When Samsung lifted internal restrictions on ChatGPT use in its semiconductor division in March 2023, engineers used the tool for tasks it was designed to assist with, such as debugging proprietary source code. They also used it for analyzing chip-testing data. But within weeks, Samsung identified at least three separate incidents in which employees had uploaded confidential company information into the system.

The exposures didn't result from an external breach or intrusion. Employees themselves entered sensitive material into a third-party AI platform while trying to do their jobs more efficiently. They didn't mean to, but they still put the company's data at risk. In response, Samsung imposed temporary restrictions on generative AI tools, including ChatGPT, on company-owned devices and internal networks. Major financial institutions, including JPMorgan, Goldman Sachs, and Deutsche Bank, had already implemented similar restrictions.

They're taking steps to protect their data, and Samsung is doing the same.

The failure in each case wasn't one of security, but one of privacy. Security keeps unauthorized actors out, while privacy limits what authorized systems can see. It's a crucial distinction, and one that enterprise AI hasn't always gotten right. Most enterprise AI deployments achieve security without achieving privacy, which means the AI provider itself can observe every prompt, every uploaded document, and every interaction the organization sends through its platform. This is a problem, because it means the AI provider has access to sensitive information.

The organizations most at risk aren't the ones that failed to adopt AI. They're the ones that adopted it enthusiastically, invested in it generously, and built their most critical workflows around it without asking the question that should have preceded every other: Who owns the intelligence this system is building? It's a question that gets to the heart of the issue, and one that companies need to consider carefully.

The ownership problem is compounded by a structural shift in how organizations acquire AI capability. Menlo Ventures reported that in-house enterprise AI development collapsed from 47% to 24% in a single year. This decline happened because organizations concluded that the cost of building was compounding faster than the value of differentiation. They're looking for ways to cut costs, and outsourcing AI development is one way to do that. However, this conclusion may prove correct on cost, but catastrophically wrong on ownership.

Companies need to think carefully about what they're giving up when they outsource AI development.

Before the next vendor renewal, companies should ask themselves three questions. First, if the organization left this platform tomorrow, what would it take with it? If the answer is nothing beyond the raw data, the intelligence the platform built from that data stays with the provider. That's a problem, because it means the company won't own the insights it's paid for. Second, what can the AI vendor observe and learn from the organization's usage patterns, even under the current contract?

This is important, because it gets to the heart of the privacy issue. Third, is AI being deployed where it matters most, or only where the ownership risk feels lowest? Companies need to consider these questions carefully, because they'll help determine whether the organization is building an asset or renting one.

The answers will reveal whether the organization is building something it owns, or just renting a tool. For companies that have invested heavily in AI, the question of who owns the intelligence is crucial. As AI continues to play a larger role in business operations, the issue of data privacy and ownership will become increasingly important. Companies must carefully consider the risks and benefits of using AI tools, and ensure that they're taking steps to protect their intellectual property. They can't afford to ignore this issue, or they might find themselves at a competitive disadvantage.

  • Samsung's semiconductor division exposed confidential data to ChatGPT in 2023
  • JPMorgan, Goldman Sachs, and Deutsche Bank have implemented restrictions on generative AI tools
  • In-house enterprise AI development collapsed from 47% to 24% in a single year
  • Menlo Ventures reported the decline in AI development
  • Three key questions to ask before vendor renewal: what data would the organization take with it, what can the AI vendor observe, and where is AI being deployed

The issue of AI ownership isn't just a technical problem, but also a business one. As companies become more reliant on AI, they must consider the long-term implications of their decisions. The question of who owns the intelligence that AI systems build is a critical one, and companies must be careful not to sacrifice their intellectual property for the sake of convenience. They need to think carefully about what they're getting into, and make sure they're protecting their interests.

As the use of AI continues to grow, the need for clear guidelines and regulations around data ownership and privacy will become increasingly important. Companies must be proactive in addressing these issues, and ensuring that they're protecting their interests. They can't just wait for someone else to fix the problem; they need to take action themselves. The consequences of not doing so could be severe, and companies that fail to take action may find themselves at a competitive disadvantage in the future. They won't be able to keep up with their competitors, and they might struggle to stay afloat.

In the end, the decision to use AI tools is a complex one, and companies must weigh the benefits against the risks. While AI has the potential to bring significant benefits, it also poses significant risks, particularly around data ownership and privacy. By carefully considering these issues, and taking steps to mitigate them, companies can ensure that they're getting the most out of their AI investments. They'll be able to protect their intellectual property, and stay competitive in a rapidly changing market. It's a challenge, but it's one that companies can't afford to ignore.