IBM has published a public commentary arguing that open-source development and strong governance are essential for long-term AI security — a response to the growing reality that frontier AI models are now capable enough to require restricted deployment. The piece arrives at a pivotal moment for the industry, as discussions about who controls the most powerful AI systems intensify.
The Case for Openness at Infrastructure Scale
IBM Senior Vice President Rob Thomas published the commentary titled “Open Source, After Mythos,” arguing that once AI becomes foundational infrastructure, closed development models become harder to justify from both a security and accountability standpoint.
“The more critical the technology, the stronger the case for openness,” Thomas wrote. “If AI is becoming foundational, then openness is no longer a debate. It is a design requirement.”
Thomas argued that restricting access to powerful AI systems may look like caution in the short term, but at infrastructure scale, security tends to improve more through open scrutiny than through concealment. He warned that concentrating understanding of frontier AI capabilities “inside a small number of companies” could heighten rather than reduce systemic risks over time.
An Industry at a Crossroads
The commentary reflects a broader debate that is gaining urgency as AI capabilities expand rapidly. Anthropic recently launched a major initiative called Project Glasswing, a collaboration with more than 40 organizations — including Apple, Amazon, Microsoft, and Google — deploying a powerful new AI model exclusively for defensive security research and committing up to $100 million in usage credits to support the effort. The model’s capabilities, including advanced vulnerability discovery, placed it in a category that Anthropic determined warranted restricted access rather than general release.
IBM’s position represents the other side of that debate: that the most powerful AI tools will ultimately be safer and more trustworthy when their development is open to scrutiny, contribution, and collaborative governance — not locked inside proprietary systems controlled by a handful of private companies.
