Weekly Review 22 May 2026

Some interesting links that I Tweeted about in the last week (I also post these on Mastodon, Threads, Newsmast, and Bluesky):

1. The hardware demand from the AI boom is causing industrial relations problems for Samsung: https://www.rnz.co.nz/news/world/595429/at-samsung-the-global-ai-boom-spurred-a-looming-strike-and-deep-divisions
2. Even at a so-called "elite" institution like Princeton, 30% of students are using AI to cheat: https://arstechnica.com/tech-policy/2026/05/ai-driven-cheating-widespread-even-at-elite-schools-like-princeton/
3. The New Zealand Defence Force really should have worked out how it was going to use AI before it rolled AI out across the entire organisation: https://www.rnz.co.nz/news/science-and-technology/594967/nzdf-still-drafting-ai-directive-months-after-rolling-out-tech
4. The explosion in AI use has had a commensurate explosion in the use of AI to attack systems: https://www.theguardian.com/technology/2026/may/11/ai-powered-hacking-industrial-scale-threat-three-months-google
5. Writing students seem to see AI as a tool to help them write, rather than a cheating tool: https://www.theguardian.com/us-news/ng-interactive/2026/may/10/fiction-writing-professor-ai
6. The New Zealand government's approach to AI is too trusting of the AI companies: https://www.rnz.co.nz/news/political/594827/polyanna-policy-is-nz-s-framework-for-ai-use-in-government-overly-optimistic
7. It's not just huge amounts of energy that gets consumed by AI, it's fresh water too: https://arstechnica.com/tech-policy/2026/05/data-center-used-30-million-gallons-of-water-without-initially-paying/
8. Using AI to create "new works" by a long-dead classic author is misleading at best-it's not the author's work, it's a guess by a machine with no soul: https://www.theguardian.com/stage/2026/may/11/moliere-ex-machina-ai-create-new-work-france-equivalent-shakespeare
9. The lack of regulation around AI kids' toys is concerning: https://arstechnica.com/ai/2026/05/the-new-wild-west-of-ai-kids-toys/
10. I feel that an AI-enabled mouse pointer might be taking things a bit too far: https://www.theregister.com/software/2026/05/13/googles-ai-enabled-mouse-pointer-understands-this-and-that/5240005
11. I tell my first-year AI students that bad data gives you a bad model. I didn't think it'd go this far though: https://arstechnica.com/ai/2026/05/anthropic-blames-dystopian-sci-fi-for-training-ai-models-to-act-evil/
12. While AI enables more attacks, it also finds more vulnerabilities for vendors to patch: https://www.theregister.com/patches/2026/05/14/welcome-to-the-vulnpocalypse-as-vendors-use-ai-to-find-bugs-and-patches-multiply-like-rabbits/5240027
13. AI can design gardens, but I don't think they're going to be very good: https://www.theguardian.com/lifeandstyle/2026/may/13/chelsea-flower-show-garden-designers-clash-over-ai
14. Distributing AI data centres to individual houses seems like a good idea. It might be even more efficient if they used the waste heat to heat the hot water: https://arstechnica.com/ai/2026/05/the-newest-ai-boom-pitch-host-a-mini-data-center-at-your-home/
15. I don't think we can trust AI enough just yet to let them run even a micronation: https://www.rnz.co.nz/news/world/595412/he-declared-a-new-country-governed-by-ai-he-s-not-sure-it-will-end-well
16. Almost 75% of AI customer server bot roll-outs fail, mostly because they don't meet expectations: https://www.theregister.com/ai-ml/2026/05/13/ai-customer-service-bots-get-rolled-back-at-74-of-firms/5239800
17. An AI Wikipedia clone that openly hallucinates everything: https://futurism.com/artificial-intelligence/deranged-wikipedia-clone-made-entirely-of-ai-hallucinations Note, not intended to actually be an encyclopedia.
18. AI is now involved in the majority of security breaches: https://www.extremetech.com/internet/ai-is-involved-in-most-modern-security-breaches-report
19. Meta has fired so many of its workers because of AI that morale has collapsed: https://futurism.com/artificial-intelligence/mark-zuckerberg-treating-workers-poorly
20. A lot of AI training is being done by people that have been made homeless by AI: https://futurism.com/artificial-intelligence/ai-industry-homeless
21. Coding is like every other skill, if you don't keep using it you lose it. So using AI to write code for you degrades your coding skill: https://futurism.com/artificial-intelligence/engineers-losing-coding-ability-ai
22. An MCP server that allows AI to fact-check: https://www.theregister.com/ai-ml/2026/05/13/see-through-local-ai-lies-with-irish-eyes/5239911
23. If you make AI use one of your employee's KPI, then they will maximise their usage of AI whether it makes sense or not: https://arstechnica.com/ai/2026/05/amazon-employees-are-tokenmaxxing-due-to-pressure-to-use-ai-tools/
24. The newest bottleneck in using AI is Wi-Fi speed: https://spectrum.ieee.org/wi-fi-enterprise-networks