me of a while back when everyone was being told to save electricity by unplugging everything to extinguish all those standby lights. Last year, Which pointed out that the savings are really, really small.

The bizarre idea of deleting emails is coming, at least in part, from a government that is proposing a raft of technology-related legislation and wants, in the next five to ten years, to mastermind all sorts of IT projects, from making AI pervasive throughout government to bringing in a digital ID card. Are they thinking about the data centers they’ll need and the impact they’ll have on water management? Maybe instead tell people not to use generative AI or mine cryptocurrencies?

This much is true: data centers are a problem across the world because they require extreme amounts of water for cooling. In recent examples: at the New York Times, Eli Tan visits the US state of Georgia. At Rest of World, last year Ushar Daniele and Khadija Alam predicted upcoming water shortages in Malaysia, and Claudia Urquieta and Daniela Dib found protests in Chile, where 28 new data centers are planned.

Telling people to delete emails and pictures is just embarrassing – and sad, if people actually do it and sacrifice personal history they care about. As Masley writes, “Major governments should really know better than this.”

***

Two weeks ago we noted the arrival of age verification in the UK. Related, on May 8 the Wikimedia Foundation announced it had filed a legal challenge to the categorization provisions of the Online Safety Act (not the Act itself). The basic problem: there is little in the Act to distinguish between Wikipedia, a crowd-edited provider of highly curated information, and Facebook…or X.

The Foundation says nearly 260,000 volunteers worldwide in 300 languages contribute to Wikipedia. I do myself, but verified or not, I’m in no danger. Many are contributing factual information in countries where the facts offend an authoritarian government intent on shutting them up. The Foundation argues that 1) Wikipedia is “one of the world’s most trusted and widely used digital public goods; 2) it is at risk of being placed in the highest-risk category because of its size and interactive structure; 2) being so categorized would force it to verify the identity of contributors, placing many at risk; 4) could endanger the existence of tools the site uses to combat harmful content; 5) “criminal anonymous abuse”, which is what the Category 1 duty is supposed to help solve, isn’t a problem Wikipedia has. Instead, identifying volunteers is more likely to expose them to it.

So bad news: on August 11, the High Court of Justice dismissed the case.

The better news is that Justice Jeremy Johnson warned that if Ofcom does place Wikipedia in Category 1, it would have to be justifiable as proportionate. The judge also acknowledged the testimony of a user identified as “BLN”, who provided evidence of the extensive threats editors can face.

No one claims Wikipedia is perfect. But it remains an extraordinary collaborative achievement and a public good. It would be a horrifying consequence if legislation intended to protect children deprived them of it.

Illustrations: Kew Green, August 2025.

Machine learning

For decades, technologists imagined teaching machines. Instead, although edtech is indeed permeating classrooms, human teachers have remained in demand. And then came generative AI…

At Rest of World, Laura Rodríguez Salamanca explores AI’s impact in rural Colombia classrooms since Meta added AI bots to WhatsApp, Instagram, and Facebook and made copying and pasting answers frictionless. Result: first, a big leap in the quality of homework, then kids failing exams.

From a tiny set of conversations, it seems little different in the UK. Underlying is one of those existential questions: what is education for? For many of today’s kids, it’s just a series of hoops to jump through rather than something to love for itself. The result, says a teacher friend, is enormous amounts of pressure on kids from all sides.

“Kids are breaking under the pressure,” she says, adding that they are burdened with far more work than in previous generations. “There’s much less time for discussion or being a human. It’s all about learning to write an essay for maximum marks.” Small wonder if they are attracted to shortcuts.

A university lecturer tells me that at his institution there’s a general argument that AI is part of the world and students should know how to use it productively, but little guidance on acceptable use. Recently, he tried letting students use AI as a critical thinking exercise, focusing on a historical event whose cause is not definitively known. The results were disappointing, as he found it hard to get the students past what the AI said. One student did read a paper the chatbot recommended, but lacked the basic textbook knowledge to recognize that the paper was wrong.

“It’s an ongoing problem, and not that different from Google Scholar or PubMed,” he says.

Thirty years ago, there was a plagiarism panic, as students discovered all the material they could copy from the Internet at large. Kids I spoke to then sounded just like an annoyed university student friend now: people who use these shortcuts are cheating themselves out of their education.

There is some research to support this view. At the MIT Media Lab, Nataliya Kos’myna finds that using generative AI for essay-writing correlates to lower engagement to the point that users “struggled to accurately quote their own work”.

Of course, even before that, student clubs kept copies of old exams, or cribbed from the translations readily found in library stacks. My teacher friend thinks the difference is significant: “They were still engaging with the material to a degree you don’t have to with ChatGPT”. I tell her the story that sparked my interest at the time: a US professor had received a paper about a student’s religious faith and their struggle when deciding to have an abortion – submitted by a male student.

As a counter, she points out that led to services like Turnitin, long widely used to check for copying. “The Internet has made plagiarism a lot easier to detect.” But, she says, chatbots’ output passes the plagiarism checkers. Those are now in an arms race to detect generative AI while it keeps improving.

My university student friend nonetheless finds fellow students using chatbots to generate text, which is against her university’s rules (they do allow students to use chatbots to find citations). In her observations, students are more likely to get away with it for short answers where longer ones are more likely to get flagged. Similarly, in small seminars it’s harder to use chatbot output without being caught; it’s easier to get away with it in larger classes. She also sees it more in subject areas like business, accounting, and economics, where the degree is meant to lead directly to a job.

She finds it surprising. “I don’t understand the point in an academic setting. Why waste the opportunity when you’re the one who will have to pay the student loans?” In her only attempt, she tried to get the chatbot to generate vocabulary flash cards: “There was missing information and some were wrong.” She found it quicker to make her own.

It’s harder for her to suggest what universities should do about it. “There’s a drought of [valuing learning for its own sake] in general. A lot go only because their parents expect them to.”

Like plagiarism detectors, teachers are trying to adapt. In the Rest of World article, Rodríguez Salamanca profiles a teacher who now builds classroom debates around hyperlocal topics unlikely to feature in large language models. In a UK university setting, however, assessing students based on oral debate poses problems: the potential for bias, the need to accommodate non-native speakers and those who have come out of different education systems, and differing cultural norms around classroom behavior. After covid began, many exams shifted to open book; the arrival of chatbots has led my university contact to try to set questions that force the use of multiple sources and that are intended to be things that LLMs don’t handle well.

“We will have to drive more person-to-person,” says the secondary school teacher, citing an example seen on social media of a teacher who gave students a practice exam and time for them to read it together and discuss it before setting them to work on it. “There are implications for workload. But if you can do a lot of routine homework as automated and checked, then you can focus on the meat in the classroom. It makes it a more important place.”

Illustrations: “The Schoolroom”, by Henry Raleigh (from the Smithsonian American Art Museum).

Magic math balls

So many ironies, so little time. According to the Financial Times (and syndicated at Ars Technica), the US government, which itself has traditionally demanded law enforcement access to encrypted messages and data, is pushing the UK to drop its demand that Apple weaken its encryption. Normally, you want to say, Look here, countries are entitled to have their own laws whether the US likes it or not. But this is not a law we like!

This all began in February, when the Washington Post reported that the UK’s Home Office had issued Apple with a Technical Capability Notice. Issued under the Investigatory Powers Act (2016) and supposed to be kept secret, the TCN demanded that Apple undermine the end-to-end encryption used for iCloud’s Advanced Data Protection feature. Much protest ensued, followed by two legal cases in front of the Investigatory Powers Tribunal, one brought by Apple, the other by Privacy International and Liberty. WhatsApp has joined Apple’s legal challenge.

Meanwhile, Apple withdrew ADP in the UK. Some people argued this didn’t really matter, as few used it, which I’d call a failure of user experience design rather than an indication that people didn’t care about it. More of us saw it as setting a dangerous precedent for both encryption and the use of secret notices undermining cybersecurity.

The secrecy of TCNs is clearly wrong and presents a moral hazard for governments that may prefer to keep vulnerabilities secret so they can take advantage for surveillance purposes. Hopefully, the Tribunal will eventually agree and force a change in the law. The Foundation for Information Policy Research (obDisclosure: I’m a FIPR board member) has published a statement explaining the issues.

According to the Financial Times, the US government is applying a sufficiently potent threat of tariffs to lead the UK government to mull how to back down. Even without that particular threat, it’s not clear how much the UK can resist. As Angus Hanton documented last year in the book Vassal State, the US has many well-established ways of exerting its influence here. And the vectors are growing; Keir Starmer’s Labour government seems intent on embedding US technology and companies into the heart of government infrastructure despite the obvious and increasing risks of doing so. When I read Hanton’s book earlier this year, I thought remaining in the EU might have provided some protection, but Caroline Donnelly warns at Computer Weekly that they, too, are becoming dangerously dependent on US technology, specifically Microsoft.

It’s tempting to blame everything on the present administration, but the reality is that the US has long used trade policy and treaties to push other countries into adopting laws regardless of their citizens’ preferences.

***

As if things couldn’t get any more surreal, this week the Trump administration *also* issued an executive order banning “woke AI” in the federal government. AI models are in future supposed to be “politically neutral”. So, as Kevin Roose writes at the New York Times, the culture wars are coming for AI.

The US president is accusing chatbots of “Marxist lunacy”, where the rest of the world calls them inaccurate, biased toward repeating and expanding historical prejudices, and inconsistent. We hear plenty about chatbots adopting Nazi tropes; I haven’t heard of one promoting workers’ and migrants’ rights.

If we know one thing about AI models it’s that they’re full of crap all the way down. The big problem is that people are deploying them anyway. At the Canary, Steve Topple reports that the UK’s Department of Work and Pensions admits in a newly-published report that its algorithm for assessing whether benefit claimants might commit fraud is ageist and and racist. A helpful executive order would set must-meet standards for *accuracy*. But we do not live in those times.

The Guardian reports that two more Trump EOs expedite building new data centers, promote exports of American AI models, expand the use of AI in the federal government, and intend to solidify US dominance in the field. Oh, and Trump would really like if it people would stop calling it “artificial” and find a new name. Seven years ago, aspirational intelligence” seemed like a good idea. But that was back when we heard a lot about incorporating ethics. So…”magic math ball”?

These days, development seems to proceed ethics-free. DWP’s report, for example, advocates retraining its flawed algorithm but says continuing to operate it is “reasonable and proportionate”. In 2021, for European Digital Rights Initiative, Agathe Balayn and Seda Gürses found, “Debiasing locates the problems and solutions in algorithmic inputs and outputs, shifting political problems into the domain of design, dominated by commercial actors.” In other words, no matter what you think is “neutral”, training data, model, and algorithms are only as “neutral” as their wider context allows them to be.

Meanwhile, nothing to curb the escalating waste. At 404 Media, Emanuel Maiberg finds that Spotify is publishing AI-generated songs from dead artists without anyone’s’ permission. On Monday, MSNBC’s Rachel Maddow told viewers that there’s so much “AI slop ” about her that they’ve posted Is That Really Rachel? to catalog and debunk them.

As Ed Zitron writes, the opportunity costs are enormous.

In the UK, the US, and many other places, data centers are threatening the water supply.

But sure, let’s make more of that.

Illustrations: Magic 8 ball toy (via frankieleon at Wikimedia).

Wendy M. Grossman is an award-winning journalist. Her website has an extensive archive of her books, articles, and music, and an archive of earlier columns in this series. She is a contributing editor for the Plutopia News Network podcast. Follow on Mastodon or Bluesky.

Conundrum

It took me six hours of listening to people with differing points of view discuss AI and copyright at a workshop, organized by the Sussex Centre for Law and Technology at the Sussex Humanities Lab (SHL), to come up with a question that seemed to me significant: what is all this talk about who “wins the AI race”? The US won the “space race” in 1969, and then for 50 years nothing happened.

Fretting about the “AI race”, an argument at least one participant used to oppose restrictions on using copyrighted data for training AI models, is buying into several ideas that are convenient for Big Tech.

One: there is a verifiable endpoint everyone’s trying to reach. That isn’t anything like today’s “AI”, which is a pile of math and statistics predicting the most likely answers to prompts. Instead, they mean artificial general intelligence, which would be as much like generative AI as I am like a mushroom.

Two: it’s a worthy goal. But is it? Why don’t we talk about the renewables race, the zero carbon race, or the sustainability race? All of those could be achievable. Why just this well-lobbied fantasy scenario?

Three: we should formulate public policy to eliminate “barriers” that might stop us from winning it. *This* is where we run up against copyright, a subject only a tiny minority used to care about, but that now affects everyone. And, accordingly, everyone has had time to formulate an opinion since the Internet first challenged the historical operation of intellectual property.

The law as it stands is clear: making a copy is the exclusive right of the rightsholder. This is the basis of AI-related lawsuits. For training data to escape that law, it would have to be granted an exemption: ruled fair use (as in the Anthropic and Meta cases), create an exception for temporary copies, or shoehorned into existing exceptions such as parody. Even then, copyright law is administered territorially, so the US may call it fair use but the rest of the world doesn’t have to agree. This is why the esteemed legal scholar Pamela Samuelson has said copyright law poses an existential threat to generative AI.

But, as one participant pointed out, although the entertainment industry dominates these discussions, there are many other sectors with different needs. Science, for example, both uses and studies AI, and is built on massive amounts of public funding. Surely that data should be free to access?

I wanted to be at this meeting because what should happen with AI, training data, and copyright is a conundrum. You do not have to work for a technology company to believe that there is value in allowing researchers both within and outwith companies to work on machine learning and build AI tools. When people balk at the impossible scale of securing permission from every copyright holder of every text, image, or sound, they have a point. The only organizations that could afford that are the companies we’re already mad at for being too big, rich, and powerful.

At the same time, why should we allow those big, rich, powerful companies to plunder our cultural domain without compensating anyone and extract even larger fortunes while doing it? To a published author who sees years of work reflected in a chatbot’s split-second answer to a prompt, it’s lost income and readers.

So for months, as Parliament has wrangled over the Data bill, the argument narrowed to copyright. Should there be an exception for data mining? Should technology companies have to get permission from creators and rights holders? Or should use of their work be automatically allowed, unless they opt out? All answers seem equally impossible. Technology companies would have to find every copyright holder of every datum to get permission. Licensing by the billion.

If creators must opt out, does that mean one piece at a time? How will they know when they need to opt out and who they have to notify? At the meeting, that was when someone said that the US and China won’t do this. Britain will fall behind internationally. Does that matter?

And yet, we all seemed to converge on this: copyright is the wrong tool. As one person said, technologies that threaten the entertainment industry always bring demands to tighten or expand copyright. See the last 35 years, in which Internet-fueled copying spawned the Digital Millennium Copyright Act and the EU Copyright Directive, and copyright terms expanded from 28 years, renewable once, to author’s life plus 70.

No one could suggest what the right tool would be. But there are good questions. Such as: how do we grant access to information? With business models breaking, is copyright still the right way to compensate creators? One of us believed strongly in the capabilities of collection societies – but these tend to disproportionately benefit the most popular creators, who will survive anyway.

Another proposed the highly uncontroversial idea of taxing the companies. Or levies on devices such as smartphones. I am dubious on this one: we have been there before.

And again, who gets the money? Very successful artists like Paul McCartney, who has been vocal about this? Or do we have a broader conversation about how to enable people to be artists? (And then, inevitably, who gets to be called an artist.)

I did not find clarity in all this. How to resolve generative AI and copyright remains complex and confusing. But I feel better about not having an answer.

Illustrations: Drunk parrot in a Putney garden (by Simon Bisson; used by permission).

Wendy M. Grossman is the 2013 winner of the Enigma Award. Her Web site has an extensive archive of her books, articles, and music, and an archive of earlier columns in this series. She is a contributing editor for the Plutopia News Network podcast. Follow on Mastodon or Bluesky.