AI snoops your email 👀

Meta Expands Nvidia Deal to Use Millions of AI Chips

Meta has struck a massive multiyear deal with Nvidia to deploy millions of chips across its AI data centers—GPUs, standalone CPUs, networking gear, and next-generation systems. Financial terms weren't disclosed, but analysts estimate the deal is worth tens of billions of dollars.

What Meta plans to use them for: Training and inference across its entire AI infrastructure—powering personalization and recommendation systems for billions of users, plus its Meta Superintelligence Labs efforts. CEO Mark Zuckerberg said the partnership supports the company's push to deliver personal superintelligence globally. 

This isn't new, but it's much bigger: Meta has been buying Nvidia GPUs for over a decade. But this deal is a significant expansion, locking in supply across multiple chip generations at a time when Blackwell GPUs are back-ordered and Rubin production is just ramping up. Meta accounts for roughly 9% of Nvidia's revenue, and this commitment effectively reserves a dedicated slice of Nvidia's pipeline.

Microsoft Copilot Bug Exposed Confidential Emails

Microsoft confirmed that a bug in its Copilot AI allowed the assistant to read and summarize customers' confidential emails for weeks, bypassing the data loss prevention policies companies had specifically set up to prevent this.

How it happened: 

• Copilot Chat's “work tab” feature picked up emails in users' Sent and Drafts folders, including ones marked with confidential sensitivity labels—the kind organizations use to protect proprietary information, legal communications, financial data, and trade secrets. 

• Microsoft traced the root cause to a code error that allowed items in those folders to be read by Copilot, even when policies explicitly blocked automated tools from accessing them.

• Customers first reported the issue in January. Microsoft acknowledged the bug in February and began rolling out a fix.

What Microsoft hasn't said: How many customers were affected, how many confidential emails were exposed, or whether the AI stored or retained any of the content it accessed during the vulnerability window.

Google Adds Music Generation to the Gemini App

Google is rolling out a music-generation feature in the Gemini app, powered by DeepMind's latest Lyria 3 model. It's in beta and available to all users 18+ globally.

How it works: Describe any song you want—something as crazy as “a comical R&B slow jam about a sock finding its match”—and Gemini creates a 30-second track with lyrics and AI-generated cover art. You can also upload a photo or video and ask Gemini to create a song that matches the mood. Users can customize genre, tempo, instruments, and vocals, or reference an artist's name for broad stylistic inspiration (though Gemini won't mimic their voice directly).

Prompt to production code

Shipper.now is a no-code AI app builder that turns plain-language prompts into fully deployed web apps—handling frontend, backend, database, auth, hosting, and deployment automatically.

How you can use it:

• Launch MVPs without hiring engineers

• Build internal tools in minutes

• Iterate on features with AI-generated updates

• Cut dev costs while reducing bugs and rework

Pricing: Paid

Your AI Marketing Agency

Protaigé is an AI-powered marketing platform that acts like a virtual agency. Its multi-agent AI system takes a single brief and produces a complete, ready-to-launch campaign.

How you can use it:

• Scale marketing output without adding headcount

• Automate compliance checks and brand-safe approvals

• Produce coordinated assets across social, email, and ads

Pricing: Free and paid plans available