NotebookLM data tables, AI browser security risks & more (December 24, 2025)

Good morning,

AI isn’t just shipping features anymore, it’s hitting limits. Pricing experiments are getting pulled, regulators are stepping in, and platforms are quietly hardening their defenses. This feels less like an innovation cycle and more like a reset.

Let’s get into it.

━━━

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

🤖 Platform Moves & Power Plays

🧠 Google upgrades NotebookLM with structured data

Google is rolling out a new NotebookLM feature that turns structured tables into usable source material. It’s designed for analysts and researchers working with spreadsheets and datasets. This pushes NotebookLM closer to a real knowledge-work engine, not just a summarization tool.

🛡️ OpenAI details how Atlas fights prompt injection

OpenAI outlined new defenses to reduce prompt injection attacks inside ChatGPT Atlas. The company is explicit that the risk never fully disappears. The takeaway is clear, agentic browsers are powerful, but security remains an ongoing arms race.

🏛️ Italy orders Meta to loosen WhatsApp AI rules

Italy told Meta to suspend its policy blocking rival AI chatbots on WhatsApp Business. Regulators say the rule restricts competition and innovation. Europe is signaling that platform-controlled AI ecosystems won’t get a free pass.

🛒 AI Meets the Real World

💸 Instacart pulls back its AI pricing experiment

Instacart is rolling back an AI-powered pricing test after customers reported higher costs. The company said the tool created confusion and trust issues. This is a reminder that AI touching prices is far more sensitive than AI touching productivity.

🗣️ Amazon plugs Alexa into Angi, Expedia, Yelp

Amazon’s Alexa assistant now connects with Angi, Expedia, Square, and Yelp. Users can book services, travel, and payments directly through voice workflows. Alexa is quietly shifting from smart assistant to transaction layer.

🧰 Tools of the Day

→ Typeless, AI-native documents without rigid templates
→ Google, core AI features spreading deeper across Workspace
→ Nano Banana Playground, playful space to experiment with generative AI

⚡ Quick Hits

→ Authors sue AI companies again over training data
→ Lemon Slice raises $10.5M for digital avatar tech
→ Silicon Valley dealmaker helped secure Intel lifeline

🧠 TLDR

NotebookLM grows up, OpenAI admits AI browsers are never fully safe, Europe cracks down on platform control, and Instacart learns that AI pricing breaks trust fast. AI is maturing, and the guardrails are finally catching up.

Cheers,
David