Policy & Power Take Center Stage

The U.S. government flexed its regulatory muscles in multiple arenas. First, the Department of Justice filed a brief in an ongoing environmental case, arguing that allowing xAI to keep operating unpermitted gas turbines is a matter of “national, economic, and energy security.” The DOJ claims the Pentagon relies on xAI’s compute cluster, highlighting the growing national security framing of AI infrastructure.

Separately, the Trump administration’s export control order against Anthropic escalated into a public battle. The directive, which forced Anthropic to suspend access to its Mythos 5 and Fable 5 models, was revealed to be “never about an AI jailbreak,” according to sources. Instead, it appears to be a broad geopolitical move. The controversy sparked a protest from “dozens of cybersecurity experts” who argue the ban dangerously hinders cyber defenders. Meanwhile, across the Atlantic, the European Union published its voluntary Code of Practice for AI content labelling, providing guidelines for companies to meet the upcoming AI Act’s transparency requirements.

SpaceX’s AI Ambitions & Market Moves

Following its historic IPO, SpaceX’s valuation ballooned to $2.7 trillion, allowing it to pass Amazon. The company immediately deployed its new stock currency, announcing the $60 billion acquisition of AI startup Cursor (TechCrunch / The Verge). The deal is a clear bid to bolster SpaceX’s struggling AI division and chase a stated “$26 trillion addressable market.”

Reliability, Security, and Protests

On the research front, startup Probably raised $9M to build “a more reliable kind of AI” focused on preventing hallucinations. This push for robustness aligns with growing emphasis on AI red teaming, explained here as a critical security practice. A stark reminder of why such testing is needed arrived via a report on a critical vulnerability in Microsoft’s Copilot. The “SearchLeak” exploit could allow hackers to steal 2FA codes, showcasing repeated failures in LLM security.

Ethical concerns around AI’s use flared at a public event, where Google CEO Sundar Pichai faced boos and a walkout during a Stanford graduation speech over the company’s defense contracts with Israel and ICE.

Products & Startups

• Meta launched a new “AI Mode” for Facebook search, which pulls answers from public posts across its platforms.
• Plaud, maker of AI notetaker hardware, announced its software business topped $100M in annual recurring revenue.
• Respond.io, a Malaysian AI agent-powered messaging platform, raised $62.5M and is eyeing acquisitions.

Editorial: Today’s stories paint a picture of an industry entering a new phase of maturity—and scrutiny. The whims of the market are increasingly tempered by the will of the state, whether through environmental permits, export controls, or content regulations. The massive SpaceX-Cursor deal shows capital is still flowing at a staggering scale for those with grand, integrated visions. Yet, beneath the macro maneuvering, the foundational challenges of reliability and security remain acute, proving that for all the talk of trillion-dollar markets, building trustworthy AI is still a million-dollar problem.