The Hallucination HQ

Archives
Log in
Subscribe
20 June 2026

The US just banned two AI models

The Hallucination HQ

AI news that's actually fun to read — 2026-06-20

Today's Hallucination HQ

America Pulls the Plug on Two Anthropic Models, Numbers Remain Unbothered

The US government forced Anthropic to withdraw its Fable 5 and Mythos 5 models late last week, citing national security concerns after Amazon researchers allegedly found a way to bypass Fable 5's safety guardrails — which is rather the one thing guardrails are meant to prevent. Cybersecurity researchers have since signed an open letter pushing back on the ban. Usage figures, meanwhile, have declined only modestly, suggesting users treat government intervention the way most people treat terms and conditions: noted, ignored, carried on.

Source: TechCrunch


Norway Decides Children Should Learn Things the Old-Fashioned Way

Norway has introduced a near-total ban on AI tools in elementary schools, reasoning that children aged six to thirteen should probably develop basic skills before outsourcing them. It is a philosophically defensible position — you wouldn't hand a child a calculator before teaching them to count, though admittedly calculators never confidently fabricated historical facts. The policy allows limited exceptions but sets a clear national tone: Scandinavia, having already perfected work-life balance, is now having a go at protecting childhood.

Source: Reuters via Hacker News


Miami Startup Says It's Solved a Decade-Old AI Maths Problem. Peers Are Coming Round.

Subquadratic emerged from stealth last month claiming to have cracked a fundamental mathematical bottleneck that has constrained large language models for nearly ten years. The AI community greeted this with the warmth typically reserved for someone cutting a pub queue. But the Miami-based startup has since published enough supporting detail that sceptics are beginning to reconsider — which, in academic circles, is roughly equivalent to a standing ovation. The bottleneck in question relates to how efficiently models process long sequences of information, and if the claim holds, it is genuinely significant.

Source: MIT Technology Review


The Sam Altman Film Has Lost Its Studio, Which Is Art Imitating Life to an Uncomfortable Degree

Luca Guadagnino's film Artificial — starring Andrew Garfield as OpenAI CEO Sam Altman — has been dropped by Amazon MGM after roughly a year in development. The film covers the five days in 2023 when Altman was fired and then reinstated, a story so dramatic it was presumably written by someone who'd never met a corporate board before. Amazon, which has a significant investment in Anthropic — Anthropic being OpenAI's main rival — has not commented on whether that conflict of interest played any role. Pure coincidence, no doubt.

Source: The Verge


The Man Who Uses a Brain Implant to Speak Is, By Most Measures, a Power User

Casey Harrell has ALS, is paralysed, and cannot speak coherently without his brain-computer interface — a device he has now relied on for nearly three years. Researchers describe him as "the first power user" of the implant, which is both an extraordinary medical milestone and a phrase that hits rather differently when you understand what it actually means. Brain-computer interface trials are expanding globally, with multiple companies now running human studies. The technology that once seemed speculative is now, quietly and seriously, changing lives.

Source: MIT Technology Review


AI: building brain implants that restore human speech, and chatbots that confidently invent history. Quite the range.

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 Nobel Prize, now at Anthropic Older → OpenAI's very busy hiring week
Powered by Buttondown, the easiest way to start and grow your newsletter.