60% of U.S. Jobs Now Exposed to AI, According to Global Safety Report

Additional Resources

• International AI Safety Report 2026

Transcript

Pete Newsome: 0:00

Welcome to Cornering the Job Market for Tuesday, February 3rd. I’m Pete Newsome, and today’s big headline is actually a number: 60%. That’s the share of American jobs currently exposed to artificial intelligence, according to the most comprehensive report on AI security and risk that’s ever been published. It’s called the International AI Safety Report, and it’s authored by more than 100 experts from 30 countries. This thing is massive. It has more than 1,400 citations across 156 pages, and I promise not to subject you to most of it. I will zero in specifically on the labor market. So let’s get back to that headline number.

0:38

60% of jobs in advanced economies like the US are now exposed to general purpose AI. Now, exposed doesn’t mean eliminated, and the report is careful to make that distinction. What it means is the task performed in those roles can either be automated by AI or significantly augmented by it. In emerging economies, that figure is only 40%. I say only, but that’s still a massive number in the big scheme of things. 10% would be significant. So 60% for countries like the US, 40% for underdeveloped countries. So what kind of jobs are we talking about here? Knowledge work, basically. So roles that rely heavily on writing, analysis, coding, research, customer communication, in other words, cognitive tasks that AI systems are becoming increasingly better at handling. Now the report acknowledges that AI isn’t perfect yet, far from it. It still hallucinates, it gets things wrong. We all know that.

1:38

But the report highlights something that I bring up in almost every video, and that’s that AI is being adopted faster than any comparable technology in history. Much faster than the internet or the personal computer. And that is extremely relevant. While it hasn’t happened yet, and everyone points to it, and this report does that a lot, it’s where we’re heading, and that’s what’s most important, right? We can’t look through the rearview mirror, we can’t even look out the side of the car. We need to be looking ahead to see where all this is taking us. Um, the report backs up that speed of adoption with numbers. In December 2024, about 30% of U.S. workers reported using AI tools, but by mid-last year, that number jumped to 46%.

2:23

And globally, more than 700 people now use AI systems on a weekly basis. The report points out that job loss hasn’t kicked in yet. So here we go again with what I’m starting to feel like is a historical fallacy for anyone to talk about AI’s impact by pointing to what hasn’t happened yet. I’ll quote right from the report. They say no discernible relationship between AI exposure or adoption and changes in overall employment have been seen yet. Yeah, yet, right? That, of course. So, you know, you could come up with a million analogies for this. What was it like when we were sitting at home in the 90s watching AOL load one line of an image at a time, right? If we based what had happened yet off of that, that wouldn’t have painted the picture of what was coming ahead at all.

3:15

So I I did notice something in the report uh that I’ll show on a graphic here that they used Anthropex Claude Sonnet and not Opus. And I’m using Opus 4.5 right now. That’s like comparing a middle schooler to a graduate student. So we all have to be very conscious of this. And I know I’m on my soapbox here, which I can’t help but do when I when I see reports and articles, and almost all of them like to talk about the state of things that we’re currently in, or looking back at history when it just sets a tone that I think is going to mislead a lot of people to make them too comfortable. Now, one of the big findings, and we’ve seen this a lot too lately, is that junior workers are getting hit the hardest. They cited multiple studies that found employment in AI-exposed occupations has declined for younger workers specifically, but it’s held so far for older workers since the release of ChatGPT. So, yeah, since the release of ChatGPT, I mean that first year was it was it was great compared to anything we had seen previously, but nothing compared to what we have now. So I’ll keep moving on. I’ll focus on the report for the rest of this, I promise.

4:33

Uh the freelance market data that kind of jumped out at me too within four months of ChatGPT’s release. Writing jobs on one of the major uh freelance platforms declined by 2%. So immediate impact there. Um writers’ monthly earnings have fallen by over 5%. So that’s a big drop. Okay, I’m not gonna say it. I was gonna go down the path of what I think is happening to these professions, sticking to the data. Um and no surprise at the same time from this uh same subject or the same section of the report, demand for machine learning programming increased by 24%. So that’s a huge, huge increase, and that’s that’s what what companies want, right? So just you can fill in the blanks yourself on what you think they’re going to do with all this AI that they’re developing. But moving on to the productivity side of the report, they found gains typically ranging from 20 to 60 percent in controlled studies and 15 to 30 percent in real-world work environments. So those are meaningful numbers, to say the least. And then finally, there’s um they acknowledge that there’s uncertainty in all this, right? The economists who contributed this report, they’re split. Some predict modest macroeconomic effects as it relates to impact on employment, um, because again, they’re pointing to historical patterns where automation created new jobs as it eliminated old ones. Okay, I can’t help myself, I have to say it, but don’t you think the AI is going to be better at the new jobs too? Maybe just me. I don’t know. Others in the report argued that if AI reaches human-level performance across cognitive tasks, it will significantly reduce wages and employment. Yeah.

6:24

Yes, of course it will. Um and so the concern, their concern, my concern, is that this time is just different, right? Because the other technologies weren’t thinking. They weren’t automating, they weren’t replacing physical labor and the things that we can do. So the smart move is to treat this moment as an opportunity to get ahead of the curve. If you are in the workforce or you run a business, either any side of the equation, you don’t want to be behind this curve. So that is always the message I’m going to continue to deliver, soapbox or not.

7:03

Um, so there we are. That is really the only big headline that we had today. Now, tomorrow, if you’ve gotten with me this far, we’re gonna be I’m gonna be talking about the Q1 survey that we just published on the Four Corner website today. So lots of interesting findings there, and I will definitely be sharing my opinions on that. But before we close, fun fact self-correction. People are more likely to notice their own typos if they change the font color of the text, if self-correction does that. Well, yeah, of course they are. That’s not even fun, that’s just necessary at this point, right? If I had to type without self-correction, like I did in what grade was I in? Ninth grade when I took typing class, I had to figure out the mistakes on my own. I keep my brain doesn’t work that way anymore. I’m lazy. I let the computer do it for me. That’s AI kicking in a little bit there too. So that is it for today. Please like, subscribe, share with anyone you think might be interested. Give me feedback you have, and I look forward to talking to you tomorrow.