🤖 Copilot Exposes Confidential Emails of Customers

🔐 Copilot Exposes Confidential Emails of Customers

🌍 Words to Worlds Is AI’s Next Frontier?

📍AI Tier Tracker

🛠Trending Tools

🥪Brief Bites

💰Funding Frontlines

💼Who’s Hiring?

Microsoft 365 Copilot deploys AI-enabled features across Teams, Word, Excel, Outlook, OneNote, and PowerPoint to streamline repetitive tasks. For example, a Data Analyst might leverage Microsoft 365 Copilot to access sales datasets within Excel and forecast consumer spending scenarios based on seasonal shopping habits.

The confidentiality breach bug, labeled with the technical identifier “CW1226324,” inadvertently allowed Microsoft 365 Copilot to access classified emails of customers, exposing sensitive information and violating privacy regulations.

This isn’t the first critical security flaw. Aim Labs discovered “EchoLeak,” which allowed bad actors to send crafted emails containing hidden instructions to businesses, compelling the business’s Microsoft 365 Copilot to access sensitive information and send it to them.

73% of enterprises experienced at least one AI-related security incident within the past 12 months, with an average cost of $4.8 million per breach. These breaches take an average of 290 days to detect and contain.

We’re witnessing the rise of zero-click cyberattacks, meaning bad actors don’t need businesses to click on a malicious link or download a shady file. For example, bad actors are exploiting “LLM scope violations,” where language models with access to confidential content are compelled to bypass security measures and expose trade secrets.

In September 2024, World Labs raised a $230 million Series B funding round at a $1 billion post-money valuation to develop Large World Models (LWMs), which learn how time and space influence the physical world.

Today’s language models, like Gemini, Claude, and ChatGPT, are designed to predict the probability of a sequence of words. For example, when given: “The dog fetched the {BLANK}!” language models ask themselves, given the words so far, what’s the most likely next word?

In this case, it might predict: “{TENNIS BALL}!” LLMs are great at understanding and generating words. They can tell you what happens, but not why or how it happens.

If we toss a tennis ball into the air, we know gravity will pull it back down. This kind of physical intuition is something we develop by observing how gravity, friction, and collision influence the physical world around us. LWMs aim to understand, simulate, and predict the dynamics of the physical world just like we do.

We rely on our physical intuition every day to predict how the physical world might react to our actions. For example, knowing how much force to use when opening a car door. It’s not some magic ability. It’s our brain constantly building an internal model of how the physical world works based on our experiences.

So, what if LLMs could also autonomously experience the forces and flows of the universe? LWMs aim to help Gemini, Claude, and ChatGPT move beyond “text,” “images,” “audio,” “video,” and “code” by learning to uncover the underlying laws that govern the physical universe.

Kana raised a $15M Seed Round for an AI-powered marketing platform.

SurrealDB closed a $23M Series A for multi-model, AI-native databases.

Temporal landed a $300M Series D to make sure your code never fails.

NVIDIA (Santa Clara, CA): GPU Architect Intern, Summer 2026

Airtable (New York, NY): Software Engineer, AI Apps, Entry-Level

OpenAI (San Francisco, CA): Digital Marketing, Growth, Mid-Level

Meta (Menlo Park, CA): Product Design, RealTime AI, Senior-Level