Daily Tech News 14 June 2026

Top Story

• The US government has responded to Anthropic's constant warnings of imaginary dangers with a global ban - except for US citizens on US soil - on access to the company's Mythos 5 and Fable 5 security-focused AI models. (Tech Crunch)
  
  This whole story is dumb but at the same time it's amusing to watch Anthropic's continuing push towards regulatory capture leave them stuck like a rat in a trap.
  
  To mix my metaphors, Anthropic CEO Dario Amodei cried "risk" one too many times and ended up being devoured by it.
  
  
  
• Speaking of dumb 75 datacenter projects totaling $130 billion in value were blocked in just the first four months of 2026. (Tom's Hardware)
  
  This leaves SpaceX, by the way, pulling in $2 billion a month renting out its datacenters to its capacity-limited competition.
  
  OpenAI CEO Sam Altman blames this trend on Chinese influence, and while Altman absolutely is a compulsive liar, that probably is a large factor. Chinese datacenter rollouts are not facing the same headwinds.
  
  
  
• Speaking of which, OpenAI is facing investigation from multiple states attorneyses generalseses. (Tech Crunch)
  
  Oh no.

Tech News

• How are the inmates at The Verge coping with our new African-American trillionaire? Good lord. (The Verge)
  
  Maintain eye contact and back away slowly. Do not engage. I repeat, do not engage.
  
  
  
• Nvidia's high-end desktop graphics card, the RTX Pro 6000 - has had a small price bump. (Tom's Hardware)
  
  The card is basically an RTX 5090 with triple the memory. It now costs $13,250.
  
  
  
• Intel meanwhile is planning to compete with AMD's relaunched Ryzen 5800X3D processor from 2022 - with a relaunch of its own Raptor Lake chips, also from 2022. (Tom's Hardware)
  
  Raptor Lake was the codename for most of Intel's 13th and 14th generation chips; power-hungry but capable, and most importantly and just like the 5800X3D, still supporting DDR4 memory. (Though Raptor Lake can also work with DDR5.)
  
  This market opportunity opened up with the AI companies consuming all the world's DRAM, so CPU makers are looking to sell upgrades to consumers wanting to keep their DRAM and upgrade their system around it.
  
  (I own 384GB worth of DDR4 modules, bought at a small fraction of today's prices. You can bet I am interested.)
  
  
  
• A British police officer has been suspended from normal duties after being caught using AI tools to fabricate evidence. (BBC)
  
  That's not the story. This is the story:

  "The officer involved has been removed from frontline duties, pending the outcome of the investigation. No arrests have been made."
  
  

   It comes in the same week that a new national centre for AI in policing, called PoliceAI, was established.

  At the launch on Wednesday, PoliceAI interim director Alex Murray said: "Crime and technology are evolving rapidly.

  "Policing must keep pace by adopting AI responsibly to catch criminals and keep people safe."

  We're sorry we got caught. We'll make sure that never happens again.
  
  
  
• Over 400 Arch Linux packages have been compromised to push a rootkit infostealer. (Bleeping Computer)
  
  Seemsbad.
  
  
  
• Arch Linux now believes the incident is under control - after detecting 1579 compromised packages. (Phoronix)
  
  It's not as bad as it seems: The packages are not in the core of the distribution, but in a secondary user-contributed repository that has long been viewed as "use at your own risk".
  
  
  
• London's Docklands Light Railway is still running on Windows XP. (The Register)
  
  Good.
  
  
  
• KPMG's report Total Experience: Redefining Excellence in the Age of Agentic AI is bullshit on more levels than the immediately obvious. (The Register)
  
  The report on the use of AI was AI-generated. Only 5 of 45 citations actually matched the material being cited.
  
  
  
• Speaking of which, I've been using Grok recently to dig into the background of a science fiction story I am working on involving highly speculative physics.
  
  It is extremely helpful at sniffing out related real-world speculative papers on the topics I am interested in, since the papers themselves are most often locked behind paywalls.
  
  But it is also very clear when it reaches its limits. LLMs cannot "clump" without careful management. Grok has "projects" limiting conceptual bleed between answers, but it doesn't have nested subprojects; I can't easily keep it from inserting irrelevant details into a question, unless I create a new project with every question and then merge the answer into the whole afterwards.
  
  (If you try it for a while you'll notice this happen as soon as you venture beyond simple question-and-answer use.)
  
  
  
• Apple's powerful new M5 Max chip is so far only available in laptops. There's just one small problem. (WCCFTech)
  
  It produces enough heat to melt the screen.
  
  
  
• The OneXPlayer X2 Mini Pro is a supercomputer in a handheld console: A Ryzen AI Max 388 APU with up to 64GB of RAM and 8TB of storage. (Notebook Check)
  
  And a price to match, unfortunately, with the cheapest model running to $2399.
  
  
  
• And finally, Shutterslop. (Nerds.xyz)
  
  The stock photography company has embraced the inevitability of AI as its market implodes.