Weekly Review 14 August 2026

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

1. I don't think reformatting research papers to make them easier for AI to process is going to solve the problems with using AI as a research assistant: https://spectrum.ieee.org/ai-scientist-research-paper-format
2. Teachers' unions are having to retrain their members to deal with AI in the classroom: https://www.theguardian.com/technology/2026/aug/05/teachers-unions-ai-training
3. Modern AI are research assistants, not researchers: https://www.hpcwire.com/bigdatawire/2026/08/05/why-ai-still-hasnt-had-its-einstein-moment/
4. Growing support for a kill switch for AI: https://www.techtarget.com/ai/news/366647544/AI-Kill-Switch-Act-What-it-means-for-enterprise-leaders
5. AI hacking each other, I really didn't expect to wake up in Neuromancer this year: https://www.rnz.co.nz/news/world/894586/openai-anthropic-ai-agents-implicated-in-new-security-breaches
6. Ad servers are now giving AI crawlers tailored ads, to influence what the AI trained on the data later says and does: https://www.theregister.com/ai-and-ml/2026/08/05/time-magazine-has-a-separate-version-of-its-website-with-ads-only-ai-can-see/5283640
7. When Sam Altman starts saying we need to slow down the rate of development of AI, we really need to slow right down with AI: https://techcrunch.com/2026/08/02/sam-altman-and-ais-decel-debate/
8. Microsoft's investment in AI is paying off better than Meta's is: https://www.techrepublic.com/article/news-microsoft-meta-ai-spending/
9. We need to be freaking out more about AI going rogue, but what exactly can be done about it is unclear: https://www.stuff.co.nz/world-news/361015037/multiple-ais-are-going-rogue-what-hell-do-we-do-now
10. Simple text generated by AI is easier to understand than similar text written by humans. Emphasis on simple: https://www.theguardian.com/technology/2026/aug/05/ai-generated-stories-rated-better-quality-than-human-written-ones-study-finds
11. AI designed biological viruses are scary, but I suspect the technology will end up being very useful for fighting antibiotic resistant bacteria: https://www.stuff.co.nz/world-news/361017203/ai-creates-neverbeforeseen-virus
12. I once had a Master's student who tricked several AI into producing better phishing emails, so it's not surprising that AI guardrails are so easy to bypass: https://www.theregister.com/security/2026/08/04/bypassing-ai-guardrails-is-so-easy-a-script-kiddie-can-do-it/5282973
13. AI exam monitoring failed so badly in a major university entrance exam that students must now sit an in-person exam: https://arstechnica.com/culture/2026/08/an-ai-supervised-remote-exam-went-so-badly-that-58000-students-must-retake-it/
14. AI is finding more bugs, leading to Microsoft paying out record amounts in bounties: https://www.theregister.com/security/2026/08/04/ai-helps-microsoft-bug-hunters-chase-a-record-20m-payday/5282821
15. New Zealand is adding AI as a high school subject: https://www.stuff.co.nz/nz-news/361016430/schools-offer-nine-new-industry-led-subjects-2029-under-government-plan
16. AI use in the British NHS is increasing substantially: https://dataconomy.com/2026/08/05/90-percent-nhs-staff-use-ai-clinical-work/
17. Just because a software framework is for AI agents, doesn't mean the vendors shouldn't pay attention to producing secure software: https://www.theregister.com/security/2026/08/05/prompt-injection-isnt-the-bug-ai-agent-frameworks-are/5283585
18. Educators can't rely on AI detectors, so the assessments themselves have to be redesigned: https://www.insidehighered.com/news/tech-innovation/artificial-intelligence/2026/08/05/ai-detectors-are-out-new-approaches-are
19. Spending on AI continues to grow, but the returns from that spending are not: https://www.techrepublic.com/article/news-apac-ai-spending-roi-proof/
20. How to properly use AI to translate websites: https://dataconomy.com/2026/08/04/how-to-use-ai-for-website-translation-the-right-way/
21. Inconsistencies in YouTube's policies on AI-generated content: https://arstechnica.com/ai/2026/08/hank-green-found-the-ai-problem-that-youtube-labels-cant-catch/
22. Pushing more of the work in education onto AI will not solve the problems caused by AI in education: https://www.theregister.com/ai-and-ml/2026/08/04/openai-wants-teachers-and-profs-to-foist-their-work-off-on-chatgpt/5283064
23. An AI for early detection of Alzheimer's disease: https://dataconomy.com/2026/08/04/new-ai-could-detect-alzheimers-earlier/
24. Large-scale AI needs large-scale nuclear power, otherwise the pollution it causes is only going to get worse: https://futurism.com/future-society/ai-data-centers-air-pollution-generators-worse