Okay, this is kind of crazy.
Anthropic just published a report on the impact of AI on jobs and the number one most at-risk profession is computer programmers. Let that sit for a second. Computer programmers. The very people who built the tools, trained the models, and wrote the code that made AI possible in the first place. The industry out of which AI was literally birthed is now the one most at risk of being consumed by it.
Is that what shooting yourself in the foot looks like? Or is the more accurate reference the one from Greek mythology, where the Titans gave birth to the gods who would eventually overthrow them?
Anyway. Back to the report.
Here is the thing though. This was not supposed to happen this way. Back in 2013, Oxford researchers predicted that 47% of US jobs were at risk of automation and the jobs they flagged were truck drivers, cashiers, telemarketers and factory workers. The low skill, repetitive, physical stuff. That was the story everyone believed for years. Goldman Sachs later revised the picture, pointing toward programmers, accountants and customer service reps as being more exposed than originally thought, but even then most people assumed the first wave would hit the bottom of the pay scale.
Turns out the data is telling a very different story.
So What Did Anthropic Actually Find?
The report, published this week and titled “Labour Market Impacts of AI: A New Measure and Early Evidence,” does something genuinely useful. Instead of theorising about which jobs AI might affect someday, it looks at what AI is actually being used for right now, using real conversation data from Claude users, and maps that against real occupations. The result is a measure they call “observed exposure.” Not what AI could theoretically do. What it is already doing, today, in the real world.
And here is where it gets interesting.
The workers most exposed to AI disruption are not who you think. They are older. They are more educated. They are higher paid. Computer programmers top the list with 75% of their tasks already covered by AI activity. Financial analysts are right behind them. Customer service representatives. Medical records specialists. These are not entry-level, minimum wage roles. These are careers people spent years building degrees around.
Image credit: Anthropic
The Hiring Freeze Nobody Is Talking About
Here is the part that stopped me mid-scroll. High-risk jobs are not mass firing people. Not yet. What they have done instead is quietly stop hiring. Entry-level hiring in the most exposed occupations has dropped 14% since ChatGPT launched. And the group feeling it hardest? College graduates aged 22 to 25, who are roughly four times more likely to be affected than older workers in the same fields.
Think about that for a moment. You spend three or four years studying, you graduate, and the entry point into your chosen career has quietly started to close. Not because companies are replacing existing staff with AI. Because they are replacing the need to hire new people in the first place. The job did not disappear. The vacancy did.
Image credit: Anthropic
The Scariest Line in the Whole Report
There is one finding in this AI jobs report that I keep coming back to and it is this: AI models are already capable of automating most of this work today. The gap between what AI can do and what it is currently doing is not a technology problem. It is a regulation and adoption problem. Companies are moving slowly. Laws have not caught up. That is the only thing standing between the current situation and a much more dramatic one.
It is not a skill issue. It is an adoption issue. And adoption, unlike skills, can change overnight.
The Safe Side of the Job Market
On the flip side, roughly 30% of the job market does not even register on the exposure index. The jobs least affected by AI? Bartenders. Dishwashers. Lifeguards. Cooks. Roles that are physical, unpredictable, and context-dependent in ways that AI simply cannot replicate yet. The irony is not lost on anyone that the jobs long considered low status are now the most structurally secure.
One More Thing Worth Saying
Anthropic deserves real credit for publishing this. They are the company behind Claude, the very tool generating much of the data in this report. They are essentially saying: here is evidence that our technology may be disrupting the labour market, here is how we are going to track it, and here is what we know so far. That kind of transparency from an AI lab is genuinely rare and worth acknowledging.
The report is not meant to be the final word. The authors are clear that actual unemployment in high-risk fields has not meaningfully increased yet. But they are building the early warning system now, before the disruption becomes undeniable, which is exactly the right time to do it.
The future is here. It is just not evenly distributed.
And right now, it is landing hardest on the people who least expected it.
If you want breakdowns like this in your inbox every week, written in plain English without the hype, subscribe to the August Wheel newsletter. The future is moving fast. Let’s figure it out together.