Technically Good ✨

Technically Good is back! With a slightly shorter edition. There's a note at the end of this newsletter if you're interested! --- # ✨ In This Edition ✨ 1. 📟 Short-Form Content: Slow it down!! It's too dangerous!!1!; malware; and "going rogue." 2. 🌞 Good News!: "No AI" seals as a sign of quality 🦭; people are flocking back to human writers. 3. 🐘 Toots and Memes and Such: Age verification, empowerment, and "going rogue" (reprise.) 4. 📯 The Post-Script: The "Inevitability" of AI. Glitched, cropped, and edited version of images by benjamin lehman and Aron Visuals on Unsplash. --- # 📟 Short-Form Content ## Chatbots ("AI") - Dario Amodei and other chatbot CEO's have decided to call for a "frontier model slowdown." Because the models are, like, too dangerous and too advanced and will kill us all and stuff like that. However, it remains to be seen whether investors and other ~stakeholders~ will be happy with any sort of slowdown. * As tante comments, new models cost a lot to develop, but don't feel meaningfully "better" to users, so a slowdown could be just what the doctor ordered for the bottom line. Which, uh. * Focusing on potential future dangers also has the convenient side effect of distracting from all the harms that LLMs are currently doing! And then there's the fact that calling AI "too dangerous" is just way cooler than admitting it doesn't work, to paraphrase The Beaverton. - Current AI discourse is, of course, dominated by pro-AI "CEO Said" journalism. That's where the "growth" and the "money" is. Which also means that, what used to be called "malware" or a virus - namely, a program that causes bad things to happen and spreads to machines it wasn't supposed to be on - now seems to be given the label of a "rogue AI." Which makes AI sound very scary and powerful, and not at all like it just wasn't programmed properly, and like the sandbox was maybe a lil porous. - Relatedly, and in what is surely shocking news to all of us, Meta's new Muse chatbot seems to ignore user permissions, and simply goes ahead and uploads things like Apple Messages to the cloud. Maybe exercise caution when you let it, uh, handle your fb Marketplace transactions for you? 😬 - As I've said before - don't let the machine de-skill you: Students who don't use AI are ahead on every measure in the long-running PISA assessment. Further to this, a UC Berkeley study has found that using LLMs actively erodes your ability to persist with difficult tasks. After only ten minutes! * At this point in the newsletter, I'd like to invite you to spend a couple of minutes thinking about whether a future in which LLM's have successfully de-skilled you, and your family and friends, and in which our collective ability to keep at hard things has been eroded, may be problematic - on both an individual and a societal scale. - AI crawlers and scrapers bringing open, community- and volunteer-maintained platforms to their knees by attempting to ingest all the knowledge that the community and its volunteers have contributed over years adds a special kind of insult to injury. ## Non-Chatbots - This is going to keep being a recurring theme: IDScan, an identity verification company, has confirmed that hackers have stolen 13 to 15 million of driver's licenses. As we've known for a while, it is not a question of whether, but when age-gating data gets stolen through hacks or leaks. * Note that it doesn't even need to be the platform for which you needed the age verification that gets hacked for your data to end up all over the web and in who-knows-whose hands, but rather the intermediary - the "verification provider" - which needs access to your ID to carry out the verification. - Canada's lawful access bill (C-22, which Technically Good has written about here) could be a barrier to a trade agreement between the EU and Canada, as it poses privacy concerns and may not be reconcilable with EU law. One of the points of contention seems to be the endangerment of e2ee (end-to-end encryption). - Flock cameras are big on privacy! For themselves. --- # 🌞 Good News! - Bad news that we need this, good news that publishers are standing up against AI: German childrens' book publisher Loewe is now adding a "No AI" label to their covers, as a sign of sustained high quality content. - Remember how AI boosterists have long claimed that professions like writing, editing, translating, and many more will definitely be replaced by LLMs very very soon? Well, we may just be seeing the opposite happen. Writer Mike Rosenberg is reporting that clients now approach him to "turn their brains back on." See also: The machine de-skilling its users. --- # 🐘 Toots and Memes and Such Chris Fletch on the one kind of age verification I can get behind: > *Plays audio recording of me getting out of a chair* There’s your age verification ... ok, maybe I'll also consider "what do you do if your SNES cartridge doesn't work?" and "does a pencil fit into the lil holes in a cassette tape? how do you know?" --- lcamtuf on "empowerment" through AI, which is how you're being made to believe that you're not being de-skilled by these chatbots: > Worried about AI? Don't be. Using agents at work simply empowers you to focus on more meaningful tasks, similarly to how self-checkout allowed cashiers to pursue higher-impact projects at the grocery store --- Matt Blaze on - hey, remember the bit on what we used to call "malware" above!? > “Our software went rogue!” No. You programmed it wrong. See also: LLM "guardrails" are just... text. Text that asks the LLM to pretty please not do a thing. --- # 📯 The Post-Script This is a fantastic short read on the "inevitability" of AI, by Lana. As a fellow child of the 80s and 90s, I was just talking about this with friends - nobody had to sell us email and the first instant messengers and brick-style mobile phones as "inevitable." We just all got one of those - in the case of physical items like a phone, as soon as we could afford to - because, like Lana says, these things solved a clear and distinct social problem. Those brick phones with a monochrome display, no internet access, and horribly screechy ringtones. (This was mine; it was magic. 💜) If you have to cram a technology into every text box, button, form, mailout, and help popup so that your users will use it, and if you have to tell them over and over again, like a broken record, that this is inevitable, there's a really good chance it's - not. --- And here's the end note that was promised: This edition has been slightly shorter as last month has been... eventful. :) That said, if any readers - or folks you know - are looking for a researcher, data analyst, tech writer, and/or all of the above rolled into one: I am available for project work, and hello [at] technically-good.ca is but a button-click away! --- Follow this blog via RSS | Find me on Mastodon | Bubbles --- ## Recent Posts - 11 Oct, 2026 💌 Newsletter #017 - Slow it down!! It's too Dangerous!!1! - 13 Sep, 2026 💌 Newsletter #016 - Chatbots and their "Safeguards," Alternative Navigation Apps - and a New Section! - 30 Aug, 2026 💌 Newsletter #015 - "Watermarks" in Chatbot Output; Meta settles; "It's Just a Tool?"

# ✨ In This Edition ✨ 1. 📟 Short-Form Content: Chatbots & their "safeguards," eavesdroppy "Smart TVs," and how age gating still is not it. 2. 🌞 Good News!: Alternative navigation apps exist!, AI bans in public schools, more digital sovereignty. 3. 🐘 Toots and Memes and Such: On valid sources, connecting people, and moving carefully. 4. 📯 The Post-Script: Blood in the Machine: The Origins of the Rebellion Against Big Tech, and Myths around AI and Labour. Original image by Tamas Tuzes-Katai on Unsplash. Glitched and cropped. --- # 📟 Short-Form Content ## Chatbots ("AI") - Things seem to be going very well over at OpenAI, which keeps bleeding executives. This time, it's the Head of Data Centres, totalling 13 (yes, thirteen) departed executives in 2026 alone. - Things are also going very well with Claude, whose "safeguards" apparently can - and have been - circumvented in the context of topics as lighthearted and good-for-humanity as "bioweapons." 😳 - All the "AI is here to stay" and the general cramming of the chatbots into everything is not making the people love the chatbots. - Muse, of Knights of Cydonia and Starlight and Madness and loads-of-other-bangers fame, lost their social media handle to Meta's new chatbot thingie. Sigh. Anyway, my favourite song of theirs is Uprising. - In more local, but not really?, news, aUS-linked network of fake websites is promoting Alberta separatism to AI chatbots. This is my surprised face: 😲 - The EFF is agreeing that training chatbots on copyrighted works is a-OK. I know that's an argument that's been kicking around, I'm just not sure... I agree with it. - In a refrain that seems to be repeating everywhere: Chatbots at work make volume of Stuff go up, and also make volume of incidents and firefighting time go up. What a deal! - (The following article contains detailed accounts of mental health concerns, a very slippery slope of AI usage, as well as suicide. Please tread carefully in reading & sharing this article if you are concerned that these topics may upset you or someone you share it with.) A very cautionary tale on "AI psychosis." ## Non-Chatbots - In the last newsletter, we learned how Meta settled in the massive lawsuit it was facing. The settlement is not good news, as it requires age gating technology for all users, and requires even more data to be collected than before. - In more news on how age gating is still not it: When age gating gets implemented, an ID needs to be scanned somewhere, and "somewhere" has a real high likelihood of also meaning "publicly accessible" down the line. See also here. Like, these will just keep coming, basically. - Smart TV's have been caught logging audio even when the screen was off, and snooping on local devices. Coooooool. --- # 🌞 Good News! - The EU is going to be tougher on regulating ChatGPT (among other platforms, including Roblox) going forward. * According to The Verge, this will include being held accountable for "mitigating risks related to ChatGPT’s impact on minors, user mental health, and the spread of illegal content." This comes after those platforms were designated "Very Large Online Search Engines." - New York City is banning AI in public schools in all grades prior to high school. This includes companion chatbots and student-facing generative AI and will affect around 600,000 students. - Non-Google map and navigation platforms exist! * There's likely not a single perfect one for all use cases, but you can check out MapQuest and OrganicMaps as possible alternatives. * I find that for cycling, OrganicMaps is actually better than Google Maps. And on MapQuest your lil icon can be a unicorn, so. - Non-Microsoft and non-Google office suite alternatives also exist! * Such as LibreOffice, which broke download records after stating it has no AI features. I use LibreOffice for all my office suite needs, and it's doing the trick. - For more digital sovereignty, give Todd Maffin's Unplug America project a look! Alternatives to US-based platforms and services across a ton of categories. - The young'uns apparently aren't super big fans of AI art, calling it "boomer art," and that made me chuckle. - This is weird territory now, as obviously kids suffering from asthma is never good news. But. The good news is that the people are speaking up, and there is a growing backlash forming against data centres & the negative effects they have on us as a society. * Such as in Memphis, where apparently zero of the 35 methane gas turbines used by that data centre are "quipped with pollution controls typically required by federal rules." Uh. * It's also really weird that the people don't love having data centres in their community when they come with a host of positive effects, such as <checks notes> threatening jobs, more expensive utilities, noise, and pollution. Truly mysterious. A baffler for the ages. --- # 🐘 Toots and Memes and Such That's right, a new section. As a millennial, I needed a place to share toots1 and memes and, well, such, that I've come across. You can, of course, also follow Technically Good on Mastodon and be served this kind of content when it's steaming fresh and piping hot. > "Growing up with the phrase “Wikipedia isn’t a valid source,” only to end up in a daily life where everyone says, “Just ask ChatGPT." by Andrea Junker > "Ai is going to lead to an abundant and prosperous future the same way that Facebook has connected me and brought me closer to my friends and family." by Brian Owen. As always: > "Move carefully and fix things." by datarama. --- # 📯 The Post-Script - I am finally reading Brian Merchant's Blood in the Machine: The Origins of the Rebellion Against Big Tech. A historical account of the Luddite movement that's weirdly relevant in 2026. - You should also check out Dr. Alex Hanna's video on Myths around AI and Labour. --- Follow this blog via RSS | Find me on Mastodon | Bubbles --- ## Recent Posts - 13 Sep, 2026 💌 Newsletter #016 - Chatbots and their "Safeguards," Alternative Navigation Apps - and a New Section! - 30 Aug, 2026 💌 Newsletter #015 - "Watermarks" in Chatbot Output; Meta settles; "It's Just a Tool?" - 16 Aug, 2026 💌 Newsletter #014 - Many, Many Leaks; Being "Opted In;" and Cursed "Podcasts" for your Children 1. yes, I still call them toots.↩

# ✨ In This Edition ✨ 1. 📟 Short-Form Content: Chatbots & their "safeguards," eavesdroppy "Smart TVs," and how age gating still is not it. 2. 🌞 Good News!: Alternative navigation apps exist!, AI bans in public schools, more digital sovereignty. 3. 🐘 Toots and Memes and Such: On valid sources, connecting people, and moving carefully. 4. 📯 The Post-Script: Blood in the Machine: The Origins of the Rebellion Against Big Tech, and Myths around AI and Labour. Original image by Tamas Tuzes-Katai on Unsplash. Glitched and cropped. --- # 📟 Short-Form Content ## Chatbots ("AI") - Things seem to be going very well over at OpenAI, which keeps bleeding executives. This time, it's the Head of Data Centres, totalling 13 (yes, thirteen) departed executives in 2026 alone. - Things are also going very well with Claude, whose "safeguards" apparently can - and have been - circumvented in the context of topics as lighthearted and good-for-humanity as "bioweapons." 😳 - All the "AI is here to stay" and the general cramming of the chatbots into everything is not making the people love the chatbots. - Muse, of Knights of Cydonia and Starlight and Madness and loads-of-other-bangers fame, lost their social media handle to Meta's new chatbot thingie. Sigh. Anyway, my favourite song of theirs is Uprising. - In more local, but not really?, news, aUS-linked network of fake websites is promoting Alberta separatism to AI chatbots. This is my surprised face: 😲 - The EFF is agreeing that training chatbots on copyrighted works is a-OK. I know that's an argument that's been kicking around, I'm just not sure... I agree with it. - In a refrain that seems to be repeating everywhere: Chatbots at work make volume of Stuff go up, and also make volume of incidents and firefighting time go up. What a deal! - (The following article contains detailed accounts of mental health concerns, a very slippery slope of AI usage, as well as suicide. Please tread carefully in reading & sharing this article if you are concerned that these topics may upset you or someone you share it with.) A very cautionary tale on "AI psychosis." ## Non-Chatbots - In the last newsletter, we learned how Meta settled in the massive lawsuit it was facing. The settlement is not good news, as it requires age gating technology for all users, and requires even more data to be collected than before. - In more news on how age gating is still not it: When age gating gets implemented, an ID needs to be scanned somewhere, and "somewhere" has a real high likelihood of also meaning "publicly accessible" down the line. See also here. Like, these will just keep coming, basically. - Smart TV's have been caught logging audio even when the screen was off, and snooping on local devices. Coooooool. --- # 🌞 Good News! - The EU is going to be tougher on regulating ChatGPT (among other platforms, including Roblox) going forward. * According to The Verge, this will include being held accountable for "mitigating risks related to ChatGPT’s impact on minors, user mental health, and the spread of illegal content." This comes after those platforms were designated "Very Large Online Search Engines." - New York City is banning AI in public schools in all grades prior to high school. This includes companion chatbots and student-facing generative AI and will affect around 600,000 students. - Non-Google map and navigation platforms exist! * There's likely not a single perfect one for all use cases, but you can check out MapQuest and OrganicMaps as possible alternatives. * I find that for cycling, OrganicMaps is actually better than Google Maps. And on MapQuest your lil icon can be a unicorn, so. - Non-Microsoft and non-Google office suite alternatives also exist! * Such as LibreOffice, which broke download records after stating it has no AI features. I use LibreOffice for all my office suite needs, and it's doing the trick. - For more digital sovereignty, give Todd Maffin's Unplug America project a look! Alternatives to US-based platforms and services across a ton of categories. - The young'uns apparently aren't super big fans of AI art, calling it "boomer art," and that made me chuckle. - This is weird territory now, as obviously kids suffering from asthma is never good news. But. The good news is that the people are speaking up, and there is a growing backlash forming against data centres & the negative effects they have on us as a society. * Such as in Memphis, where apparently zero of the 35 methane gas turbines used by that data centre are "quipped with pollution controls typically required by federal rules." Uh. * It's also really weird that the people don't love having data centres in their community when they come with a host of positive effects, such as <checks notes> threatening jobs, more expensive utilities, noise, and pollution. Truly mysterious. A baffler for the ages. --- # 🐘 Toots and Memes and Such That's right, a new section. As a millennial, I needed a place to share toots1 and memes and, well, such, that I've come across. You can, of course, also follow Technically Good on Mastodon and be served this kind of content when it's steaming fresh and piping hot. > "Growing up with the phrase “Wikipedia isn’t a valid source,” only to end up in a daily life where everyone says, “Just ask ChatGPT." by Andrea Junker > "Ai is going to lead to an abundant and prosperous future the same way that Facebook has connected me and brought me closer to my friends and family." by Brian Owen. As always: > "Move carefully and fix things." by datarama. --- # 📯 The Post-Script - I am finally reading Brian Merchant's Blood in the Machine: The Origins of the Rebellion Against Big Tech. A historical account of the Luddite movement that's weirdly relevant in 2026. - You should also check out Dr. Alex Hanna's video on Myths around AI and Labour. --- Follow this blog via RSS | Find me on Mastodon | Bubbles --- ## Recent Posts - 13 Sep, 2026 💌 Newsletter #016 - Chatbots and their "Safeguards," Alternative Navigation Apps - and a New Section! - 30 Aug, 2026 💌 Newsletter #015 - "Watermarks" in Chatbot Output; Meta settles; "It's Just a Tool?" - 16 Aug, 2026 💌 Newsletter #014 - Many, Many Leaks; Being "Opted In;" and Cursed "Podcasts" for your Children 1. yes, I still call them toots.↩

# ✨ In This Edition ✨ 1. 📟 Short-Form Content: Random word choice watermarks in chatbot output; Getting "opted in" to Google scanning your stuff; Meta settles; Books are destroyed to feed the chatbots; EU mandates driver-facing cameras. 2. 🌞 Good News!: More transparency in online shopping; Mastodon 5.0; AI-free, digitally sovereign document editing; Your old phone as a NAS. 3. 📯 The Post-Script: Reading There's no such thing as Just a Tool by Iris Meredith. Original image by Natalia Y. on Unsplash, glitched and edited. --- # 📟 Short-Form Content ## Chatbots ("AI") - OpenAI's Head of Ethics has left the company, after less than a year in the role. "There currently is no replacement lined up, suggesting shifting priorities at the company." That's probably a good sign of what's to come. Right? - Google now scans your emails and attachments. Apparently, if you use Gmail, you were "opted in" automatically. Here's some tips on how to opt out. Reminder that there are email services available that aren't google. - Anthropic is going to start "watermarking" its chatbot output. How will it do this? By "subtly altering the word choices" such that the watermarking will be "indistinguishable to readers." * This only sounds like a reasonable and normal thing to do if you assume that synonyms are exchangeable in all contexts and senses, which is, unfortunately, not how language works. Or writing. Or words. Not that the tech dudebros care about any of that if Line Go Up. - NVIDIA is buying Hugging Face. Certainly not the case that the "AI economy" is the same $5Tn or whatever many trillions of dollars going round and round. - What at first was just a hunch that booksellers had when strange bulk orders came in has now been confirmed: Yes, AI firms are buying up tons of rare books - only to train their chatbots on them. Destructively. Destroying knowledge to train a chatbot that makes people cognitively dumber. You can't make this up. * McSweeney's article on this, by Jack Loftus, is once again solid gold: > The older books are my favorite because they arrive with such rich history. Thick leather covers, gold inset lettering, the musty smell of age, and sometimes a handwritten note from 1894 that I don’t bother to read, and the scanners largely skip since it’s not integral to our performance KPIs. - A medical transcription chatbot claimed that a patient was doing drugs (of the illegal kind) - when she wasn't. A serious error with potentially grave consequences, not only in terms of the medical diagnosis and treatment patients receive. This quote by the patient, Ms Green, stuck with me - AI involvement, such as AI transcription, is presented to us as the "default" (despite it being horrendously inaccurate), and opting out feels like being difficult or "annoying": > "I don't want to be in her bad books or be that annoying patient who takes up more of her time," she said. - Microsoft's Copilot was asked about its protective mechanisms - and "eventually revealed undocumented URL parameters and protective measures," which enabled the researchers to build manipulative URL's. - The province of Alberta, Canada, is hosting multiple Town Halls around AI data centres, which could make power in Alberta even more expensive than it already is. The company, of course, disputes that. * Anyways, Albertans have shown up and they have been very clear, via jeers and boos and comments and heated questions, that they're not interested in seeing AI data centres in their province. * There is also this handyAlberta Data Centre Watch website, where users can explore planned projects. - Very unsurprisingly to anyone that's been paying attention even slightly, and has a passing understanding of how capitalism works, AI is not getting us closer to that shorter work week. Gasp! - So this is one way that your data can end up somewhere completely different from where you thought it might end up: Google buys crashed airline Spirit's data at auction, because AI. This includes "over 100 million emails, 30 million recorded phone calls, reams of stuff from Teams, Oracle, and SAP." - On Bank Robbing Robots, by user @lproven@vivaldi.net: > I invented a robot that can rob a bank. Of course, it robs these banks on its own, so I can't be held legally responsible when it robs a bank. I mean, all I did was invent it, it's not my fault if it works exactly the way I intended. Also, I get to keep all the money from the banks it robs. Sorry if any of that money was yours, but now that l've invented the Bank Robbing Robot, the Bank Robbing Robot is inevitable. ## Non-Chatbots - The big thing this week is of course the settlement in Meta's trial revolving around harms to its young users. You're probably wondering why, when Meta is going to pay up to $18bn, as well as going to implement restrictions that, for example, allow only two hours of Instagram and Facebook per day, this isnt found in the Good News section. It is partially good news, but: The amount of settlement money is a drop in the bucket given Meta's earnings - and the settlement means that months of evidence will not be heard. Meta has also denied wrongdoing. - In times of natural disasters, when people are trying to exchange valuable (and at times life-saving) information with each other, turning to a "walled garden" may not be the way to go. - Google is introducing its "extended sideloading flow," essentially adding hurdles to "sideloading" apps. "Sideloading" meaning "installing apps from anywhere that isn't Google's store." Which is kind of wild, actually, when you think of desktop computing. Really, it's just "installing software" - except there's now this added layer of some corporation getting a say in what you install on a device that you own? It's all in the name of avoiding risk and danger, of course. * Somehow we're also supposed to rejoice that they walked back their original plans and we now "only" have to deal with a one-time 24hr waiting period. Uh-huh. Talk about anchoring. Here's an earlier article on the original plans. - As part of the EU's General Safety Regulation, every new car and van sold in the EU is now mandated to have a cabin-facing camera - as in, a camera that watches the driver. Primarily, these are meant to ensure the driver is not engaging in distracted driving, for example by emitting warning sounds or buzzes if the driver glances away for too long. That is probably a good thing, seeing the epidemic of drivers focused on the phone in their crotch instead of the road, or what is going on around them. That said, any recording of this kind carries privacy risks (and, to a certain extent,"normalizes" surveillance) - there are strict rules around what this camera is allowed to do, as per Reclaim The Net: > "The system cannot use facial recognition or any biometric identification of the people inside the car. The data it generates is not allowed to leave the vehicle, and it cannot be passed to third parties." Let's hope that this will remain true. --- # 🌞 Good News! - A German court has ruled that online shopping platforms need to be more transparent in regards to their recommendation systems - such as which factors significantly determine the selection - as well as to make it easier for users to de-personalize their feeds. - Mastodon 5.0 will have a host of great new features, such as a redesigned composer, a reworked visibility/discovery settings section, and DM's being distinguished better from regular posts. If you're not on Mastodon yet, give it a try - unlike BlueSky, which claims to be decentralized but actually kind of isn't?, Mastodon definitely is. - LibreOffice is still around, doesn't use any generative AI, and is a very decent office suite. I use it for all my local document writing and spreadsheeting needs. The difference between, say, Word and Writer feels likethis little meme. * Also, reminder that "The Cloud" is "someone else's computer," and if that computer happens to be under the purveyance of someone who isn't completely rational, that could be a problem. File under: Digital sovereignty. * ... especially since also, someone can, in fact, just turn off your cloud, leaving you without your data. * On a completely unrelated note, getting off of US tech, especially of the cloud persuasion, is probably a good idea, now more than ever. Fore more ideas on that, check this excellent article by Paris Marx. - This is a neat idea, especially in light of the skyrocketing RAM and storage prices, which are of course necessary to keep Line Going Up and so that upper management can have emails summarized that someone else used a chatbot to write: Your old phone as a NAS! (Note that I have not tried this myself; as always, use caution when installing new apps/features on your devices, both in terms of security but also your saved data.) - The Fairphone 6+ is now available in Europe and the USA! Fairphones are cool because they are highly modular and repairable - the camera breaks, you can just replace the camera module. By yourself. By only purchasing the new camera module. Repairability and sustainability, heck yeah. --- # 📯 The Post-Script Reading There's no such thing as Just a Tool by Iris Meredith (@iris_meredith@mastodon.social). It has it all: Philosophy, urbanism, software development, AI. Something for everyone. You should probably read it. > It's quite possible for a tool to be used for good things, and in fact for it to be necessary, but for the viewpoint that it fosters to nonetheless be one that we shouldn't encourage. > [W]hen I hear a tech person say that something's "just a tool", I can't help but read it as an unwillingness to take responsibility. --- Follow this blog via RSS | Find me on Mastodon | Bubbles --- ## Recent Posts - 13 Sep, 2026 💌 Newsletter #016 - Chatbots and their "Safeguards," Alternative Navigation Apps - and a New Section! - 30 Aug, 2026 💌 Newsletter #015 - "Watermarks" in Chatbot Output; Meta settles; "It's Just a Tool?" - 16 Aug, 2026 💌 Newsletter #014 - Many, Many Leaks; Being "Opted In;" and Cursed "Podcasts" for your Children

# ✨ In This Edition ✨ 1. 📟 Short-Form Content: Random word choice watermarks in chatbot output; Getting "opted in" to Google scanning your stuff; Meta settles; Books are destroyed to feed the chatbots; EU mandates driver-facing cameras. 2. 🌞 Good News!: More transparency in online shopping; Mastodon 5.0; AI-free, digitally sovereign document editing; Your old phone as a NAS. 3. 📯 The Post-Script: Reading There's no such thing as Just a Tool by Iris Meredith. Original image by Natalia Y. on Unsplash, glitched and edited. --- # 📟 Short-Form Content ## Chatbots ("AI") - OpenAI's Head of Ethics has left the company, after less than a year in the role. "There currently is no replacement lined up, suggesting shifting priorities at the company." That's probably a good sign of what's to come. Right? - Google now scans your emails and attachments. Apparently, if you use Gmail, you were "opted in" automatically. Here's some tips on how to opt out. Reminder that there are email services available that aren't google. - Anthropic is going to start "watermarking" its chatbot output. How will it do this? By "subtly altering the word choices" such that the watermarking will be "indistinguishable to readers." * This only sounds like a reasonable and normal thing to do if you assume that synonyms are exchangeable in all contexts and senses, which is, unfortunately, not how language works. Or writing. Or words. Not that the tech dudebros care about any of that if Line Go Up. - NVIDIA is buying Hugging Face. Certainly not the case that the "AI economy" is the same $5Tn or whatever many trillions of dollars going round and round. - What at first was just a hunch that booksellers had when strange bulk orders came in has now been confirmed: Yes, AI firms are buying up tons of rare books - only to train their chatbots on them. Destructively. Destroying knowledge to train a chatbot that makes people cognitively dumber. You can't make this up. * McSweeney's article on this, by Jack Loftus, is once again solid gold: > The older books are my favorite because they arrive with such rich history. Thick leather covers, gold inset lettering, the musty smell of age, and sometimes a handwritten note from 1894 that I don’t bother to read, and the scanners largely skip since it’s not integral to our performance KPIs. - A medical transcription chatbot claimed that a patient was doing drugs (of the illegal kind) - when she wasn't. A serious error with potentially grave consequences, not only in terms of the medical diagnosis and treatment patients receive. This quote by the patient, Ms Green, stuck with me - AI involvement, such as AI transcription, is presented to us as the "default" (despite it being horrendously inaccurate), and opting out feels like being difficult or "annoying": > "I don't want to be in her bad books or be that annoying patient who takes up more of her time," she said. - Microsoft's Copilot was asked about its protective mechanisms - and "eventually revealed undocumented URL parameters and protective measures," which enabled the researchers to build manipulative URL's. - The province of Alberta, Canada, is hosting multiple Town Halls around AI data centres, which could make power in Alberta even more expensive than it already is. The company, of course, disputes that. * Anyways, Albertans have shown up and they have been very clear, via jeers and boos and comments and heated questions, that they're not interested in seeing AI data centres in their province. * There is also this handyAlberta Data Centre Watch website, where users can explore planned projects. - Very unsurprisingly to anyone that's been paying attention even slightly, and has a passing understanding of how capitalism works, AI is not getting us closer to that shorter work week. Gasp! - So this is one way that your data can end up somewhere completely different from where you thought it might end up: Google buys crashed airline Spirit's data at auction, because AI. This includes "over 100 million emails, 30 million recorded phone calls, reams of stuff from Teams, Oracle, and SAP." - On Bank Robbing Robots, by user @lproven@vivaldi.net: > I invented a robot that can rob a bank. Of course, it robs these banks on its own, so I can't be held legally responsible when it robs a bank. I mean, all I did was invent it, it's not my fault if it works exactly the way I intended. Also, I get to keep all the money from the banks it robs. Sorry if any of that money was yours, but now that l've invented the Bank Robbing Robot, the Bank Robbing Robot is inevitable. ## Non-Chatbots - The big thing this week is of course the settlement in Meta's trial revolving around harms to its young users. You're probably wondering why, when Meta is going to pay up to $18bn, as well as going to implement restrictions that, for example, allow only two hours of Instagram and Facebook per day, this isnt found in the Good News section. It is partially good news, but: The amount of settlement money is a drop in the bucket given Meta's earnings - and the settlement means that months of evidence will not be heard. Meta has also denied wrongdoing. - In times of natural disasters, when people are trying to exchange valuable (and at times life-saving) information with each other, turning to a "walled garden" may not be the way to go. - Google is introducing its "extended sideloading flow," essentially adding hurdles to "sideloading" apps. "Sideloading" meaning "installing apps from anywhere that isn't Google's store." Which is kind of wild, actually, when you think of desktop computing. Really, it's just "installing software" - except there's now this added layer of some corporation getting a say in what you install on a device that you own? It's all in the name of avoiding risk and danger, of course. * Somehow we're also supposed to rejoice that they walked back their original plans and we now "only" have to deal with a one-time 24hr waiting period. Uh-huh. Talk about anchoring. Here's an earlier article on the original plans. - As part of the EU's General Safety Regulation, every new car and van sold in the EU is now mandated to have a cabin-facing camera - as in, a camera that watches the driver. Primarily, these are meant to ensure the driver is not engaging in distracted driving, for example by emitting warning sounds or buzzes if the driver glances away for too long. That is probably a good thing, seeing the epidemic of drivers focused on the phone in their crotch instead of the road, or what is going on around them. That said, any recording of this kind carries privacy risks (and, to a certain extent,"normalizes" surveillance) - there are strict rules around what this camera is allowed to do, as per Reclaim The Net: > "The system cannot use facial recognition or any biometric identification of the people inside the car. The data it generates is not allowed to leave the vehicle, and it cannot be passed to third parties." Let's hope that this will remain true. --- # 🌞 Good News! - A German court has ruled that online shopping platforms need to be more transparent in regards to their recommendation systems - such as which factors significantly determine the selection - as well as to make it easier for users to de-personalize their feeds. - Mastodon 5.0 will have a host of great new features, such as a redesigned composer, a reworked visibility/discovery settings section, and DM's being distinguished better from regular posts. If you're not on Mastodon yet, give it a try - unlike BlueSky, which claims to be decentralized but actually kind of isn't?, Mastodon definitely is. - LibreOffice is still around, doesn't use any generative AI, and is a very decent office suite. I use it for all my local document writing and spreadsheeting needs. The difference between, say, Word and Writer feels likethis little meme. * Also, reminder that "The Cloud" is "someone else's computer," and if that computer happens to be under the purveyance of someone who isn't completely rational, that could be a problem. File under: Digital sovereignty. * ... especially since also, someone can, in fact, just turn off your cloud, leaving you without your data. * On a completely unrelated note, getting off of US tech, especially of the cloud persuasion, is probably a good idea, now more than ever. Fore more ideas on that, check this excellent article by Paris Marx. - This is a neat idea, especially in light of the skyrocketing RAM and storage prices, which are of course necessary to keep Line Going Up and so that upper management can have emails summarized that someone else used a chatbot to write: Your old phone as a NAS! (Note that I have not tried this myself; as always, use caution when installing new apps/features on your devices, both in terms of security but also your saved data.) - The Fairphone 6+ is now available in Europe and the USA! Fairphones are cool because they are highly modular and repairable - the camera breaks, you can just replace the camera module. By yourself. By only purchasing the new camera module. Repairability and sustainability, heck yeah. --- # 📯 The Post-Script Reading There's no such thing as Just a Tool by Iris Meredith (@iris_meredith@mastodon.social). It has it all: Philosophy, urbanism, software development, AI. Something for everyone. You should probably read it. > It's quite possible for a tool to be used for good things, and in fact for it to be necessary, but for the viewpoint that it fosters to nonetheless be one that we shouldn't encourage. > [W]hen I hear a tech person say that something's "just a tool", I can't help but read it as an unwillingness to take responsibility. --- Follow this blog via RSS | Find me on Mastodon | Bubbles --- ## Recent Posts - 13 Sep, 2026 💌 Newsletter #016 - Chatbots and their "Safeguards," Alternative Navigation Apps - and a New Section! - 30 Aug, 2026 💌 Newsletter #015 - "Watermarks" in Chatbot Output; Meta settles; "It's Just a Tool?" - 16 Aug, 2026 💌 Newsletter #014 - Many, Many Leaks; Being "Opted In;" and Cursed "Podcasts" for your Children