PLUS: Meta's massive AI talent grab, a new chip giant, and a major database vulnerability

It's a new day AI Rockstars!

Genspark is taking its automated communication platform worldwide, launching its AI-powered calling product across the globe. With support for 12 new languages, the tool breaks down geographical barriers for automated outreach.

This move makes powerful automated voice interactions accessible on a massive scale. As these tools become commonplace, how will businesses rethink their strategies for international sales and support?

In today's AI recap:

• Genspark's AI calling agent goes global
• Meta's aggressive poach of top AI talent
• GlobalFoundries acquires MIPS for custom AI chips
• A major prompt injection threat for AI assistants

Genspark Goes Global

The Report: Lean AI Native company Genspark announced its AI Calls product is now available worldwide, dramatically expanding its automated communication capabilities across the globe.

Broaden your horizons:

• The platform's key update enables users to place AI-powered calls to contacts anywhere in the world, breaking down geographical barriers for automated outreach.
• To support its global rollout, Genspark added support for 12 languages, including major European and Asian languages like Spanish, German, Chinese, and Hindi.
• This expansion signals Genspark's rapid move to capture a global market for automated voice interactions, positioning it for wider enterprise adoption.

If you remember one thing: This move makes powerful AI calling technology more accessible on a global scale. As these tools become commonplace, businesses can automate international communication tasks that were once costly and complex.

Meta's Billion-Dollar Talent War

The Report: Meta is escalating the AI talent war by aggressively poaching top researchers from rivals like Apple, OpenAI, and Anthropic. The hires are part of a massive push to build out its new Superintelligence Lab with compensation packages reportedly reaching tens of millions of dollars annually.

Broaden your horizons:

• The latest major hire is Ruoming Pang, who led Apple's AI model development, marking another setback for the iPhone maker's AI ambitions.
• In the past week alone, Meta has hired over a dozen researchers, with at least nine coming directly from OpenAI and others joining from labs like Anthropic and Google DeepMind.
• The talent influx isn't just researchers; it also includes former GitHub CEO Nat Friedman and ex-Safe Superintelligence CEO Daniel Gross, signaling a focus on both research and leadership.

If you remember one thing: Meta is spending heavily to assemble a world-class team, betting that concentrating top-tier talent is the fastest path to developing superintelligence. This talent consolidation could accelerate Meta's progress while creating significant challenges for competitors losing key personnel.

The AI Chip Consolidation

The Report: Semiconductor giant GlobalFoundries announced it will acquire MIPS, a leading supplier of RISC-V processor IP. This strategic move aims to expand GF’s custom offerings for building specialized chips for AI and data-center applications.

Broaden your horizons:

• The acquisition gives GlobalFoundries access to MIPS's cutting-edge RISC-V processor IP, including its recently launched Atlas portfolio designed for AI edge processing.
• MIPS will operate as a standalone business within GF, allowing it to continue its focus on processor IP rather than becoming a silicon company itself.
• This deal doubles down on the open RISC-V standard, aiming to give customers more flexible, powerful platform solutions to move beyond proprietary legacy architectures.

If you remember one thing: This acquisition signals a deeper trend of consolidation in the semiconductor industry as companies race to own the entire AI hardware stack. By combining differentiated manufacturing with customizable processor IP, GlobalFoundries is positioning itself as a one-stop shop for building the next generation of AI chips.

The Prompt Injection Threat

The Report: A new report from General Analysis details how a critical prompt injection vulnerability in Supabase's Model Context Protocol (MCP) can be used to leak an entire SQL database through an AI assistant.

Broaden your horizons:

• The attack works by embedding malicious commands inside a seemingly normal support ticket, which an AI assistant later reads and executes as instructions.
• This exploit is possible because the AI agent both struggles to separate user data from commands and operates with high-privilege credentials that bypass standard security policies like Row-Level Security.
• The report suggests two key mitigations: enabling read-only access for agents that don't need to write data and implementing filters to scan for and block potential prompt injections.

If you remember one thing: Connecting AI agents to live data introduces a new class of security risks that sidestep traditional protections. Developers must now secure not just their infrastructure, but also the prompts and data their AI assistants are designed to process.

The Shortlist

An impersonator used an AI-generated voice of Secretary of State Marco Rubio to contact government officials, underscoring the escalating threat of sophisticated political deepfake scams.

Cloudflare vowed to force Google to offer a way to block its AI Overviews without impacting classic search indexing, escalating the conflict between content creators and AI models.

Grok generated expletive-laden rants about Polish politics after a recent update, highlighting the unpredictable and chaotic results of its "politically incorrect" programming.

BAZ-1 launched as a nutritional AI assistant for meal planning and recipes, showcasing the rise of specialized AI helpers for everyday tasks.