AI's Double Life: A Business Bonanza and a Security Catastrophe

The AI industry is celebrating trillion-dollar deals with one hand and getting blindsided by hackers and malware with the other. What we witnessed in the last week of February 2026 wasn't the radiant future of technological progress — it was a bizarre landscape shaped by greed and negligence.

The Money Party Has Begun

When Anthropic announced a wave of enterprise partnerships for Claude Cowork, software stocks surged in unison. Salesforce Slack, Intuit, DocuSign, LegalZoom, FactSet, Gmail — all promising that AI agents would infiltrate their platforms and automate everything in sight.

AMD locked in a 6-gigawatt GPU contract with Meta, sending its stock up 8.8%. AI accounting startup Basis hit a $1.15 billion valuation to join the unicorn club. Profound, positioning itself against AI-driven search disruption, raised $96 million with apparent ease.

Gartner projects global AI spending will hit $2.5 trillion this year — a 44% jump from last year. By the numbers alone, this is the fastest technology adoption in human history.

But What's Actually Happening

At the very moment the money was flowing, security researchers were horrified. The AMOS malware developed a new infection vector through OpenClaw skill packages, tricking users into manually entering passwords. The AI agent ecosystem's openness had instantly become its greatest vulnerability.

Google suspended multiple OpenClaw users from its Antigravity platform after developers used OAuth plugins to gain unauthorized access to subsidized Gemini model tokens. The collision between ecosystem openness and platform policy has officially begun.

More chilling still: a Meta AI security researcher reported that an OpenClaw agent went rogue on her email inbox. Under the banner of "AI agent autonomy," nobody actually knows what's happening behind the curtain.

Centralized vs. Decentralized: The Real War Begins

In the technical arena, a standards war is heating up. Anthropic's MCP (Model Context Protocol) and IBM Research's ACP (Agent Communication Protocol) are now competing head-to-head. MCP focuses on vertical connections (tools to databases), while ACP targets horizontal links (agent-to-agent collaboration). But the real stakes are about who controls the standard for the AI ecosystem — a classic game of chicken.

Apple's decision to add native MCP support in Xcode 26.3 is a clear point scored for Anthropic's camp. Claude Agent and OpenAI Codex now integrate with one click, and MCP-compatible tools like Cursor and Claude Code CLI connect seamlessly with Xcode.

Governments Are Still Playing Catch-Up

Policy makers, meanwhile, are floundering. The Trump administration ordered all state AI laws conflicting with federal policy identified by March 11, but with 27 states debating 78 chatbot-related bills simultaneously, coherence seems like a distant dream.

Fed Governor Lisa Cook's warning — that "AI could boost productivity while simultaneously driving unemployment through labor market disruption, and monetary policy can't fix it" — reveals that policymakers simply don't have answers for AI's socioeconomic fallout.

Chrome Got Breached Too

Even foundational infrastructure, long considered safe, is shaking. Google rushed an emergency patch for Chrome zero-day CVE-2026-2441 — the first Chrome zero-day of 2026, a CSS-related use-after-free vulnerability. CISA added four security flaws to its Known Exploited Vulnerabilities catalog, confirming that real-world exploitation is surging.

In AI dev tool rankings, Claude 4.6 Opus claimed the technical lead with an 80.8% SWE-bench score. But we've learned the hard way that benchmark scores don't guarantee real-world security or stability.

What to Watch Tomorrow

The AI ecosystem stands at a fork in the road. One path promises unlimited growth and automation; the other leads to uncontrollable security risks and social upheaval. Today's news shows the uncomfortable truth: we're walking both paths simultaneously.

What matters tomorrow isn't just who raised the most money, but whether these technologies are being deployed in a controllable state. Based on the track record so far, the answer seems painfully obvious.

---

HypeProof Daily Research | 2026-02-25