AI is not stealing your data
Configuration, not conspiracy: what AI vendors actually do with your data...
There is a new favourite bogey man in the AI chatter. Apparently the big AI models are going to steal your data and use it to destroy your business. They are hooking you into a cycle of data sharing that you can’t escape. I have even seen this described as data expropriation.
Many of the people spreading this myth are AI advocates. Economists, policy influencers, VCs who want to invest in the next big thing, Alex Karp, even Satya Nadella are all on board with the narrative.
The basic premise is simple. When AI is used in your business it captures information about how people use the software. The AI providers then use this to train their models.
Why is this different from any other software? It’s not. Every app you use from Facebook to your highly secured corporate SAP instance is learning from how you use it.
The experts pushing the story describe this as a trap. As the model learns about your business, you become more dependent on the AI and it is impossible to escape.
Salesforce would describe this as configuration and integration. Every bit of large corporate IT does it. They say changing CRM or ERP is like ripping out your own heart.
This is true even for obviously commoditised applications. Next time you hear someone complain how much worse the user experience of Outlook is compared to their personal Gmail, ask them whether they have ever thought about switching the corporate platform. They will most likely just sigh and shake their head. Too difficult.
Does this mean you are paying twice for B2B software - once with money and once with data? No. The data side of the deal is that the software vendor uses what it learns to make better products.
SAP and Salesforce have built great businesses on top of this. In return, there are thousands of businesses using those same applications to deliver enormous value.
The relationship is symbiotic, not adversarial. Software vendors need your business as much as your business relies on their product.
Maybe OpenAI and Anthropic will do the same? So what?
That possibility at least explains the motivation of Microsoft and Palantir in spreading these myths. Karp and Nadella run businesses that already have oceans of this kind of data. Anything that stops new entrants matching this is clearly in their interests.
Those expert voices will throw in a further twist. The “so what” is that AI models will be able to copy your business and compete you out of existence. This is just fantasy:
“The real moat is not the data!”
Maurizio
Yes your software “knows” a lot about your business but not everything is in the data. If it was, why are the AI labs investing heavily in deployment capability? Consulting was supposed to be one of the easiest jobs for AI to automate. Yet billions are being spent to build more consulting capability.
AI is a tool that consultants can use to deliver value to their clients. If they want to compete with that, they need to build far more than just a better model.
The competitive advantage of your business does not sit in files and documents. AI expands your capability to use your own data for your own business. The questions that matter are: How can I use AI to deliver more value to my customers? How can I leverage AI to drive business growth?
None of the answers flow from restricting the ability of your AI to use your data. And yes, that data will also help those you share it with grow their own businesses. Do you want to share it with those making the most of the AI opportunity, or stick with vendors who are behind the innovation curve?


