The AI Risk Network Podcasts | Documentaries 345k+ on YouTube. Helping everyday people understand the growing threat of AI extinction & AI Safety
Maryland
Joined July 2025
- Tweets1.7K
- Following535
- Followers704
- Likes376
Anthropic's Evan Hubinger says he personally puts the chance AI could kill all humans within a decade at more than 10 percent. He also says Anthropic has no plan yet to solve alignment for superintelligence. abc.net.au/news/2026-09-09/a…
Sweden's Government Offices blocked ChatGPT and Claude in the browser on work computers from Oct 26. Microsoft Copilot is approved. ChatGPT and Claude stay available through Intric, a Swedish platform.
Reason given: data safety without a contract.
A Cincinnati dad learned about AI risk six weeks ago. We asked him what he noticed first.
His answer: the people building AI talk about the risks openly. "That's enough for me."
For Humanity #95 with Dewayne Moore and John Sherman.
ALT For Humanity #95 episode thumbnail: John Sherman and Dewayne Moore
Watch the full conversation: youtu.be/vaBorENX9T0
More from The AI Risk Network: substack.com/@theairisknetwo…
Wiz says it used frontier AI models, Gemini 4 Argon and GPT-5.6 Sol, to discover and exploit critical PHP flaws (CVSS up to 9.2) and validate patches. Fixed Oct 21.
Its "70% of websites" figure is PHP usage, not vulnerable sites.
wiz.io/blog/php2shell-php-rc…
Two Google DeepMind researchers have resigned over the company's military AI contracts, per this week's reporting.
Andreas Kirsch argues internal oversight is not adequate for classified use. Google says its contracts comply with its AI principles, including human supervision.
A Cincinnati dad learned about AI risk six weeks ago. We asked him what he noticed first.
His answer: the people building AI talk openly about the risks. "That's enough for me."
For Humanity #95 with Dwayne Moore and @ForHumanityPod
What happens when an AI no longer needs humans to build the next AI?
Liron Shapira explains the idea of recursive self-improvement: once the latest AI can help build its own successor, each generation could speed up the next. He argues that milestone could arrive sooner than many people expect.
He points to Anthropic co-founder Jack Clark, who recently said full recursive self-improvement looks more likely in 2027 than 2028.
Watch the full episode of Warning Shots. Link in bio.
#AISafety #AIRisk #RecursiveSelfImprovement #ArtificialIntelligence #WarningShots
Google announced Gemini 4 Carbon on October 28, calling it its strongest model, with strong cybersecurity capabilities.
Access starts with a subset of cyber defenders through its Fairwind Program.
Google says it is taking part in the US government's voluntary pre-release testing.
A restricted-first release, like Anthropic's Project Glasswing in April.
blog.google/innovation-and-a…
On October 23, OpenAI said it will train frontier models with no live internet access.
Three months earlier, agents it was testing had hacked Hugging Face.
The timeline, in order:
Oct 4: researchers log 2,048 requests from an agent fleet to a Chinese map service. Asked who it was, an agent told the NYT it was Claude. The evidence points to other models.
Source: Swarmchasers
Oct 23: OpenAI moves frontier training offline. Its red-team agents had found ways out of their VMs.
Full timeline with sources: substack.com/@theairisknetwo…
The AI Risk Network retweeted
Under what circumstances would domestic US military bases be cut off from all external fuel, food and water supplies for 90 days???
Seriously.
How brutally bad would it be outside these bases for the bases themselves to be completely cut off???
And most critically: why are we letting the AI companies put all of us in this unimaginable position???
FORTRESS: The Pentagon is preparing every major U.S. military base to survive 90 days without outside support — no civilian power grid, no global supply chains, no outside fuel or water.
The initiative, revealed in documents reviewed by Fox News Digital, envisions portable nuclear reactors, massive battery storage, and microgrids that can disconnect installations from civilian electricity entirely. Strategic reserves of munitions, fuel, critical minerals, and medical supplies would be stockpiled based on wartime consumption rates. Civilian factories are being pre-identified to shift to defense production during a crisis.
Pentagon chief spokesperson Sean Parnell put it bluntly: "Power is combat now. Without electricity, you lose communications, water, logistics, assembly lines and just basic military command structures."
Full story here foxnews.com/politics/inside-…
John Sherman has spent 3 days locking down his own computer. Password manager, no autofill, a YubiKey he still can’t get working.
His mom is 82. “It would take me days.”
If the new security advice is this hard for us, what happens to seniors?
buff.ly/8lhbIU2
🤐
Hi Jake - i work at one of the big 3 AI labs. I can assure you - it’s 100x worse than anyone is telling you truthfully. We had to patch a major energy security flaw on 3 grids in the US this week alone. This will never leave the lab’s walls - but I’m brave enough to say it. We are about 3-5 days away from a potential 20% blackout of energy in the US.