VIENNA — AI and open source are critical for Europe’s tech future, extending to the economic bloc’s digital sovereignty and growth opportunities through investing in open source. Discussions during the Linux Foundation’s release of the Open Source Maturity in Europe report at the Open Source Summit Europe here this week underscored what is at stake.

“I don’t think Europe can afford to miss another technology wave,” said Gabriele Columbro, General Manager of the Linux Foundation Europe and Executive Director of the Fintech Open Source Foundation, while speaking at OSS. “This is not just about economics — it’s also about digital sovereignty.”

In the survey, 43% of respondents believe AI/ML would benefit most from being open source, highlighting the growing importance of open and transparent AI development in Europe. The survey findings indicate a strong interest in open source AI and machine learning (AI/ML), as respondents ranked AI/ML as the technology that would benefit most from open source. This ranked higher than operating systems (40%), cybersecurity (30%) and cloud or container technologies (29%).

The survey also reveals that investment in open source as an alternative to technology monopolies has emerged as the top priority, with 58% of respondents seeing this as the top priority for open source investments in 2024, compared to 39% in 2023. Security concerns about open source have also shifted, with 76% of respondents believing open source software is more secure than proprietary software.

However, much work still needs to be done to build on the increasing level of confidence in open source, especially in Europe. While companies like Mistral AI and Spotify dynamically support open source, Europe still lags behind its U.S. counterparts in tech. This is particularly true in the cloud space, where Europe lacks significant competitors to Google Cloud, AWS and Azure.

“There’s so much discussion on digital sovereignty, especially on the cloud side, and honestly, there are different degrees of technical savviness and politics within member states. Many efforts to create a sovereign European cloud have failed, partly because it’s so much harder as an afterthought,” Columbro said. “Once the hyperscalers are there, like Kubernetes, it’s much harder to build momentum for an alternative. Timing is now of the essence.”

@lf_europe, @finosfoundation, @theASF ‘s Gabriele Columbro: Europe will play a critical role in AI. #OSSummit pic.twitter.com/o5LDM3FDvt

— BC Gain (@bcamerongain) September 16, 2024

Licensing Challenges

A definition for AI/ML open source that adheres to the open source spirit and historical licenses is still in development, not only in Europe but globally. Licensing systems for AI/ML open source will be different and pose challenges not seen since the early days of the open source community.

Currently, what is deemed “open source” for AI/ML is not always truly open source. For example, Meta calls its Llama large language model (LLM) open source when it clearly isn’t. Other LLM providers are following Meta’s example.

With AI/ML, there must be a shift in the notion of open source as defined by the Open Source Definition—this shift is coming. The current definition, stewarded by the OSI initiative, doesn’t immediately apply to AI. “AI is code, so sure, you can use the open source definition for that,” Columbro said. “But AI is also training data, and that’s not covered by the Open Source Definition.”

The Linux Foundation is working on the Coalition for Content Provenance and Authenticity (C2PA), which deals with content provenance. They also have a license for open data, which maps the concept of open source licenses. Then there are the weights — the weights of the large language models (LLMs). “You train the model based on the input dataset, which creates weights and parameters that inform decisions,” Columbro said. “That is also a form of data, but it’s more like instructions in a certain format, so it’s not exactly code. The first problem is that there is no unified definition of open source AI because it’s a combination of code, data and models.”

@lf_europe, @finosfoundation, @theASF ‘s Gabriele Columbro: @lf_europe, @finosfoundation, @theASF ‘s Gabriele Columbro: «OS wouldn’t be here if it was for another fundamental ingredient, and that ingredient is trust. »@mindthegabz @thenewstack @linuxfoundation #opensourcesummit pic.twitter.com/BNdotiTXfS

— BC Gain (@bcamerongain) September 16, 2024

The Linux Foundation recently released the Model Openness Framework at KubeCon + CloudNativeCon EU in Paris in March. This framework evaluates AI language models “based on their openness,” Columbro said, describing it as a “gradient rather than a binary definition.” However, a more exact definition is to come.

“The Open Source Initiative has been working on defining open source AI for the past two years, and they should release their definition in November,” Columbro said. “OSI’s approach will focus on the principles, providing a binary definition of open source AI — either it is or it isn’t.”

The Model Openness Framework is more nuanced, “addressing different levels of openness,” Columbro said. One of the challenges is the lack of a clear definition. Companies may claim their model is open, but it often comes with licenses that impose restrictions, even though they label it as open source.

EU on Open

« Why Open Source AI Matters for Europe » at @OSSummit Europe: Justin Colannino, Microsoft; Sachiko Muto, OpenForum Europe (talking policy here); Stefano Maffulli, Open Source Initiative; Cailean Osborne, @thelinuxfoundation. @thenewstack pic.twitter.com/Yd3GMdN6JB

— BC Gain (@bcamerongain) September 18, 2024

Indeed, the lack of a clear definition of open source AI is problematic, said Sachiko Muto, senior researcher at the RISE Research Institutes of Sweden and chairman of OpenForum Europe. “The working definition tends to exclude projects that involve monetization,” Muto said. However, open source collaboration has “tremendous power in technology development, and the open source community is pretty good at mobilizing to influence policy as well,” Muto added. “Given that we don’t have many people on the ground in Brussels, we’ve still managed to get heard. Moving forward, though, we need to stay engaged throughout the entire policy process.”

Referring to the report, Columbro noted during the keynote how open source is “fully recognized as a piece of critical infrastructure,” and how “old myths” about security “continue to be dispelled.” Columbro concluded, “So it should be no surprise that open source has become such a big focus of EU investments over the last several years, and we hope, and are confident, this will continue. Every time open source grows in its maturity and impact, it’s going to face new challenges, whether regulatory challenges or others. We are effectively entering a new era of open source software regulation.”