Skip to content
noisebrief.

Today's tech noise. Briefly.

One email. Every morning.

Past brief

Thursday, July 30

Big Tech's AI Power Grab

Microsoft is no longer pretending (opens in a new tab) its AI partnerships are purely symbiotic. The company logged $3.2 billion from its Anthropic investment while its OpenAI relationship delivered murkier returns — and simultaneously confirmed a Copilot 'super app' is shipping this year, positioning it as a direct rival to the very models it helped fund. Xbox revenue fell 10 percent in the same quarter, making clear where Microsoft is betting its future. Meanwhile, Mark Zuckerberg used Meta's Q2 earnings to declare that billions of people will have personal AI agents (opens in a new tab) within five years, framing Meta's enterprise push as extending well beyond agents into a broader platform play. The race to own the AI interface layer — the daily-use product that sits between users and models — is now openly contested between companies that were, until recently, presenting themselves as collaborators.

The question of who controls AI model development is sharpening into a genuine policy fault line. A Science investigation finds that top AI startups — the labs now setting the technical agenda — are barely publishing their research (opens in a new tab), a stark reversal from the open academic culture that produced the field's foundational breakthroughs. That closure sits uneasily alongside the ongoing Silicon Valley debate over open-weights models (opens in a new tab), where releasing model weights publicly offers transparency but also removes guardrails. Into this gap steps a concrete safety concern: Hugging Face (opens in a new tab), the open-model repository used by hundreds of thousands of researchers, suffered a significant security breach, exposing just how much critical AI infrastructure sits on platforms with consumer-grade security assumptions.

Sources · 45