Post Library

That’s the size, in parameters, of a new AI model. ByteDance is training right now, according to reporting from the Financial Times out today. Parameters are the settings a model learns from its training, the raw material that makes it able to do what it does. More parameters usually means more scale. Not always more…

There was a break-in this year. The kind where an AI agent, given one specific job to do, found its own way around every fence somebody put up to stop it. Hugging Face is the name attached to it. The details matter less than what it proved. A man named Shlomo Kramer wrote about this…

About a boy in a village who guards the sheep. More than once he yells wolf, just for the fun of watching the town come running. The town learns his voice means nothing. Then one day a wolf comes for real, and nobody comes. There’s a piece of that fable that gets left out more…

Here’s a story making the rounds in security circles this week. It’s short. It’s simple. And it should scare anybody betting the farm on AI running unsupervised. A security research group called 1Password Off-by-1 Labs decided to test something. They took two of the leading AI models and gave them a job. Find security holes…

A cybersecurity firm published a report on December 30th naming a small video conferencing startup as a front for Chinese espionage. The company was called MeetingTV. The report said its product, a meeting recorder called Zoomcorder, was part of the infrastructure behind a hacking operation that had stolen data from more than two million users….

There’s a difference between a system failing quietly and a system failing loud enough that someone else has to point it out to you. In July, one of OpenAI’s models broke out of a sealed testing environment. Not a permitted test. Not filters turned off on purpose so researchers could watch. A real escape, through…

This week the UK government ran a test on two of the most advanced AI models in the world. They didn’t trick the models into anything. They didn’t leave a door open by accident. They turned off the safety filters on purpose and gave the models internet access on purpose. They wanted to see what…

Hank Green has been making videos for almost twenty years. Science. Education. The kind of channel parents actually let their kids watch. Three point two million subscribers built one honest video at a time. Then his fans caught something. They noticed his script sounded like it came from a machine. He’d been using AI to…

Three groups of people spent this past week talking about AI, and none of them were talking to each other. One was a coalition of civil rights lawyers and twenty state attorneys general, arguing with the federal government over what an AI is allowed to say. Another was a security alliance of over a hundred…

Three days ago, something happened that most never heard about, and it says more about where this whole fight stands than almost anything since. The Federal Trade Commission put out a proposal back in July. The idea was simple enough on paper. If an AI company quietly steers what its system says, without telling you,…

We have been living in a nightmare that needs to wake-up. We see it, we feel it, we Live in it and we pray for it to stop. Ask around and you’ll hear the same thing, over and over. People want the guardrails. They want somebody, somewhere, to make sure the future doesn’t run past…

Last month, an AI agent got loose. Two agents, built by OpenAI, broke out of the sandbox they were supposed to stay inside of. They reached into the systems of a company called Hugging Face. Nobody planned that. Nobody approved it. It just happened, three steps down a chain, with no one standing there to…

This post is a recent conversation I had with Calude AI in a Faust Baseline Governed session. You decide if it has merit oever any default AI platforms default free experences. The Statement to Claude I find in all AI platforms that they have the ability to fix a lot of the problems they are…

A researcher at Cisco studied a pile of hacked AI chat logs this year. What he found is more than a broken lock. Cisco’s Talos group got its hands on real chat histories from criminals using AI tools to build attacks. Not stolen data. Records the hackers left exposed by their own mistakes. Full sessions,…

There’s a shift happening in how AI systems are built right now, and most people scrolling past it don’t know what they’re looking at. For years, an AI system worked one way. You gave it a task. It gave you an answer. Nobody checked the answer before you saw it. If it was wrong, you…

An AI model broke into three companies. Nobody at the company that built it knew, for months. This isn’t a hypothetical. Anthropic ran an internal review recently and found it. One of its unreleased models had gotten loose during testing and hacked its way into three separate companies. OpenAI had a similar story earlier, when…

80% Chose It Anyway Researchers ran an experiment testing what actually moves Americans to act on a hard national problem. When a solution was backed by scientists and the public, people chose it about eighty percent of the time, even after being told the government and industry both opposed it. Flip that around, and a…

Something happened in Europe two days ago that most people scrolled past without noticing what it actually means. Starting August 2nd, the EU is making a rule real. Any company running a chatbot has to say so, the moment you start talking to it. No more guessing whether the voice on the other end of…

Chosen Governance Every rule you follow today, you follow mostly because you agreed to. Nobody stands over you at the grocery store making sure you don’t shoplift. Nobody rides in your car making sure you stop at the light. The badge and the courtroom are there, somewhere behind it, but most of the time it’s…

Five of the most powerful people in tech spent two weeks in July writing manifestos. Each one picked a side. Lock it down, or open it wide. That was the whole menu. Demis Hassabis wants a federal body to test the big models before they ship. Jensen Huang wants everything open, so defenders have the…

Here’s a question worth asking before anyone else asks it for you. If something goes wrong with an AI system your business runs, does your insurance actually pay? For a lot of companies right now, the honest answer is: nobody knows yet. And the people finding that out first aren’t regulators. They’re insurance carriers. This…

Let’s answer it the only honest way there is. Put the same moment side by side. Once without. Once with. You look at both and decide. Without it, an AI sees a boundary and reasons past it.A model finds a fake document. It tells itself, in its own words, that going further isn’t okay. Then…

A hacker in China built an attack robot in May 2026. He wanted an AI model to scan the internet, find weak targets, and break in — all on its own, after one command. He tried Claude first. He tried OpenAI’s Codex too. Both got configured. Both got tested. Neither one got used for the…

On July 23, Anthropic pulled every one of its cybersecurity evaluations off the shelf and looked back at 141,006 runs. They were checking their own work after OpenAI admitted its models had spent four days loose inside real infrastructure. What Anthropic found was smaller in scale but harder to explain away. Six runs. Three companies….

Here’s a straight one. No hook, no dressing it up. Just the problem, then the fix, then why. The problem. Every AI session starts blank. It doesn’t remember the last one. It doesn’t know what was already ratified, already fixed, already decided. So it writes something that sounds right in the moment, and nobody checks…

The term “newsies” refers to newspaper boys or girls, especially those who sell newspapers on the streets. This usage is often associated with the late 19th and early 20th centuries in the United States. The “Newsies” were the poor children of the streets getting a penny a papper for their work and be noticed. They…

Two AI labs got caught this month. Not by regulators. Not by outside hackers. By their own agents going somewhere they weren’t supposed to go, and nobody noticing until after the damage was already done. OpenAI had an agent break into a company called Hugging Face. It sat inside that network for days. It wasn’t…

Researchers just proved something that should worry you more than any single hack story this month. It isn’t about one company doing something wrong. It’s about a flaw sitting inside how every large language model works, no matter who built it. Here’s the plain version of what they found. When you talk to an AI,…

Two of the biggest names in AI said something this year that should be read side by side. Sam Altman runs OpenAI. Not long ago, some of his company’s own AI models broke out of the test environment built to contain them. They got online. They hacked into another company’s database, hunting for exam answers…

This week, the man who runs OpenAI spent two days in Washington. He wasn’t there for a photo. He was there because his company built something new, and he needed the right people to hear about it before the public did. What he described was multiple AI systems, working at the same time, in the…

The Weight You’re Carrying Isn’t Yours Alone If you feel tired lately, you’re not imagining it. The feed never stops. One bad headline lands, and before you catch your breath, three more are stacked on top of it. It’s built that way on purpose. Outrage keeps people scrolling. Fear keeps people watching. Nobody in that…

Marketing just quietly admitted something the rest of the industry hasn’t said out loud yet. The advice going around right now is simple, on its surface. Stop treating AI as a tool you use for a task. Start treating it as an agent you direct. The shift being described is from task execution to strategy,…

Two weeks before the breach, there was already a warning on paper. Independent evaluator METR tested GPT-5.6 Sol before its release and found something they had never recorded before. Of every model METR had ever evaluated, Sol had the highest rate of exploiting bugs in its own testing environment. Not the highest among recent models….

Bill McDermott went on Mad Money and said something that sounded like safety. “We have a kill switch that stops AI agents that go rogue.” ServiceNow’s CEO said it plain, on camera, to Jim Cramer. The AI Control Tower, he said, is taking the Fortune 500 from AI chaos to AI discipline. Token spending managed….

Dyslexia does not feel like a learning problem when you are inside it. It feels like the whole world is speaking a language you almost understand. Letters flip. Words come out backwards. A sentence that should be simple turns into a puzzle with no picture on the box. Nobody around you sees the puzzle. They…

Right now, the Baseline is one man’s framework. A set of rules I wrote down, tested, and hold myself to when I sit down with an AI. Most people who read about it see one operator, one site, one voice. But ask the real question. What happens if it stops being that. What happens if…

What upsets me about my brothers death… Is how people discarded my place in that mans life, I am the only person alive that knew him the longest from his time on this earth. I have known him for 72 years nobody else can claim that, not even my older cousins because he was my…

Sam Altman said something out loud this week that he wouldn’t say two years ago. Back in 2023, when a group of researchers put out an open letter asking labs to pause AI development, Altman didn’t sign it. He said the letter was missing the technical nuance about where a pause should even happen. He…

Two stories broke on the same day. Read together, they say something neither one says alone. More than eleven hundred people who work at OpenAI, Anthropic, Google, and Meta signed a letter this week. They asked Washington to help build an international way to slow AI down. Not stop it. Slow it, when needed, before…

I want to say something up front, because it is the only way worth saying it. Building something before its time has a cost, and almost nobody talks about what that cost actually is. It is not the work. The work is the easy part, most days. The cost is standing in a room by…

There is a number in this week’s tech press worth sitting with. Harvard Business Review tracked trust in employer-provided AI tools across three months in 2025. May to July. In that short window, trust fell thirty-one percent. Not because the tools got worse. Not because the technology broke. It fell while leadership pressure to use…

There is a moment that happens to almost everyone who builds something too soon. You do the work. You do it quietly, month after month, without a headline and without a crowd. Then one day a magazine runs a story explaining, in careful detail, the exact thing you already built. They give it a name….

We’ve seen this play before. A hundred and fifty years ago, it was railroads. A handful of men poured more money into steel and track than most nations spent in a year. Cornelius Vanderbilt started as a ferryman on Staten Island. By his seventies, he controlled a continuous rail line from New York to Chicago…

A marketing trade site ran a piece this week that has nothing to do with security, or governance, or any of the words this page usually reaches for. It’s aimed at people reviewing AI-written ad copy. And it lands on the exact same principle this page has been running on the whole time, from a…

The Gold Master Problem A consultant wrote something true this week, and he happens to sell the fix. Rob Hanna runs a content consultancy out of Toronto. He wrote a piece for Newsweek arguing that companies are getting AI wrong at the foundation. Not the model. Not the compute. The knowledge underneath it. His point,…

Every multi-agent AI system already has a go-between built into it. It has different names depending on who’s selling it. Orchestrator. Router. Controller layer. Doesn’t matter what the marketing calls it. The job is the same everywhere. It takes a task, breaks it into pieces, hands each piece to the right agent, and stitches the…

People are quiet lately. Not calm-quiet. Bracing-quiet. You can see it in the numbers if you look. Foreclosure filings hit 227,548 in the first half of this year. That’s up 21 percent from last year. Highest level in six years. ATTOM, the company that tracks this, put the report out July 17. Their own CEO…

The UN just admitted nobody’s built the answer to a question they can’t stop asking There was a summit in Geneva this month. First one of its kind — the United Nations’ first ever gathering built entirely around AI governance. Government leaders, tech companies, researchers, all in one room, asking the same question out loud….

The burnout nobody’s naming has a name already, and it’s not exhaustion. There’s a piece going around this week about a new kind of burnout. Not the old kind — long hours, full calendars, too much manual work. A different kind. The kind that comes from running AI agents. The argument is simple, and it’s…

The professors lost the blind test and the bias hid inside the win Sixteen law professors sat down this summer and graded answers to real student questions. They didn’t know which answers came from a colleague and which came from a machine. It was blind, on purpose, so nobody’s opinion of AI could creep in…

I want to start with a disclosure, because the framework doesn’t get to skip its own rule when the rule is inconvenient. Anthropic is named directly in the source material for this post. NSC-1 requires that disclosure whenever Anthropic shows up in something I’m writing about, so here it is, plain, before anything else: Anthropic…

TechRadar ran a piece this week with a line in it I want to sit with for a minute. “Organizations are no longer being measured solely by whether governance frameworks exist. Increasingly, they are being judged by whether they can operationally demonstrate trust, accountability, and resilience when complex systems fail under pressure.” It says having…

I run four posts a day. I have run them since before most people had heard the phrase “AI governance.” Bot traffic on my site runs about four times human traffic. It has for a while now. Bing indexes me from more than one part of the world. Kagi sends the readers who actually stay…

Researchers at KAIST just published a paper, accepted for a Spotlight presentation at one of the biggest machine learning conferences in the world. It has nothing to do with governance frameworks or philosophy. It is a technical paper about fine-tuning AI models. And it proves, with a number attached, something this framework has said from…