The Big Deals & Blocked Deals

In a significant regulatory move, China has ordered Meta to unwind its $2 billion acquisition of AI agent startup Manus, dealing a major blow to Mark Zuckerberg’s strategic push into the space. The months-long probe concluded with a veto, signaling heightened scrutiny over foreign tech giants’ AI ambitions. Read more on TechCrunch.

Meanwhile, OpenAI and Microsoft announced an amended partnership agreement, aiming to simplify their complex relationship and provide “long-term clarity” as they continue scaling AI innovation. Read more from OpenAI. In other corporate news, Google DeepMind is partnering with the Republic of Korea to accelerate scientific breakthroughs using frontier AI models. Read the announcement.

The AI Agent Frontier

The concept of AI agents—autonomous programs that can perform tasks—is moving from theory to tangible products, but not without issues.

• The AI Phone? Rumors suggest OpenAI could be developing a phone where AI agents fundamentally replace traditional apps, with mass production speculated for 2028. Read more on TechCrunch.
• Security Warning: Google researchers are warning that malicious web pages are “poisoning” enterprise AI agents via indirect prompt injections. Hidden instructions embedded in HTML are creating digital booby traps in public data scraped by these agents. Read more at AI News.
• Learn to Build Them: For developers, Google and Kaggle have launched a new “Vibe Coding” course focused on building generative AI applications and agents. Check out the course.

Ethics, Power & Odd Listings

As AI integrates deeper, its societal and operational impacts are coming into sharper focus.

• Canva’s AI Glitch: Canva has apologized after its new “Magic Layers” AI feature was found automatically replacing the word “Palestine” in user designs, raising serious questions about hidden bias in generative tools. Read the report at The Verge.
• Powering the Future: On the infrastructure side, MIT researchers unveiled “EnergAIzer,” a faster method to estimate AI power consumption in seconds, helping data centers reduce wasted energy. Read the MIT News article. Looking even further ahead, Meta has inked a deal for space-based solar power beamed to Earth at night. Details on TechCrunch.
• AI in Design & Real Estate: Automakers like GM and Nissan are using AI to accelerate car design, moving from traditional sketches to AI-generated 3D concepts. See the concepts at The Verge. In a sign of the tech wealth times, a Bay Area homeowner is listing a 13-acre property with an unusual requirement: payment in Anthropic equity. The bizarre listing on TechCrunch.

Editorial: The Messy Birth of Agentive AI

Today’s news paints a picture of an industry in transition, barreling toward an “agentive” future where AI doesn’t just answer questions but acts. The ambition is clear: OpenAI’s rumored phone, Meta’s foiled acquisition, and new developer courses all point in one direction. Yet, the road is already littered with real-world friction—geopolitical blocks, security vulnerabilities from poisoned web data, and ethical stumbles like Canva’s biased filter. This isn’t just a tech shift; it’s a messy integration into the complex systems of global politics, human discourse, and digital infrastructure. The principles announced by leaders (see OpenAI’s guiding principles) will be tested not in labs, but in these exact kinds of chaotic, real-world interactions.