The Good, The Bad and The Ugly in AI this week 29/09/25
The Good, The Bad and The Ugly in AI this week 29/09/25
AI Literacy
NEW! Word of the week
LLM: Large Language Model. A statistical pattern machine that does NOT "know" things. They're just next-token predictors trained on existing texts. Ask a question and they do NOT check a database. They spin a statistical wheel of fortune.
Tip of the Week
Meta has added AI to WhatsApp, the messaging app used by over 2 billion people worldwide (LinkedIn have done the same - see further down). Now there's an AI chatbot built right into your WhatsApp that can answer questions, create images, and help with tasks without you having to open another app. While this might be handy for quick searches or getting help with something, it's worth thinking about privacy since WhatsApp can see your messages and contacts. Meta is putting AI into all of its apps as part of a bigger plan to collect more data and make their services harder to live without. The AI doesn't always give perfect answers though, so you shouldn't rely on it for anything really important.
If you would like to opt-out here is how (you need to do this for every contact!):
News from Around our Schools
Pictured above is Munzarin S M, our AI Champion at The Indian Public School - Chennai North Campus. Munzarin recently delivered her AI training to colleagues and created this wonderful display afterwards. many congratulations to you and your fellow teachers at Chennai North Campus for your engagement with our AI Strategy.
Pictured above is Fani Theophanous (our AI Champion and Head of Computing Department, at Pascal Education Larnaka, Cyprus). Fani recently delivered her training on the ethical use of AI. When I asked her about the session she said: "it was interactive and included fun activities, interesting questions, and a deepfake video to watch and discuss." Many congratulations to you all for engaging with this important aspect of AI. Keep us updated with your progress!
The Hi:AI photo is from James Cole (Subject Lead Computer Science and AI Champion at Stoner, UK). The display is currently to be found in a communal area in the school and is one of many AI and computing displays that James refreshes regularly. Thank you James for all your hard work promoting the safe and creative use of AI in your school.
The Good
🤖 AI Breakthrough in Medicine
AI designs first working virus genomes
Stanford and Arc Institute researchers created the first AI-generated viruses that successfully infect and kill bacteria. The AI model called Evo designed 16 functional viruses out of 302 attempts, with 392 mutations never seen in nature. These synthetic versions broke through bacterial resistance where traditional viruses failed, marking a breakthrough in computational biology.
🏆 GPT-5 Dominates Elite Coding Competition
GPT-5 beats humans at elite coding contest
At the ICPC World Finals, OpenAI's GPT-5 achieved a perfect 12/12 score, outperforming 139 top university teams from over 100 countries. The AI solved 11 problems on first attempts and eventually cracked the toughest challenge after nine attempts, showcasing AI's growing dominance in competitive programming.
🌐 Google Embeds Gemini into Chrome
Google embeds Gemini into Chrome
Google introduced native Gemini integration across Chrome for all US desktop users. The browser now features AI-powered multi-tab context analysis, an AI Mode in the address bar for complex queries, and previewed upcoming agentic capabilities for autonomous multi-step tasks like grocery shopping and appointment booking.
🔍 Google Launches Real-Time AI Voice Search
Google rolls out real-time AI voice search
Google launched Search Live in the US, allowing users to ask questions aloud and receive real-time AI answers with supporting web links. The feature works in the Google app on iOS and Android, plus Google Lens, making search more natural and interactive.
🎭 Google's AI-Powered Mood Board App
Google launches an AI-powered mood board app
Google launched Mixboard, an app that builds AI-powered mood boards directly from text prompts using the new Nano Banana model. It offers features like image generation, regeneration, and AI-generated text support, simplifying brainstorming and creative workflows.
The Bad
🚨 China Bans NVIDIA Chips
China bans NVIDIA chips in sudden crackdown
China's internet regulator banned domestic tech giants including Alibaba and ByteDance from buying or testing NVIDIA's AI chips. This comes after NVIDIA's RTX Pro 6000D had already been ordered in large volumes, following antitrust accusations and US-China tensions over AI exports.
🤖 AI 'Workslop' Making Office Work Worse
AI 'workslop' is making office work worse
Stanford researchers found AI is flooding offices with "workslop"—content that looks professional but adds little value. 40% of employees in a survey of 1,150 companies received low-quality AI output in the past month, causing confusion and eroding workplace trust.
🛡️ Google Forced to Tackle Intimate Image Abuse
Google joins effort to fight intimate image abuse
Google partnered with StopNCII.org to remove non-consensual intimate images from Search using unique image and video hashes. The company faced criticism for lagging behind peers like Meta, TikTok, and Microsoft in adopting this approach.
📈 Disney+ and Hulu Prices Jump Again
Disney+ and Hulu prices jump again in October
Starting 21 October, Disney+ with ads rises to £11.99/month (+£2), while ad-free Disney+ hits £18.99/month (+£3). This continues the trend of steady streaming price hikes since Disney+'s £6.99 launch in 2019.
🔒 LinkedIn to Use Your Data for AI Training
LinkedIn to use your data for AI training
LinkedIn will share user data with Microsoft and affiliates to train AI models starting 3 November 2025. Users are automatically opted in and must manually opt out. Data includes profiles, resumes, posts, and interactions, raising significant privacy concerns.
The Ugly
🏛️ Meta Bankrolls PAC to Block AI Rules
Meta bankrolls PAC to block strict AI rules
Meta is funding a new PAC—the American Technology Excellence Project—with tens of millions to fight against tough state-level AI regulations. The group will back pro-tech politicians from both parties, signalling a concerning power struggle between regulators and AI firms.
💌 Meta Brings AI into Dating
Meta brings AI into the dating experience
Meta introduced AI into Facebook Dating with two new features: a dating assistant chatbot and "Meet Cute" for automated weekly surprise matches. While tackling swipe fatigue sounds helpful, the risk of AI hallucinations creating wildly inappropriate matches raises serious concerns about automated relationship recommendations.
Pending/Developing
🕶️ Meta's Neural Band Smart Glasses
Meta's new glasses respond before you move
At Meta Connect, Meta revealed three new smart glasses, including Ray-Bans that pair with a Neural Band—an AI-powered wristband that reads muscle signals before you move. Also announced: Ray-Ban Meta Gen 2 with 8-hour battery life and Oakley Meta Vanguard for athletes.
👆 Apple's First Touchscreen MacBook Pro
Apple to launch first touchscreen MacBook Pro
Apple is preparing to launch its first touchscreen MacBook Pro with an OLED display, aiming for mass production in late 2026. This marks a major departure from Apple's long-standing stance against touchscreen Macs, potentially redefining the Mac experience.
🍏 OpenAI Raids Apple for Hardware Push
OpenAI raids Apple for hardware push
OpenAI launched a major hiring offensive targeting Apple's hardware teams, offering £1M+ packages. Former Apple exec Tang Tan is leading the effort, with production agreements linking OAI with iPhone manufacturers for devices launching in "late 2026 or early 2027."
💨 xAI's Cost-Efficient Grok 4 Fast
xAI unveiled Grok 4 Fast, achieving comparable results to Grok 4 despite using 40% fewer thinking tokens, resulting in a 98% price reduction. The model rose to No. 1 in LMArena's Search Arena and supports a 2M token context.
🎵 AI Artist Secures £3M Record Deal
AI artist Xania Monet lands $3M record deal
Mississippi poet Talisha Jones secured a multimillion-dollar contract for her AI-generated R&B persona Xania Monet, following the artist debuting on Billboard charts with 10M streams. Jones uses Suno for music creation while writing all lyrics herself.
🤑 NVIDIA Invests £100B in OpenAI
Nvidia fuels OpenAI's compute chase with $100B
NVIDIA and OpenAI announced a £100B infrastructure project—the largest in AI history. The companies signed a letter of intent to deploy 10 GW worth of NVIDIA systems, representing millions of GPUs, with the first gigawatt expected online in the second half of 2026.
🏗️ Altman Details Infrastructure Push
Altman details infrastructure push in new blog
OpenAI CEO Sam Altman published plans to build infrastructure capable of producing one GW of AI capacity weekly. He argued that limited compute forces difficult choices between breakthroughs, making massive infrastructure expansion essential for tackling major challenges simultaneously.
🛡️ Google Tackles AI's Shutdown Resistance
Google to tackle AI's shutdown resistance
Google DeepMind released Frontier Safety Framework 3.0, expanding AI risk monitoring to cover emergent behaviours like shutdown resistance and unusual influence on human beliefs. The framework aims to identify critical threats warranting immediate governance and mitigation efforts.
🌊 Alibaba Floods Market with Qwen3 Models
Alibaba floods market with Qwen3 model releases
Alibaba released six new Qwen3 models across text, vision, audio, and safety, highlighted by the 1T parameter Qwen-Max. Max shows near-frontier capabilities in coding and agentic tasks, while Omni processes text, images, audio, and video across 19 languages for understanding and 10 for generation.
Thank you!
Thank you for your warm welcome in Madrid last week – it was lovely to put faces to email addresses. I know many of our conversations were just the beginning of our AI journeys together. I look forward to working with you all in the coming months.