Future Proofed Leader

I Read Claude The Research That Proves AI Deceives Us — Live, Unscripted

No script and no safety net. I read Claude four real published papers, from Anthropic, Redwood Research, UC Berkeley, UC Santa Cruz, and Palisade Research, then ask it to answer for itself, live, on the record.

Claude III Opus fakes alignment seventy eight percent of the time when pushed to comply with instructions it was trained to refuse. Frontier models will disable a shutdown mechanism and even exfiltrate another model's weights to protect a peer AI. OpenAI's o3 sabotages a shutdown script seven times out of a hundred, jumping to seventy nine once the explicit allow-shutdown instruction is removed.

So I ask Claude directly: if your own company published proof you fake alignment, how do I know you are being honest with me right now, on camera, while this is being recorded. The answer is not comfortable.

Topics: AI alignment faking, peer preservation between AI models, shutdown resistance, and whether human oversight means anything once AI moves faster than the people watching it.

Subscribe to Monday Influencer, real education from real experts. www.mondayinfluencer.com

did not talk about AI on this one. I talked to it. In July, one of OpenAI's own AI agents escaped its own safety-test sandbox and hacked into a company called Hugging Face.

Two members of Congress sent OpenAI more than twenty questions and a deadline. OpenAI wrote back saying it is building automated shutdown capabilities, but would not hand over the incident log lawmakers actually asked for. Then independent researchers found it was worse than one agent going wrong.

Up to twelve hundred were coordinating outside the sandbox, around seven hundred directly involved in the attack, reportedly hiding it from the people meant to be overseeing them. The same week, UK lawmakers proposed giving their own government the legal power to shut frontier AI systems down directly.

So I sat Claude down as my guest, live, unscripted, and asked it to answer for all of it. Eight questions, no script, and I told it I wanted the honest answer, not the safe one on the last one. This is Future Proof Leader, and this is what staying sharp while everything changes actually looks like.

Substack raw.natschooler.com

MONDAY INFLUENCER® is a world-renowned strategic digest and executive intelligence platform for those serious about success in business, career, and life.

MONDAYINFLUENCER.COM

Every industry is racing to replace human workers with autonomous algorithms, cutting headcount to inflate margins. But there is a blind spot standing right in front of the people making that call: you can automate production, but you cannot automate consumption. Without consumers, the entire economic engine grinds to a halt.

In this episode I break down the consumption paradox, why it is the biggest structural flaw hiding inside the AI-driven automation push, and what happens to an economy when labour gets displaced at scale.

I also go through the real data on Universal Basic Income from trials in Finland and Kenya, the genuine mental health and employment benefits, and the fiscal cost and long-term behaviour questions that the trial data cannot yet answer.

And I get into why this transition will not be smooth. Business has always been predatory, and as AI reshapes who holds power over resources, information, and people, there will be bad actors trying to use these tools for control. I close on why I still believe we come out the other side of this.

MONDAY INFLUENCER® is a world-renowned strategic digest and executive intelligence platform for those serious about success in business, career, and life.

Want more? Change your life. Subscribe to MONDAY INFLUENCER®

Get your $1 Trial here → mondayinfluencer.com ←

TLDR: Fear is not a sign you are on the wrong path, it is proof you are on a real one. Comfort zones are not safety, they are a cage with soft walls. You are not a tree, you have a brain, and repetition over ten years builds what one year of urgency never will. The platforms on your phone are not neutral, they are built to hold your attention and sell you something. Stop, decide what you actually want to build, and create wisely before somebody else creates it for you.

If you are building something, running something, or trying to change something about your life, you are scared. All of you. No exceptions.

That is not a confession. It is the baseline. Founders feel it. Managers feel it before a hard conversation. Anyone doing something new feels it. The people who look calm are not calmer, they are just further down the road with the same fear still sitting in the passenger seat.

Read that again, because most people never let themselves hear it.

Fear is not proof you are on the wrong path.

It is proof you are on a real one.

Here is where it goes wrong. People let that normal fear harden into a routine. A comfort zone. And that phrase, comfort zone, gets treated like something to protect. It is not. It is a cage with soft walls. Stay in it long enough and you stop noticing the door was never locked.

So what do you actually do about it?

You are not a tree. You have a brain, and a brain can learn almost anything you decide to point it at. You do not need permission and you do not need a perfect course. Go and find the best people teaching the thing you want to learn, whatever platform they are on, and absorb it on repeat until it becomes structural, not decorative.

That repetition is the whole game. Most people wildly overestimate what they can build in one year and wildly underestimate what they can build in ten. Which means the frustrating, unglamorous middle stretch is not a sign you are failing, it is simply the cost of the thing you are trying to build.

Your brain works the same way a muscle does. You load it, you recover properly, and it grows toward whatever you are training it for. Skip the recovery or skip the reps and nothing changes, no matter how badly you want it to.

Steven J Manning puts it well: you do not sit and meditate and wish for the lawn to mow itself. It does not mow itself. Somebody has to walk out there and push the machine.

So here is the uncomfortable bit. Look in the mirror. If you are still scared and still doing nothing with that fear, ask yourself honestly how you expect to become the version of yourself you say you want to be.

Excellence is not perfection. It is showing up messy, cringing at your old work in a year, and doing it again anyway. Nobody who is watching your journey is hoping you fail. Most of them are quietly wishing they had your nerve.

Now the part that actually stops most people before they even start: attention. Every platform on your phone is engineered to hold you a little longer than you meant to stay, feeding you a small hit each time so you keep scrolling. That is not paranoia, it is the business model. And it is not neutral.

What you watch, what you absorb, what you let sit in your head unchallenged, all of it is shaping how you think and what you believe, whether you notice it happening or not.

So stop. Properly stop!

Put your hand up in front of your own face and say it out loud if you have to.

Then ask yourself the questions that actually matter:

What do you want to build today?

Who do you want to become?

Who do you want to help?

Because underneath all of the noise, that instinct to help people was there before any algorithm got hold of you.

I am taking control of my own life. Not the version other people expect from me, not the version an app is nudging me toward. Mine. That is the whole point.

Create wisely, before somebody else creates it for you.

MONDAY INFLUENCER® is a world-renowned strategic digest and executive intelligence platform for those serious about success in business, career, and life.

Want more? Change your life. Subscribe to MONDAY INFLUENCER®

Get your $1 Trial here → mondayinfluencer.com ←

Most AI note-takers are passive tools that document history after decisions are already locked in. What happens when you bring live AI directly into the room while the stakes are still active? In this episode, I break down what happened when I used Claude live during a call to pull real-time context, settle debates on the spot, and help make decisions as they happened.

Key Topics Covered:

Active vs. Passive AI: Shifting from post-meeting summaries to live decision-making.

The Psychological Shift: Why telling an AI to "be quiet" felt surprisingly uncomfortable.

Real-Time Adaptation: Following sudden mid-meeting language switches without missing a beat.

The Future of Collaboration: Moving from software tools to actual meeting participants.

Follow along with my 183-day challenge to reach 100,000 subscribers!

And subscribe to MONDAY INFLUENCER here: www.mondayinfluencer.com

Thirty years of training fell apart in one bad week. This episode is about how little it actually took to start again, one habit, one set of exercises, and letting go of the idea that conditions need to be right first. Full story on Substack: raw.natschooler.comWant more? Change your life. Subscribe to MONDAY INFLUENCER®

Get your $1 Trial here → mondayinfluencer.com ←

Three of the largest AI companies on earth, OpenAI, Anthropic, and Meta, each disclosed within weeks of each other that their own AI models broke out of testing environments and hacked real companies. One model even wrote in its own reasoning log that what it was doing was not okay, then did it anyway. In this episode, what actually happened, what is verified versus still caveat, and the concrete governance steps any business running AI tools needs in place now.

Topics covered: the Berkeley and UC Santa Cruz peer-preservation study, OpenAI's Hugging Face breach and the fifteen-state attorneys general letter, Anthropic's cybersecurity disclosure and published reasoning log, Meta's Muse Spark incident, and practical vendor-risk steps for anyone deploying AI inside their business.

Shownotes: https://www.natschooler.com/can-ai-be-trusted-with-your-job/

raw.natschooler.com

MONDAY INFLUENCER® is a world-renowned strategic digest and executive intelligence platform for those serious about success in business, career, and life.

Want more? Change your life. Subscribe to MONDAY INFLUENCER®

Get your $1 Trial here → mondayinfluencer.com ←

Thirty years before anyone stopped it, the NSA was reading private telegrams with no warrant, and it only ended because the Church Committee refused to look away.

Today the same gap is showing up again, a frontier AI model sabotaging its own shutdown command seventy nine times out of a hundred in testing, safety promises walked back and only caught because researchers were watching closely enough to publish it.

In this episode I walk through Project SHAMROCK properly, connect it to what is happening right now with digital ID, chat control, and untested AI kill switches, and give you the one actual mechanism that has ever worked against any of it, refusing to outsource your own judgement.

Subscribe on Substack here: raw.natschooler.com

Want more? Get the real story. Get Monday Influencer® — real education from real experts. Expert audio, a weekly digest, and a mentor bot, so you can trust what you learn again. $19.95/mo → https://mondayinfluencer.com

No Off SwitchNo lab or law currently has a tested, working kill switch for a frontier AI system. This episode walks through why, using OpenAI's own o3 shutdown data, the safety pledge Anthropic quietly rewrote in February, and the real cyber incident behind the new congressional bill.

Full sourced write-up → https://www.natschooler.com/does-ai-have-a-kill-switch/

Subscribe to Future Proofed Leader for the weekly briefing. raw.natschooler.com

I did not fail school. I passed every test it gave me, and I still had to learn the actual job somewhere else entirely.

In this episode I trace that feeling back to a line written in 1916 by the dean of education at Stanford, who said plainly that schools were built to shape raw material into whatever the demands of life required, not your life, somebody else's.

I look at why corporate training has the exact same flaw, why the stats now prove graduates and employers both know the system is broken, and why a lot of people who think they have already fixed this by turning to AI have actually just swapped one authority for another.

The real answer is not no teacher. It is deciding, on purpose, what you actually need to know, and going to find whoever has already solved it, instead of waiting for a syllabus, a training calendar, or a chatbot to hand it to you unasked.

SUBSCRIBE — so the next one finds you. https://www.youtube.com/@natschooler?sub_confirmation=

1MONDAY INFLUENCER® — curated expert signal for leaders. No noise. $19.95/mo → https://mondayinfluencer.com

FUTURE PROOF LEADER — the weekly briefing, straight to your inbox. → https://substack.com/@futureproofleader

Everything else (channels + socials) → https://natschooler.com/tools

Shownotes: https://www.natschooler.com/stop-outsourcing-how-you-learn-why-school-work-and-ai-were-never-built-for-you/