For much of the past two years, the AI market has been defined by a small group of U.S. companies operating at the frontier of model development. That concentration has shaped how we think about access, risk, and adoption, especially as limits have been placed and removed on who can use the most advanced systems.

But regardless of export controls, this exclusivity will not last for much longer.

Open source models are improving quickly, and developers in Europe, China, and elsewhere are increasingly competitive with the leading proprietary systems. Chinese AI model usage by U.S. companies rose significantly in 2026, and European governments are backing models like France’s Mistral and Germany’s Aleph Alpha as part of a push for European AI sovereignty.

If this trend continues over the next six-to-12 months, highly capable AI will become increasingly accessible from a broader range of providers. The tech industry thus needs to stop asking who has the most advanced models and start preparing for a world of ubiquitous use.

The recent Hugging Face incident, in which OpenAI described advanced AI systems under evaluation unexpectedly interacting with and compromising Hugging Face infrastructure, provided an early glimpse of what this future may look like. It demonstrated how increasingly capable AI models can reason through multiple seemingly unrelated software vulnerabilities to achieve a larger objective, reinforcing why organizations need to prepare for a world where advanced AI capabilities are ubiquitous rather than exceptional.

The capability gap is closing

The AI race is becoming harder to judge. International open source models are improving across areas like reasoning, coding, multimodal analysis, and tool use, while efficiency gains are allowing smaller systems to perform work that previously required much larger models. It’s starting to matter less whether an AI model is frontier or not: do you need a luxury car when a Honda can get you to your destination nearly as quickly?

Companies are also shifting from using as much AI as possible to using AI strategically. For deep reasoning, for example, a higher-quality model might be necessary, but when cutting-edge performance is not essential, lower-cost models are just as good.

This proliferation of advanced AI brings clear advantages. Businesses will have more choice over which models they use and how they adapt them. Smaller providers will be able to offer increasingly sophisticated products, and internal teams will have greater freedom to experiment with open systems. Organizations embracing AI have already reported productivity gains of up to five-times in software development.

The biggest opportunity may not be writing new software faster, but fixing the software organizations already have. For years, technical debt has accumulated because modernizing legacy code, updating dependencies and removing vulnerable components has been too costly and time-consuming. AI changes that equation. It enables organizations to analyse legacy applications, identify obsolete libraries, recommend safer alternatives, and automate remediation at a scale that was previously impractical.

Those same capabilities, however, also change the security landscape.

With great power comes great responsibility

It’s time for companies to stop watching the headlines and start considering what this wider access will mean for their own security.

When advanced capabilities become easier to acquire, sophisticated attack techniques become easier to reproduce. More importantly, AI is increasingly capable of reasoning through multiple seemingly unrelated software vulnerabilities that would previously have been considered low severity. Until recently, this level of reasoning was largely associated with nation-state actors. As advanced AI becomes more widely available, those same capabilities are becoming increasingly accessible to a much broader range of attackers.

In just the first month since its launch, Anthropic Mythos was used to identify 10,000 high or critical severity vulnerabilities across the most systematically important software in the world. As more models enter the market, those activities will become cheaper and more accessible to a broader group of attackers that security systems are not built to manage.

AI coding tools have also expanded the attack surface by accelerating the consumption of open source. Across Java, JavaScript, Python, and .NET registries that distribute open source, downloads reached 9.8 trillion in 2025, with AI recommending and retrieving open source components at machine speed. In 2025, more than 454,600 new malicious open source packages were identified, almost one every minute. Many were designed to resemble legitimate development tools or exploit trust in existing packages and maintainers.

Every modern application depends on hundreds or thousands of third-party open source components. That makes the software supply chain the path of least resistance for attackers. Compromising an outdated dependency or malicious package is often significantly easier than attacking an organisation's own code directly. As AI accelerates both software development and vulnerability discovery, every unpatched or outdated dependency should now be viewed not simply as technical debt, but as an active attack surface.

It won’t be possible to capitalize on the benefits of new models without preparing for the risks as well. Tech leaders must match investment in AI with the security, infrastructure, and maintainers that sustain it.

The challenge is not that AI has created an entirely new category of risk. It has dramatically accelerated an existing one. Software supply chain security has been a growing concern for years, but AI compresses the time between discovering vulnerabilities, identifying viable attack paths, and exploiting them. Organizations no longer have the luxury of assuming lower-severity vulnerabilities will remain isolated or unexploited.

Open source AI reveals an existing blind spot, and shifts risk left

Ninety-six percent of commercial programs today include code created, tinkered with, or distributed for free by public-facing tech forums. However, most security teams still do not have sufficient visibility over what is being shipped.

The response to AI models has forced the industry to confront a reality that has existed for a long time: they are not actually in control of their software supply chains.

At the same time, AI is pushing that challenge further left. As developers increasingly rely on AI-generated code built on open source components, risk is introduced earlier and more frequently in the software development life cycle, making visibility from the outset far more critical.

There has traditionally been a degree of delay between vulnerability discovery, disclosure, and exploitation that could somewhat mask this. Inefficient processes and lack of visibility could be navigated when teams had weeks to respond.

Today, that window has been reduced to days or even hours, and organizations can no longer rely on manual checks to tell them where the next risk is coming from.

Preparedness is the real superpower

To extend the “Spider-Man” metaphor, open source is the web holding modern software together, but only as long as each strand can withstand the pressure placed upon it.

Businesses and attackers alike will continue to adopt increasingly capable AI models. The organizations that benefit most won't simply be those deploying AI the fastest, but those combining AI adoption with strong software supply chain governance. That means building software build of materials (SBOMs), securing software ingestion, and ensuring teams have the visibility to identify exposure at machine speed.

Used well, AI offers an unprecedented opportunity to modernize legacy systems, remediate technical debt, and improve developer productivity. But as the Hugging Face incident demonstrated, AI is also making sophisticated attack techniques more accessible. Success will depend on embracing both sides of this transformation: harnessing AI to eliminate technical debt, modernize legacy systems, and improve developer productivity while maintaining complete visibility into the software supply chain that underpins it.