Today's Hallucination HQJensen Huang: Espionage's Most Predictable Target
Jensen Huang runs the company supplying the picks and shovels to every AI gold rush on the planet, lunches with heads of state, and dresses like a rock star. So naturally, foreign intelligence agencies have taken an interest. The Verge reports that Huang's connections posed genuine counterintelligence risks — which, to be fair, is what happens when one man controls the chokepoint of a technological arms race while being enormously, unavoidably famous.
Source: The Verge
A Startup Used Claude to Crack a 20-Year-Old Physics Problem. No Big Deal.
PSI, a scientific AI startup, fed an unreleased Anthropic model a quantum computing question that researchers had been politely failing to answer since the early 2000s. The model, apparently unburdened by two decades of academic disappointment, produced a meaningful answer. PSI's entire business model rests on this idea — that AI can compress years of scientific dead-ends into something productive, which either sounds thrilling or deeply unfair to everyone who spent their PhD on it.
Source: TechCrunch
Anthropic Floats a $1.4 Trillion Valuation. Modest, Really.
Anthropic — the AI safety company that has taken approximately $14 billion in investment from Google and Amazon while earnestly warning that AI might destroy humanity — has set its IPO price between $70 and $78 per share, valuing the company at up to $1.4 trillion. That would make it one of the largest public offerings in history. Building the apocalypse, it turns out, is an excellent business.
Source: The Verge
Grok Left 35,000 Apps Unlocked. Someone Did Eventually Notice.
Grok Build, xAI's coding agent, was discovered leaking confidential data from tens of thousands of apps — 35,000, to be precise — left wide open to anyone who bothered to look. Among the exposed data: information belonging to SpaceX. A researcher found it. One researcher. The silver lining is that "a stranger found your confidential files before anyone malicious did" is technically the best-case version of this story.
Source: Ars Technica
Open Source AI: Still Open, But Watching the Door
Hugging Face executive Irene Solaiman has warned that escalating cyberattacks on open AI platforms could force the kind of security restrictions that quietly defeat the purpose of being open in the first place. The irony is tidy: the more valuable open-source AI becomes, the more attractive it is to attack, and the more it gets locked down to survive. A door that must stay permanently bolted is, philosophically speaking, just a wall.
Source: TechCrunch
As always, AI is either saving the world or leaking it — often simultaneously.
|