Artificial intelligence is starting to feel less like a new gadget and more like the plumbing of modern computing. That is the argument IBM is making in a new commentary about where AI is heading and what that means for how it should be built. The company believes the industry is reaching a point where openness is not just a philosophical preference. It may become a practical requirement.
According to IBM Software executive Rob Thomas, technology tends to follow a familiar pattern. Early on, companies treat new ideas as tightly controlled products. That approach can help move fast and keep the user experience consistent. But over time those products evolve into platforms, and eventually into infrastructure that other systems rely on. Once that happens, the expectations change.
IBM believes AI is now entering that phase.
The company points to the growing influence of large AI models across everything from cybersecurity to software development. When organizations start depending on a technology to write code, secure systems, and guide business decisions, it stops being a simple tool and begins acting more like a foundational layer of computing.
That is where the open versus closed debate becomes more complicated.
IBM notes that closed systems can work well when a technology is still young. A single company can move quickly and control the direction of development. But once the technology becomes critical infrastructure, relying on one organization to anticipate every possible problem becomes unrealistic. No company can predict every vulnerability, every misuse, or every operational challenge that might emerge.
Recent developments in AI security research highlight that tension. New models are demonstrating the ability to find software vulnerabilities and even generate exploit code at levels that rival highly skilled human experts. That capability can obviously be used for defense, but it also raises uncomfortable questions about how much visibility the broader security community should have into the systems producing those results.
IBM’s answer is not secrecy.
The company argues that history shows security often improves when systems are exposed to scrutiny rather than hidden behind closed doors. Open source software has long relied on that principle. By allowing researchers, developers, and defenders to inspect code, test assumptions, and identify weaknesses, open development spreads the responsibility for improving security across a much larger group of people.
That does not eliminate risk, of course. But IBM suggests it changes how risk is handled. Instead of trusting a small group inside one company to catch every flaw, a wider community can stress test the technology under real world conditions.
In the age of advanced AI models, that philosophy may matter more than ever. If these systems are capable of discovering vulnerabilities on their own, concentrating knowledge about them inside a few private organizations could create new blind spots for the rest of the industry.
IBM also pushes back on the long standing claim that open source destroys business value. The company argues that what actually happens is a shift in where that value lives. When the foundational layer becomes shared, competition moves upward into services, reliability, orchestration, and specialized expertise.
The technology industry has seen this play out repeatedly. Open operating systems, open developer tools, and open cloud technologies did not wipe out innovation. In many cases they expanded the market by allowing more companies and developers to participate.
IBM believes artificial intelligence could follow the same path.
There is another piece of the puzzle too. Access shapes influence. When only a small group controls a technology, that group inevitably shapes how it evolves and where it is used. Broader access allows researchers, startups, governments, and institutions to take part in guiding that evolution.
For IBM, that is the real takeaway from the current moment in AI. The issue is not just what these models can do. It is who can examine them, question them, and improve them over time.
If AI is becoming foundational infrastructure for the digital world, IBM’s position is simple. Openness should not be treated as an optional feature. It should be part of the design from the start.
