Microsoft's AI ultimatum, ChatGPT wrapped & more (December 23, 2025)

Good morning,

Another day, another reminder that AI is moving fast, sometimes faster than the guardrails meant to contain it. Today’s stories span security gaps, year-end nostalgia, trillion-dollar infrastructure bets, and the growing tension between ambition and execution across Big Tech.

Let’s get into it 👇

━━━

Powered by Aloa • AI Research & Education Hub

━━━

Want to engage in conversation about AI news? Join the Byte-Sized Reddit Community

🔐 Platforms, Power & Pressure

🧠 AI Browsers May Always Be Hackable

OpenAI says AI browsers may always be vulnerable to prompt injection attacks, even with better safeguards. The company is effectively admitting that letting AI interpret the open web comes with permanent risk. This frames security not as a solvable bug, but as a constant tradeoff as AI agents browse and act autonomously.

🎧 ChatGPT Gets Its Spotify Wrapped Moment

ChatGPT launches a year-end review showing users their most-used prompts and habits from 2025. It’s playful, but also revealing how deeply embedded AI has become in daily workflows. Expect this kind of personalization to quietly increase user lock-in.

⚡ Alphabet’s Energy End-Run

Alphabet plans to buy Intersect Power as data center demand collides with grid constraints. Rather than wait for utilities and regulators, Google is going straight to generation. This is Big Tech acknowledging that AI scale is now an energy problem, not just a compute one.

🌍 Global AI Reality Check

🪟 Microsoft’s AI Adoption Problem

Microsoft has a problem nobody wants to buy or use, according to critics pointing to clunky products and unclear value. Despite massive investment, execution gaps are becoming harder to ignore. It’s a reminder that distribution alone doesn’t guarantee love.

🧑‍💻 GLM-4.7 Goes Open Source

GLM-4.7 goes live and open source with major gains in coding benchmarks. China’s open model ecosystem keeps accelerating, especially in developer-focused use cases. This adds pressure on Western labs that are still keeping their best models closed.

🇯🇵 Japan Goes All-In on AI

Japan unveils an AI-first basic plan to embed AI across government, industry, and education. Unlike splashy moonshots, this is about systemic adoption. National AI strategies are starting to look less theoretical and more operational.

🛠️ Tools of the Day

→ Extract by Firecrawl – Turn websites into structured data pipelines for AI workflows.
→ ClickUp – Project management with deep AI features for planning and execution.
→ Aident AI – AI agents designed to automate research, writing, and operational tasks.

⚡ Quick Hits

→ YouTube will let creators build simple AI-powered games on the platform.
→ ChatGPT adds a StubHub integration for live event discovery.
→ Australia’s Goodman plans $9.3B investment in European data centers.
→ OpenAI reports increased child safety reporting to US authorities.
→ Investors outline what it really takes to build an AI startup in 2026.
→ AI job cuts accelerate as companies credit automation for layoffs.

🧾 TL;DR

AI is colliding with reality, security flaws may be permanent, energy is the new bottleneck, open models are gaining ground, and not every Big Tech AI bet is landing. The next phase is less hype, more consequences.

Until tomorrow,
David