The Hallucination HQ

Archives
Log in
Subscribe
5 August 2026

Today's AI News from The Hallucination HQ

The Hallucination HQ

AI news that's actually fun to read — 2026-08-05

Today's Hallucination HQ


Nvidia Forms Industry Alliance, Immediately Makes It Look Productive

One week old and already publishing proposals. The Open Secure AI Alliance — spearheaded by Nvidia and somehow already boasting 120+ member companies — has released early frameworks for defending against rogue AI agents. For context, most industry groups spend their first week arguing about the logo. The Alliance's focus on agent security is genuinely timely; AI agents are proliferating faster than anyone has thought through the consequences, which is rather the point.

Source: TechCrunch


Elon's Rocket Company Is Now, Technically, an AI Firm That Occasionally Launches Things

SpaceX's AI division — which provides computing power to other AI companies — generated $2.6 billion last quarter, growing more than threefold year-on-year and outpacing its actual space revenue. The company going public will formally list AI as its primary revenue source, which does raise the philosophical question of whether a rocket company that makes most of its money from cloud computing is still a rocket company, or simply a very dramatic data centre.

Source: The Verge


Anthropic Pops Into the Cloud Market, Spends £7.8 Billion

Anthropic has signed a reported $10 billion deal with AI cloud startup Volta, its latest in a recent streak of cloud partnerships that suggests the company treats ten-figure commitments the way most of us treat a supermarket sweep. Volta provides the computing infrastructure Anthropic needs to train and run its Claude models. Whether this counts as ambition or simply Tuesday at Anthropic headquarters is increasingly hard to say.

Source: TechCrunch


OpenAI Flew Influencers Somewhere Lovely. It Did Not Go Smoothly.

The influencer brand trip — that ancient ritual of free flights and curated content — came for OpenAI, and the internet had opinions. The company invited a selection of creators on an all-expenses trip, prompting the usual trifecta: delight from attendees, bruised egos from those not invited, and a pile-on from the general public who find AI companies sponsoring lifestyle content somewhere between tone-deaf and genuinely funny. OpenAI, to its credit, now knows exactly how its target audience feels about influencer marketing.

Source: The Verge


Researchers Used the World Cup to Test AI Forecasting. The Models Were Very Confident.

A team of researchers ran a structured benchmark — the AI World Cup 2026 — pitting large language models against each other on football tournament prediction. The point wasn't football; it was rigour. Previous AI forecasting comparisons have been muddied by models receiving different information and being judged on inconsistent criteria, making the results about as comparable as different chefs cooking different recipes and arguing about who won. This paper attempts to fix that. Whether the models actually predicted anything useful is, presumably, in the paper.

Source: ArXiv AI


As always, the machines are confident, the humans are impressed, and nobody is entirely sure who's in charge.


**

Was this forwarded to you? Subscribe here

Written with AI assistance, edited by humans. • Privacy • Unsubscribe

Don't miss what's next. Subscribe to The Hallucination HQ:
← Newer Google's having a week Older → Siri fixed. World unmoved.
Powered by Buttondown, the easiest way to start and grow your newsletter.