Our company’s mission, inscribed in our bylaws, is:

“To develop, maintain at the state of the art, and sustain a complete suite of open source tools for data science to benefit France, Europe, and the world.”

Today, I’m proud to announce the launch of the independent body that holds us to it: the Probabl Mission Alignment Committee.

Why this mission matters

Supporting the scikit-learn ecosystem that millions of data scientists depend on every day

Scikit-learn is the Python library for machine learning. It has been downloaded over 4.6 billion times, with more than 200 million downloads last month. It’s a dependency for over 1.3 million GitHub repositories and over 28,700 open source packages. It’s cited in over 130,000 academic papers, including more than 7,300 Nature publications.

It’s taught in classrooms and bootcamps across the world, and used in research labs and production systems across every major industry, from manufacturing and energy to healthcare and financial services. More recently, as we can see from the skyrocketing downloads, it’s used by agents, too.

In short, scikit-learn is the open digital infrastructure of data science and machine learning that millions of practitioners depend on every day.

Scikit-learn is the open standard for machine learning, but it is not alone. When a data scientist opens a Jupyter notebook, they may reach instinctively for a set of tools: numpy and scipy for scientific computing, pandas for data wrangling, matplotlib for data visualization, scikit-learn for machine learning, PyTorch for deep learning, Transformers for accessing millions of pretrained models hosted on the Hugging Face Hub, and so on. The chart below illustrates a number of such open source tools.

The broader ecosystem of Python libraries for data science and AI follows a similar pattern to scikit-learn: massive, global adoption of these tools, which too often are developed and maintained by small teams of maintainers with uncertain long-term funding. This is the open source sustainability problem, and it is not hypothetical. It is happening now, to tools that millions of people and organizations depend on every day.

The open digital infrastructure of data science and AI

There is a useful analogy here, and it is not flattering to how we’ve treated open source software. We would not build roads and then expect the crews who paved them to keep them maintained for free, indefinitely, out of passion for asphalt. We would not build a public water system and then shrug when it starts to decay because no one could be found to fund its upkeep.

Yet this is roughly how large swaths of the software industry have treated open source. We have extracted enormous value from it. We have built trillion-dollar businesses on top of it. And we have systematically underinvested in the people and organizations who maintain it.

We should consider the Python libraries shown in Figure 1, among many other important ones that are not shown, as core open digital infrastructure of data science and AI. In particular, scikit-learn is open digital infrastructure because it is:

• Non-excludable: Anyone, anywhere, can use it freely.
• Non-rivalrous: One person’s use does not diminish another’s.
• Foundational: It sits beneath and enables a vast ecosystem of tools, research, and commercial applications, with over 200 million monthly downloads, over 1.3 million dependent repositories on GitHub, and over 28,700 dependent open source packages. Given this scale of adoption and dependency, the cost of replacing it would be staggering.

Scikit-learn is a global open patrimonial asset

Scikit-learn is at the same time a European and a global open source success story. It was created at Inria in Paris-Saclay, France in 2007 and the majority of the core maintainers are based in Europe, with a center of gravity in France and Germany.

In light of its widespread use, it represents one of the rare examples of genuine European leadership in the global AI software stack — comparable in this regard to ASML in semiconductors. The difference is that ASML has a market capitalization in the hundreds of billions based on a unique hardware business. Scikit-learn’s maintainers, until Probabl was founded, were largely sustained by the goodwill of research institutions and a handful of forward-thinking companies willing to sponsor the project.

Scikit-learn is also, proudly, a global open source project, with core maintainers and both technical and non-technical contributors from many countries, including the USA, Canada, China, Japan, India, Australia, and others. The maps of issue and PR contributors on OSS Insights show how global the scikit-learn community truly is.

The need for stewardship to sustain critical open source software

Open source projects — be it critical libraries like scikit-learn or PyTorch, or any other project in AI or beyond — are not self-sustaining. Behind every release is a team of engineers fixing bugs, reviewing pull requests, adopting new APIs, updating documentation, and keeping pace with a field that moves extraordinarily fast. Without sustained investment, even the most widely used tools in the world quietly decay.

This is where stewardship comes in. Stewards — whether companies like Probabl, Red Hat, or Hugging Face; foundations like the Linux Foundation, the Python Software Foundation, or the PyTorch Foundation; or research institutions like Inria — are the organizations that support the sustained, long-term health of open source projects. They may employ maintainers or fund contributors. They may apply for grants, manage governance, nurture communities, and make the unglamorous investments — documentation, security audits, backward compatibility — that keep a project trustworthy and usable over years and decades.

Stewardship is not always glamorous work, and it is rarely the thing that gets celebrated. But without it, the open source ecosystem that data science and AI depend on does not hold together. The question for every critical open source project is not whether it needs a steward, but whether it has one, and whether that steward has the resources and the commitment to do the job properly.

What we’ve done so far as a mission-driven company

Probabl was founded with the conviction that open source tools and open source digital infrastructure of such importance need a steward — one with the resources, the expertise, and, uniquely, the legal obligation to keep it running. That is what our Société à Mission status means in practice.

Since our founding, we have:

• Invested ~€2 million cumulatively in the development, maintenance, and sustainability of scikit-learn, skore, skrub, and skops employing core maintainers and contributors.
• Helped win competitive grants supporting the development of scikit-learn: from NASA ROSES program with Quansight Labs, and from the Essential Open Source Software for Science program by the Chan Zuckerberg Initiative and the Wellcome Trust with NumFOCUS.
• Launched the free and open source Skolar MOOC, ensuring that world-class machine learning education remains freely accessible to everyone. Since July 2025, over 13,700 unique users have levelled up their machine learning skills with Skolar.
• Produced over 115 tutorial videos, explaining technical concepts and best practices in machine learning with Python. All videos are freely available on YouTube.
• Raised €18.5 million in funding — one of the largest combined pre-seed and seed rounds ever raised in Europe for a commercial open source company — to sustain this mission long-term.

Introducing Probabl’s Mission Alignment Committee

A mission written into law requires oversight. That’s not a burden — it’s the point. Anyone can claim to care about open source. A Société à Mission has to prove it, annually, to a committee of experts and to the French government.

Our Mission Alignment Committee is composed of five domain experts drawn from open source software, enterprise technology, and the AI ecosystem. They are empowered to review our evidence, contest our findings, access our financial information, and escalate concerns directly to our Board. They produce an annual Mission Alignment Report submitted to an independent 3rd party auditor.

The five members of our inaugural Mission Alignment Committee are:

• Peter Wang, Chief AI & Innovation Officer, Co-Founder, Anaconda Inc.
• Emily Omier, Positioning consultant for open source companies and co-founder of Open Source Founders Summit
• Mark Surman, President, Mozilla
• Arnaud Le Hors, Senior Technical Staff Member, Open Technologies — Open Source Security and AI, IBM
• Cailean Osborne, PhD, Head of Ecosystem Development, Probabl

I’m grateful to each of them for lending their expertise, their independence, and their credibility to this endeavor.

Statements from our Mission Alignment Committee

Peter Wang, Chief AI & Innovation Officer and Co-Founder of Anaconda Inc.:

“Machine learning has become central to every aspect of human life, and trusted, innovative open source projects like scikit-learn are a crucial part of that foundation”

Emily Omier, Positioning consultant for open source companies and co-founder of Open Source Founders Summit:

“It’s important for data science tools to remain as independent and transparent as possible; keeping them open source is the best way to do so.”

Mark Surman, President of Mozilla:

“Open source creates choice. This is no different in data science and AI. With open source tools like scikit-learn and any-llm, data scientists can train, compare, and choose the best models for their use cases. Teams can swap models without rebuilding. Enterprises can switch vendors without starting over. That chain of freedom runs from the individual data scientist all the way to the top — but only if the foundations stay open and it takes dedicated people to develop, maintain, and sustain those foundations. That’s what Probabl exists to do.”

Arnaud Le Hors, Senior Technical Staff Member, Open Technologies — Open Source Security and AI, IBM:

“In an era where we see some companies pull the rug on open source communities by suddenly switching to a business license, it is great to see projects like scikit-learn that so many depend on in the data science and AI space being backed by companies like Probabl that are committed to staying on the open source path.”

Cailean Osborne, Head of Ecosystem Development at Probabl (our company representative):

“Open source is the bedrock of data science and AI. Python libraries like scikit-learn, pytorch, and many more are used daily by millions in classrooms, labs, enterprises, and public institutions all over the world. But open source projects like these — and crucially the communities that underpin them — don’t just sustain themselves. Their continued development requires dedicated stewardship and investment. Probabl was founded with this conviction, and that’s why I’m proud to be part of the team.”

What comes next

The Mission Alignment Committee will meet twice a year. It will produce an annual report. It will challenge us — and we welcome this, because that is exactly what it is there for.

The goal was never to build a compliance function. The goal was to build something that lasts — a sustainable model for stewarding open digital infrastructure for data science and machine learning that the world depends on. The Mission Alignment Committee is how we prove, year after year, that we mean it.

If you work in data science or AI, the tools that Probabl stewards are almost certainly part of your daily life. We don’t ask for anything in return. But we do ask you to care — about who maintains the foundations you build on, about whether they are adequately supported, and about what happens if they are not.

For more from Probabl

• Follow our latest updates on LinkedIn
• Subscribe to our monthly newsletter
• Check out over 100 tutorial videos on our YouTube account
• Level up your machine learning skills for free with Skolar

Probabl is a mission-driven company: This is what that means in practice was originally published in Probabl on Medium, where people are continuing the conversation by highlighting and responding to this story.

By Yann Lechelle, Executive President & Chairman of Probabl
& Daniel Whitenack, CEO of Prediction Guard

The transition from traditional software engineering to agentic AI has introduced a fundamental crisis of operational authority for enterprises.

In the classical software development lifecycle, the concept of ownership was binary: source code was either proprietary or open source, and dependencies were tracked through relatively static manifests generated as part of the release process.

Somewhere in the early 2010s, software began eating the world in various forms of subscription services — from Infrastructure-as-a-Service to Platform-as-a-Service via APIs, all the way to Software-as-a-Service — obfuscating every bit of software dependencies formerly processed by the owners of older licensing models.

Today, as enterprises increasingly adopt AI through usage-based APIs, managed cloud services, and third-party model integrations, the infrastructure of intelligence has become dangerously opaque. To make matters worse, more and more enterprises today operate in a state of shadow AI — the unsanctioned use of AI tools by employees without formal oversight by IT or InfoSec teams. As a report by IBM highlights: over a third of employees acknowledge sharing sensitive work information with AI tools without their employers’ permission, and the risks compound at every layer of the stack, including the model lifecycles, data lineage and flow, connected tools like MCP servers and the runtime behaviors.

The results are concerning: Enterprise leaders struggle to know the full extent of their ownership of data science or AI in their organization.

For enterprise leaders navigating this challenge, the AI Bill of Materials (AIBOM) is a concept worth understanding. Analogous to the Software Bill of Materials (SBOM) that is increasingly used by enterprises to understand and secure their software supply chains, the AIBOM is a standardized, machine-readable inventory of every model artifact, tool connection, dataset, and infrastructure component involved in building and operating an AI system. The AIBOM gives enterprise leaders answers to questions like what models are we running, their degree of openness, and who controls them? What data flows through them? What third-party dependencies and tool connections are active? Where are our compliance exposures?

In this article, we argue that AIBOMs are a promising governance tool for enterprise leaders to understand, strategize, and take action on their ownership and control of AI or the lack thereof.

The Crisis of AI Ownership and Control

The rapid adoption and development of agents has led to a fragmented AI stack that extends far beyond single models towards distributed harnesses supporting agentic workflows and automation. Enterprises may stitch together third-party APIs like Azure OpenAI or AWS Bedrock, various vector databases, and agentic frameworks, creating a web of integrations where true control and ownership is absent. In these environments, security policies are often enforced externally by vendors, and the handshakes between models and internal tools are opaque. While traditional models presented risks like bias and data leakage, agents introduce a new level of risk. They don’t just provide outputs; they take actions. This shift dramatically compounds risks like unauthorized actions and privilege escalation

The Loss of Ownership and the Risks of Lock-In

True AI ownership starts with control and capability. One cannot govern what one does not own, audit, and actively manage. However, the current paradigm of AI consumption, characterized by a reliance on external vendors for security boundaries, update cycles, and alignment decisions, leaves enterprises with limited agency over their own intelligence assets.

There is a trend in industry to prioritize the magic and speed of frontier AI agents over verifiable control and management of the underpinning supply chain, pushing enterprises toward all-in cloud-only or AI-only strategies. This creates a pay-as-you-go trap, where on-demand pricing for compute and tokens leads to uncontrollable OPEX expansion, benefiting the supplier’s profits rather than the enterprise’s long-term value. When an organization gives a supplier open-ended access to its bank account while losing the ability to understand or reuse its own experiments, it loses its freedom of movement.

Separately, when such systems are managed by US providers — as is often the case given their prevalence — they fall under powerful US extra-territorial laws such as the Cloud Act or FISA (particularly Section 702), giving the US administration the ability to tap into potentially sensitive data that belongs to individuals and corporations, regardless of their citizenship or place of incorporation. By default, AI systems handle lots of potentially sensitive data.

Ownership is not just a legal status but an operational capability. It means owning the lifecycle, versioning, governance enforcement, and runtime behavior of every asset in the AI system. It is about operational agency and the ability to control the connective tissue of the stack. This ensures the organization can audit (and own) every interaction between a model and its data sources, regardless of whether they reside in a private cloud or a hybrid environment. This requires a move from AI consumption (vendor-dictated) to AI operation (enterprise-owned).

How AIBOMS can help enterprises own their AI

What are AIBOMs and how they differ from SBOMs

The historical origins of the Bill of Materials (BOM) lie in manufacturing, where it served as the definitive list of raw materials and sub-assemblies required to manufacture a product. In the digital realm, this concept first evolved into the Software Bill of Materials (SBOM) to address the complexity of modern software supply chains, where open source components often comprise the majority of a code base. Initially concerned primarily with licensing, the SBOM eventually became a regulatory and security cornerstone, notably codified in the United States through Executive Order 14028, which mandated SBOMs for software purchased by the federal government. The Cyber Resilience Act in the EU additionally introduces attestation programs recommending the production of SPDX 3.0 SBOMs.

However, SBOMs are inadequate to capture the nuances of machine learning (ML) and AI systems, exposing a critical documentation gap. Traditional SBOMs were designed for static logic (code that remains constant between versions). AI systems, conversely, are typically non-deterministic entities; their behavior is determined not just by code, but by the distribution of training data, the stochastic nature of inference, and the continuous evolution of model weights through retraining or fine tuning. The actual AI model often constitutes only a small fraction of a production system’s code, yet it introduces the most significant uncertainty. This complexity necessitated the development of the AIBOM as an extension of the SBOM, treating artifacts like datasets and model architectures as first-class citizens in the supply chain.

The shift to AIBOMs began in earnest around 2021, when working groups within the SPDX and CycloneDX communities realized that existing standards could not meaningfully capture AI-specific artifacts like fine-tuned checkpoints and data pipelines. The motivation was clear: transparency and explainability are essential for trust, and without a machine-readable record of model origins, organizations could not verify the security, provenance, or governance enforcement of the AI systems they deployed.

While the AIBOM remains the major parent umbrella term used to describe the broad requirement for transparency in AI systems, a more specialized concept, the Agentic Bill of Materials, has emerged to capture the complexities of agentic systems. Because agents tend to be active rather than passive and backed by a distributed agent harness, evolving AIBOM specifications may need to track the specific risks of autonomous action, such as unauthorized actions and privilege escalation, invocation of an inventory of tools, operational independence, required human-in-the-loop approvals, and connections to MCP servers. But the reality is that the AI industry is evolving faster than the development of standards and governance frameworks like AIBOMs. Closing the gap will require open, cross-industry collaboration on tooling and specifications that no single vendor or organization can deliver alone.

The Seven Pillars of an AIBOM

A functional AIBOM must provide a structured inventory across multiple layers of the AI ecosystem to enable effective security operations. The complexity of modern AI means that a simple list of models is insufficient; the AIBOM must include key layers making it actionable and comprehensive.

1. The Data Layer

The success of an AI model is inextricably linked to its training data. The data layer of an AIBOM captures the provenance, licensing, and processing history of all data assets. This includes:

• Training Data: Metadata on origin, sampling methods, and any preprocessing steps like normalization or scaling.
• Inference-time Data: Records of data accessed during production via feature stores, vector stores, or APIs.
• Data Stores: Documentation of the underlying storage systems, including vector databases and their specific configurations.

2. The Model Layer

This layer tracks the models themselves and their iterative evolution. In an AI system, the model layer ensures that teams maintain control over versions and avoid stealth updates from providers that can introduce regressions.

• AI Models: Pre-trained weights with information about providers, licensing, origin, etc.
• Fine-tuned Checkpoints: Specific versions customized for internal, domain-specific, task-specific, or other use cases.
• Model Configurations: Hyperparameters, architecture details, and deployment contexts.

3. The Dependency Layer

AI systems and agents are rarely standalone; they exist atop complex, and often distributed, software stacks. The dependency layer identifies vulnerabilities in the ML/AI frameworks (e.g., PyTorch, scikit-learn), SDKs (e.g., LangChain), and other runtime dependencies necessary for training and serving. Additionally for agentic systems, this layer may identify an inventory of tools, functions, and/or MCP servers available to one or more agents.

4. The Infrastructure Layer

Tracking the hardware and resources supporting AI workloads is crucial for both cost optimization and sovereignty. This pillar includes details on GPUs/TPUs used, network paths between components, and the specific cloud regions or on-premises boundaries where workloads execute.

5. The Policy and Governance Layer

Unique to the AIBOM is the need to document the controls applied to the system. This includes things like NIST and OWASP-aligned policies enforced by a governance control plane, such as prompt injection filters and sandboxed execution environments. Critically for agents, this layer may define autonomy tiers (i.e., does the agent need human-in-the-loop approval to perform certain actions like spending money or deleting files).

6. The Observability and Runtime Layer

A dynamic AIBOM must evolve alongside the system. This layer includes information on how model drift, performance metrics over time, and identified risks such as bias or sensitive information disclosure flow to an audit log.

7. The Compliance Layer

Especially for regulated industries, the AIBOM must include fields for human oversight design, fairness assessments, and evidence of compliance with frameworks like NIST 600–1, AIUC-1 or the EU AI Act.

Technical Implementation of AIBOMs

For an AIBOM to be useful in an enterprise context, it must be machine-readable and interoperable across different tools. Two dominant standards have emerged: SPDX (ISO/IEC recognized) and CycloneDX (OWASP-managed). Application of these standards to AI systems and agents is being actively explored by, e.g., the OWASP AIBOM project.

SPDX 3.0: Detailed Profiles for AI and Datasets

The System Package Data Exchange (SPDX) 3.0 specification represents a major re-architecture, introducing Profiles to simplify the documentation of specific use cases.26 The AI Profile and Dataset Profile were separated to clarify the distinct provenance requirements of models and data.

The primary class in the AI Profile is the AIPackage, which inherits from the Software Package class but adds 36 specialized fields, including:

The DatasetPackage in the Dataset Profile includes mandatory fields for datasetType and optional fields for anonymizationMethodUsed, confidentialityLevel, and dataPreprocessing. This allows for a granular view of the ingredients used in training, ensuring that licensing and privacy obligations are met.

CycloneDX 1.7: Security-First ML-BOM

CycloneDX focuses on integration with security workflows and DevSecOps tools. It introduced ML-BOM support in version 1.5, allowing for the embedding of model cards and extensive property taxonomies.

The CycloneDX approach uses the machine-learning-model component type and provides specific objects for:

• Model Card Data: Intended uses, technical limitations, and performance tradeoffs.34
• Citations and Evidence: Added in version 1.7 to enable verifiable chains of provenance, allowing authors to declare where specific data originated (e.g., a build system or manual input).
• Formulation: Describes how the component was manufactured, such as the specific steps in a training pipeline.

Engineering teams often find CycloneDX more agile for rapid vulnerability matching, while legal teams prefer SPDX for its extensive license expression language.

Building an Enterprise AIBOM Program: A Step-by-Step Guide

Transitioning to an AI strategy that you fully own requires a methodical approach to AIBOM management. Engineering leaders should view the AIBOM as infrastructure, not paperwork.

Step 1: Define Scope and Ownership

Not every notebook or prototype requires a full AIBOM, but production systems and externally sourced models do. Ownership should be clearly assigned: MLOps teams manage the automation pipeline, data scientists provide model lineage, and compliance teams maintain oversight.

Step 2: Inventory of All Assets

Map every entry point for AI in the organization, including SaaS tools with embedded AI and internal prototypes. Gaps usually appear at integration points, so a comprehensive inventory of internal, cloud, and third-party models is essential.

Step 3: Automate Extraction via CI/CD

Manual documentation does not scale. Metadata collection should be embedded into training and deployment pipelines.

Step 4: Validate and Sign

Once data is collected, it must be validated against SPDX or CycloneDX schemas. This prevents mismatches that compromise audit readiness. Crucially, the AIBOM should be digitally signed and versioned alongside the AI system it describes. This enforces integrity and ensures that the version history is tamper-proof.

Step 5: Operationalize via the Control Plane

Integrate the AIBOM with a control plane to enforce real-time policies. This allows the AI system to:

• Audit Handshakes: Govern every interaction between models, tools, and internal systems.
• Monitor for Drift: Automatically flag when a model’s runtime behavior deviates from the documented parameters.
• Roll Back Changes: Use the immutable AIBOM logs to investigate failures and roll back to a known-safe state.

Moving Toward Active AI Ownership with AIBOMs

The transition from AI consumption to AI operation is not merely a technical shift but a fundamental requirement for the modern enterprise. As AI becomes the central nervous system of enterprises in regulated industries, the illusion of control provided by vendor-defined security boundaries is no longer sufficient. Operational resiliency and control is defined by the ability to own, audit, and actively manage the entire lifecycle of intelligence.

The AI Bill of Materials stands at the center of this transformation. It is a technical solution that promises benefits for transparency, security, and regulatory compliance, and provides enterprise leaders with the agency to own and control their AI.

AIBOMs are, of course, not a silver bullet. The reliability of AIBOMs will initially be constrained by supplier compliance, requiring time and internal validation efforts. The lesson from SBOMs is that they become reliable only through demand for quality and standardization, meaning that the organizations that start requiring AIBOMs and hold suppliers accountable for quality today will be the ones who shape what good looks like tomorrow. Crucially, tracking AIBOMs is the first step toward moving from symbolic governance to active authority, ensuring that the enterprise has the ultimate authority over how its intelligence operates.

In 2026 and beyond, the most successful enterprises will not necessarily be those with the most advanced frontier models, but those that understand their AI context best and maintain full operational ownership over their assets. Reclaiming ownership is a design principle, not a checkbox, and the AIBOM is an essential tool to help enterprises be resilient going forward.

Yann Lechelle is the Executive President and Chairman of Probabl, the company by the creators of scikit-learn.

Daniel Whitenack is CEO of Prediction Guard, a company providing an enterprise AI control plane that enforces AI governance within a company’s infrastructure (including the ability to track and export AIBOMs).

We thank Arnaud Le Hors and Cailean Osborne for their input on earlier drafts.

For more from Probabl

• Follow our latest updates on LinkedIn
• Subscribe to our monthly newsletter
• Check out over 100 tutorial videos on our YouTube account
• Level up your machine learning skills for free with Skolar

The Enterprise Case for AI Bills of Materials: Understanding AI Ownership, Risk, and Control was originally published in Probabl on Medium, where people are continuing the conversation by highlighting and responding to this story.

In the world of technology, we often talk about “infrastructure” in terms of cables, data centers, and cloud regions. But for the global data science community, the most critical infrastructure isn’t made of silicon or steel–it’s made of code. Specifically, it is built upon open source libraries like scikit-learn.

When we founded Probabl, we did so with a profound sense of responsibility. As an organization founded by the creators of scikit-learn, we recognized that this library had grown beyond a mere tool; it had become the de facto open standard for machine learning worldwide.

Today, I am proud to announce the launch of Scikit-learn Central, a new digital hub designed to visualize, unite, and stimulate the sprawling ecosystem that has grown around this foundational library.

The scale of a global standard

To understand why stewarding scikit-learn is important, one only needs to look at the numbers.

Scikit-learn has surpassed 4.1 billion downloads. Every single month, another 160 million downloads occur as data scientists, developers, and students pull the library into their environments. On GitHub, it serves as the foundation for over 1.3 million repositories and 27,000 packages. It is a project that has been sustained by the efforts of over 3,100 contributors and that has received over 65,000 stars. For comparison with other popular libraries for machine learning like XGBoost and deep learning like PyTorch and TensorFlow, take a look at the chart below that shows yearly downloads from PyPi and Conda. Looking at this metric alone over the past few years, scikit-learn commands a 70%+ growth year-on-year which is staggering.

Downloads and stars only tell part of the story. The impact of scikit-learn is also measured in the progress of human knowledge. The seminal paper, “Scikit-learn: machine learning in Python,” has 128,960 citations, and scikit-learn is cited in over 7,000 Nature publications. This is world-leading science in action. From identifying genetic markers in cancer research to optimizing the complex climate modeling models, scikit-learn provides the mathematical building blocks that allow experts in biology, physics, and ecology to apply machine learning to their specific domains without needing a PhD in applied mathematics.

Mission-driven stewardship

Probabl is not a typical tech company. We are a Société à Mission under French law–a status similar to a B-Corp, with our mission integrated into our corporate bylaws: “To develop, maintain at the state of the art, and sustain a complete suite of open source tools for data science to benefit … the world.”

We believe that open source tools of this magnitude require a sustainable commercial model that respects the community. Our mission is to shepherd scikit-learn, helping the community use it and improve it. That means ensuring the core remains robust while fostering the “extended universe” of libraries that make the workflow complete.

Introducing Scikit-learn Central

The scikit-learn ecosystem is vast. While the core library provides the supervised and unsupervised algorithms, the wider community has built specialized tools to handle the nuances of modern data science.

Scikit-learn Central is our attempt to make this ecosystem navigable. It is a catalog of the “building blocks” that turn a simple model into a production-ready pipeline. When you visit the catalog, you see the full breadth of what is possible:

• Data preparation: Tools like Skrub simplify the often-tedious process of data cleaning and feature engineering, ensuring that the “garbage in, garbage out” mantra doesn’t derail your project.
• Workflow acceleration: Skore helps data scientists move faster, with smarter cross-validation, automated evaluation reports, and methodological guidance — catching common pitfalls before they reach production.
• Powerful predictions: We highlight the deep integration with libraries like XGBoost, which rely on scikit-learn’s API to deliver state-of-the-art gradient boosting.
• Operational excellence: For those moving to production, MLFlow provides the MLOps framework necessary to track and deploy models at scale.
• Trust and explainability: SHAP offers data scientists tools to evaluate and understand why a model makes the decisions it does.
• Tabular foundation models: The new wave of Tabular Foundation Models often depend implicitly on scikit-learn. For instance, TabICLv2 uses scikit-learn during pretraining to generate good synthetic data that shapes the final model.

An invitation to the wider community

At Probabl, we know that the most compelling arguments for open source aren’t found in documentation, but in execution. That is why, alongside the catalog, we are building a library of use cases.

We are inviting data scientists from every corner of the globe to share their code and their stories. Whether you are using libraries for time-series analysis in finance, nilearn for neuroimaging, or building a custom fraud detection engine for a global bank, your work can inspire and educate others.

Scikit-learn succeeded because it made complex mathematics accessible to everyone. Scikit-learn Central aims to succeed by making the entire ecosystem accessible. We are building a future where open-source machine learning is sustainable, transparent, and more powerful than ever.

I invite you to explore the catalog on scikit-learn central, contribute your use cases, and join us in this next chapter.

For more from Probabl:

• Follow our latest updates on LinkedIn
• Subscribe to our monthly newsletter
• Check out over 100 tutorial videos on our YouTube account
• Level up your machine learning skills for free with Skolar

Announcing Scikit-learn Central was originally published in Probabl on Medium, where people are continuing the conversation by highlighting and responding to this story.

By Yann Lechelle, Executive President of Probabl

GitHub just released fresh data from the GitHub Innovation Graph, providing accessible data and aggregated statistics on software development activity around the world in the millions of public repositories hosted on the platform. As the software company committed to the stewardship and long-term success of scikit-learn, this data is invaluable for us at Probabl in helping us to understand global trends in open source and to situate scikit-learn in them.

The data paints a clear picture of the global nature of open source. We see that the EU is now the world leader by some metrics like git pushes, unthroning the USA for the first time. At the same time, the USA leads when it comes to the number of developers on GitHub (30 million!). Countries like India, the UK, Brazil, Korea, Japan, and China, among many others, are also world-leading hubs of open source software development on GitHub [1].

The data also underlines that national or regional economies are not islands: open source thrives because of global interlinkages and collaboration. For example, in 2025, EU-based open source projects received millions of code contributions from developers across the world, including the USA (over 934,000), UK (over 553,000), and India (over 347,000).

These stats resonate with us at Probabl as the steward of scikit-learn, the global open standard framework for machine learning with 3.9 billion downloads and 1.3 million dependants on GitHub. While most core contributors are based in the EU (e.g. France and Germany), scikit-learn also counts on core contributors from the USA, Australia, and China. Also, scikit-learn receives most of its contributions, in particular issue reports and pull requests, from developers all over the world. As this map from OSS Insights shows, most pull requests come from developers in the USA, India, the UK, Canada, Japan, and China, among many others. The bottom line: the scikit-learn ecosystem is no exception to the global trend of open source ecosystems shown by GitHub, scikit-learn ecosystem is global.

In terms of users, scikit-learn has the wind in its sails with a growing curve of downloads (around 150 million new downloads per month). Interestingly, when viewed alongside libraries like PyTorch, the global standard for deep learning, scikit-learn maintains a significantly broader footprint in terms of downloads, reflecting the vast and growing demand for classical machine learning across almost every industry [2].

At Probabl, we’re regularly analyzing the underlying causes and trends behind the skyrocketing growth of the scikit-learn ecosystem. One thing is for sure: the users and downloaders are global. Hey GitHub Research Team, we should partner up and investigate usage trends together!

Thank you to the GitHub team for making this data available. We recommend reading more about the GitHub Innovation Graph and diving into the data. And, of course, a big thank you to everyone worldwide who contributes to scikit-learn in one way or another!

Notes

[1] We suspect China is underrepresented in these numbers, possibly due to unreported geolocations and/or use of Chinese alternatives like gitee. For example, there’s a scikit-learn mirror on Gitee: https://gitee.com/mirrors/scikit-learn

[2] We acknowledge that pypi downloads are not the only way to download a python library and therefore the download stats may be incomplete for scikit-learn and/or PyTorch.

For more from Probabl

• Follow our latest updates on LinkedIn
• Subscribe to our monthly newsletter
• Check out over 100 tutorial videos on our YouTube account
• Level up your machine learning skills for free with Skolar

The global pulse of open source: Insights from the GitHub Innovation Graph and scikit-learn was originally published in Probabl on Medium, where people are continuing the conversation by highlighting and responding to this story.

Trade and geopolitical tensions between the US, Europe, and Asia are no longer limited to traditional industries. The digital sector is now at the forefront, with structural consequences for investment, economic models, and technological dependencies. In this context, tariffs — real or anticipated — are becoming catalysts for change, forcing the European tech ecosystem to rethink its strategy.

The Domino Effect of U.S. Retrenchment

As U.S. authorities increasingly shield their domestic market and national champions, American investors are concentrating their capital on low-risk, low-exposure projects. This trend is already visible in a decline in transatlantic deal flow, particularly at the early and growth stages.

For European investors, this contraction of U.S. capital inflows opens a window. But it’s a double-edged sword: it requires taking more risk on local technologies that are often less mature or underfunded. This forced rebalancing could, over time, favor the emergence of more autonomous regional dynamics.

SaaS Price Inflation: An Inevitable Consequence

In this tense context, the SaaS economy — dominated by U.S. providers — is directly impacted. Two forces are converging toward price increases:

1. Potential EU retaliatory measures, such as digital tariffs or market access restrictions.
2. Price adjustments by U.S. providers, who may raise international rates to offset declining volumes or the increased complexity of operating across fragmented markets.

In both scenarios, European companies could find themselves trapped in an expensive dependency model, with limited bargaining power and a lack of pricing transparency.

Open Source as a Strategic Model

This environment is driving companies and ecosystems to seek more sustainable alternatives. Open source is making a comeback — not just for technical reasons, but as a strategic choice:

• Fewer dependencies: open technologies allow greater control over infrastructure and product roadmaps.
• Increased interoperability: they facilitate integration in hybrid and regulated environments.
• Digital sovereignty: they offer a concrete path to technological autonomy for Europe.

These benefits are no longer theoretical. They are becoming increasingly tangible, backed by a new generation of investors.

Investment Theses Are Evolving

Funds like Serena, Red River West, and Runa Capital are now fully integrating open source into their investment theses. This shift is significant: it reflects the realization that closed, centralized, and geography-agnostic models are less resilient in a fragmented world.

These funds are betting on hybrid models, where value is created as much by community as by product performance, and where regulatory alignment becomes a competitive advantage.

Europe at a Crossroads

Europe has long suffered from being caught in between — neither fully sovereign nor fully aligned. But the new geopolitical context creates room for political, technological, and economic realignment.

Investing in open technologies, supporting local ecosystems, and shaping European standards is no longer idealism — it’s a strategy for resilience, economic efficiency, and global positioning.

Conclusion: A Moment to Seize

Technological decoupling, if understood correctly, can be an opportunity for Europe. It calls for a paradigm shift — literally and figuratively: moving away from passive consumption models, investing in the code economy, and building the conditions for long-term trust. Open source is not an alternative — it may well be our best option.

Anyone who knows me knows that I’m a fanboy, the fruit kind. I like robots too, but I find them not sufficiently human, pun intended.

That being said, from time to time, I can really appreciate what Google can do with its soft skills (you know that thing a few humans possess; typically not an engineer’s forte). Today, I would like to highlight a practical case where Google, with a simple email, shows much greater empathy and at least gives the illusion of caring for its developer community. Here’s a copy of the email many of us received last week with my numbered (x) comments below:

Subject: Google Play Developer Program Policy Update

Making Google Play a great community for users is an important part of helping build a successful platform for you as a developer to distribute your apps. From time to time, we update our content policies as part of an ongoing effort to provide a secure and consistent experience for users. We strive to maintain clear guidelines for you as developers, to help us grow this platform and engage with those users (1).

This email is to notify you that we’ve made some changes to our policies; here are some highlights (2):

- Streamlined the ads policy, with guidance on interstitial ad behavior, and a new “System Interference” provision, which prohibits ads in system notifications or home screen icons, and requires user consent when an app changes specified settings on a device (3)

- A revised hate speech policy that provides more comprehensive coverage, while recognizing Google Play’s role as a platform for free expression

- Clarification that the gambling policy extends to all games that offer cash or other prizes; that virtual goods and currency in games are subject to the payment processor policy, that incentives should not be provided to users to rate an app; that artificially inflating an app’s install count is prohibited, and that the Google Play Program Policy applies to all developer information or content made available on the Store.

Please review the Google Play Developer Program Policy to see all the changes and make sure your app complies with our updated policies.

Any apps or updates published after this notification (4) are immediately subject to the latest version of the Program Policy. If you find any existing apps in your catalog that don’t comply, we ask you to fix and republish the application within 30 calendar days of receiving this email. After this period, existing applications discovered to be in violation may be subject to warning or removal from Google Play.

We recognize that some developers will need to change their app and advertising practices (5) to comply with the revised policy, but we believe these changes will help ensure all users and developers can maintain confidence in the standard of apps available on Google Play. Our aim is to foster a high standard of app behavior, so you will be able to take advantage of Google Play as a successful platform to distribute your apps and continue to grow your business. (6)

Regards,
Google Play Team

Google Inc.
1600 Amphitheatre Parkway
Mountain View, CA 94043

You have received this mandatory service announcement to update you about important changes to your Google Play service or account.

First of all, we receive an email, instead of being told to just re-validate the terms typically at a time when we’re busy in our app submission workflow process. The above email update is asynchronous, and makes it much easier to process calmly.

Then, the soft touches:

(1) Restate and remind the context; policies are not there to make life difficult for developers, but rather to help grow a platform and raise the bar. Granted, Google has a lot of catching up to do here! But stating a shared goal puts me on the same team and puts my mind in a positive mental state to process the new “constraints”…

(2) Highlights! THANK YOU SO MUCH. With Apple, every new policy update makes me feel like I need to use the diff online command tool to find out what crept up and what will bite me back down the line. Here, the highlights tell me what I’m in for. 3 bullet points. Probably only one or two concern me. I can digest it directly from the source, right there. Or I can dig deeper by reading the entire updated policies, but probably don’t need to. Don’t let me guess, or worse, discover the biased version from an (over) analysed blog post which inevitably is looking for the unwritten catch (which will add unwarranted anxiety).

(3) Short list, unafraid of using the “ad” word including technicalities like interstitial (Google understands and accepts ads, fully). Here, plain catching up to Apple who’s had similar guidelines for a while. But introducing those changes is a huge deal in the Android world. That email was definitely the right thing to do.

(4) This whole paragraph is EXPLICIT, and leaves no doubt as to what is likely to happen and when. New apps, updates, and existing apps. On the fruit side, we actually never know if it will apply to new/updates and/or existing apps…

(5) Apologize, by empathetic. Nice touch. Doesn’t hurt. App publisher teams work as hard on an individual basis as any Google or Apple employee (if not more?). Thanks for recognising the work that actually makes YOUR platform stronger. Actually, since most publishers won’t turn ROI positive, then double thank them.

(6) Re-state the intent. It’s for the greater good. No shortcuts allowed. Life is hard. Suck it up. The platform rules. Publishers are guests. I knew it when I joined.

In essence, the above memo is a slam dunk announcing tough new constraints for a number of developers. But it’s a rallying call. It’s subtle in its own way, and it’s done right.

My feeling with Apple is that the developer relationship is brutal, in your face, with little to no empathy, Darwinian and unapologetic (except when there is an explicit mess, e.g. dev tools downtime). Apple does most things right, including letting us know when things change. But there is often too much room for doubt, anxiety, interpretation. It’s unnerving.

I suck it up not because I’m a fanboy; I do so because in my opinion, Apple’s end product is superior, by a large margin. Still, no excuse for not using soft skills to manage the developers expectations… could go a long way to strengthen the platform, and the entire business.

Less than 6 months ago, following my 3 years stint as Scaleway’s CEO, I set out to investigate the realm of possibles by publishing yet another manisfesto, to discover “Future Proof Inc.”

Shortly after publishing the post, I received an avalanche of introductions, suggestions for fields to investigate, actual job offers, invitations to join entrepreneurial ventures…

This resulted in an intensive 7 weeks marathon, packing my agenda with no less than 140 meetings, averaging 5 per day, half of which were in person in and around Paris, the rest by phone or video conference. This period was mindblowing, allowing me to meet incredible entrepreneurs, out of this world projects. Trust me, there is no lack of energy and opportunities out there, right here, right now, in France, and certainly elsewhere…

I was able to identify no less than 50 projects to investigate, covering a variety of topics, including AI, cloud, robotics, quantum computing, future of work, venture capital… The opportunities varied in terms of maturity, ranging from ideation, early stage, late stage, and established/mature.

Out of those 50 projects, a dozen opportunities emerged (actual offers and soft offers), leading to top 3 candidates.

Now, going back to my earlier manifesto, the following themes and concepts were on my radar :

• Technology for good, or technology advancement for mankind
• Entrepreneurship & value creation; seed, growth or hyper growth
• General management, with or without business development
• Corporate Social Responsibility: social, environmental, corporate citizenship
• Computing, cloud computing, tooling, open-source & software ecosystems
• Deeptech & moonshots
• European tech sovereignty & excellence

I am happy and proud to report that what I ended up choosing checks all of the boxes! Before the reveal, here’s how it unfolded :

It all started on January 13th 2023, when François Goupil reached out to meet with me. François and I knew each other from when he was working for HUB France AI (of which I am a cofounding board member), a think-tank created after Cedric Villani issued his first report on AI in 2018. More recently, as VP Tech & Solutions at the Paris Open-source Experience Summit, I had invited François to lead the Machine Learning & analytics track. I had the pleasure to also welcome him on a panel where we discussed “Open source as-a-service : benefits & limits”.

François is currently in charge of fostering growth of Scikit-learn at Inria, the French National Institute for Research in Digital Science and Technology. Scikit-learn is one if not the most popular machine learning open-source library and framework, possibly used by all data-scientists on the planet. It has been developed over the past decade by researchers, data- and computer-scientists, typically building the tools they needed to conduct their research. Even though the library is now mature, it remains work-in-progress, as with any piece of code, but it is alive and well thanks to a dedicated team of core-developers and maintainers.

It turns out that Inria has been supporting most of the cost to sustain a good chunk of the core team, via the Scikit-learn consortium, over a good decade. Everyone agrees that this piece of technology is in fact essential to research, to academia and to the industry. It needs to be sustained. Inria should be celebrated for having invested all of these years, for allowing this project to thrive even though it is not actual research, but rather a byproduct of research, a tool to facilitate research. Of course, this could not have happened without the relentless and visionary work from Gael Varoquaux, Olivier Grisel and Alexandre Gramfort, among other core-developers and contributors. This piece of technology is a gift to humanity, allowing seamless data exploration, getting machine insights beyond the biological limits of direct human possibilities.

The French government is very clued in right now, and it starts with our President Emmanuel Macron who has put the French Tech high in his agenda. Indeed, France has been working on fostering innovation and industrial digital transformation in general since his first mandate. Rightly so, France and Europe need to be in the race in order to be relevant for the decades to come; afterall, the foundations we are building today will be delegated to our children and the future generations. As part of France 2030 and the National Strategy for Artificial Intelligence, the government has tasked Inria to build up data-science commons with, above and beyond Scikit-learn. Other initiatives seek to finance and bolster the private tech sector, hopefully leading to more unicorns, decacorns and centaurs, with a high risk/return factor, but also with growth benefits for France and Europe at large. Here however, the goal is to level the playing field through open-sourcing by reducing core dependencies on commercial lock-in and extra-territorial technologies and services, and indirectly improve sovereignty.

François set-out to convince me to join. He barely had to twist my arm! Indeed, the mission checked all the boxes… I quickly met with Gael Varoquaux, co-founding core-contributor and guardian of the Scikit-learn philosophy all these years, and then the rest of the team. Somehow, François and Gaël managed to convince the Inria management that I was the right person for the job, or rather, for the mission. Time will tell if the casting is the right one, but I already feel humbled, honored and bolstered by the challenge.

But as always, things are never that simple. The mission requires a rocket. Today, I can only reveal the first stage of the rocket since the second stage isn’t built yet!

Without further ado, I am proud to officially announce my new role as Entrepreneur-in-Residence at Inria, which means spending a lot of time with the team in the Alan Turing building on the École Polytechnique campus.

My immediate goal is to structure a vehicle to accomplish the mission. I actually started nearly 3 months ago, and we’ve made great progress already.

If you’re interested to know more, then please contact me.

Otherwise, rendez-vous in the Fall for the next milestone and announcement ;-)

30 years ago or so at The American University of Paris, I chose to do my Bachelor of Computer Science Senior Seminar on Neural Networks. Pretty obscure topic at the time! It took months to put everything together: gather the literature from actual books, learn some of the theory, code the feed-forward and back-propagation algorithms and the interface in Objective-C on a 32-bit 8MB RAM NeXT computer, create a dataset to represent digits in 5x7 grids, annotate, feed the data into the algorithms, and wait long hours to conclude that it wasn’t working, test additional layers with various number of artificial neurons, fail, try again, repeat… until the model or rather the effort was deemed acceptable enough. The goal was to investigate and write a paper about the technology, not achieve industrial strength accuracy. (Shout out to my co-author Bassel Haidar 👋)

Today [March 29, 2023], just for the heck of it, I asked ChatGPT this: “write a python script that loads the digits dataset from sklearn and then setup a pytorch deep learning model”.

30 seconds later, I had a decent piece of code. But it failed to run with a logical error somewhere in the middle of the code.

Nice try, but frustrating! My request was reasonable and a fairly trivial thing to ask a massive LLM in 2023… (!)

I thought that perhaps my python packages were incompatible with what ChatGPT had learnt and that I now needed to take over. Typically, I would hit the documentation online and start debugging… but no. I flexed my “lazy muscle” and challenged the bot’s response to give it the feedback it deserved for sweating it out in the first place. So I wrote back: “in the above, I get an error here: loss = criterion(outputs, labels)”

ChatGPT responded: “I apologize for the error in my previous code. The issue is that the PyTorch CrossEntropyLoss function expects the target label tensor to have a shape of (batch_size, ) rather than (1, ).” with the updated code.

Barely looking at the fix, I ran the updated code on my MacBook Pro M1. 30 seconds later, I had a fitted model with 97% accuracy to recognize digits.

Great, so the AI did meet my expectations after all (wait, 97% only ?). What I did not expect was the error in the middle of the code and the casual apology with a correct fix upon request!

Multiple take-aways here:
- a lot of the DL tech has matured over the past 30 years, thanks to massive contributions by Geoffrey Hinton, Yann LeCun, Yoshua Bengio, Andrew Ng, Fei-Fei Li, their colleagues and successors
- the tech is sufficiently well packaged to be harnessed by a junior computer scientist, a hobbyist or a tinkerer able to express complex ideas in simple terms
- even though there is a human in the loop here, twice (me, with the prompt & the error fix request), there is a little bit of meta action going on: LLM deep learning inferring deep learning code without breaking a sweat…

Yup, this changes everything.
The next decades are going to be interesting.

Unless things go much faster than we are allowed to know…
https://futureoflife.org/open-letter/pause-giant-ai-experiments/

TL;DR After 3 intense years as CEO of Scaleway, I am now investigating new opportunities. To explore the realm of possible to the fullest extent, to discover the right “Future Proof Inc.”, I hereby invite you, my benevolent network and readers, to participate by sharing 3 ideas with me: people to meet, job opportunities you may be aware of, topics to research, entrepreneurial ideas... Indeed, I truly believe in the wisdom of (selected) crowds, and you are all collectively clued in in ways that I am not, in a market that is perfectly imperfect. Also, some of you know me more than I know myself, so I should listen! Either way, if you are interested to participate in this live crowdsourcing experiment, please read below for deeper insights, or go wild and hit submit.

Every once in a while, we reach the end of a cycle in our professional careers, as entrepreneurs or as employees.

If I look back over the past two decades or so, I have been mostly an entrepreneur, iterating between each venture by exploring new models: with and without venture capital funding, with and without co-founders, or joining as a late founder of sorts, mostly acting as CTO, COO, or CEO. All of those ventures went full cycle, i.e. they were acquired, all six of them. Five of them were tech product companies, including one in deep-tech. One was a mobile app development agency. All were B2B companies including two that attempted a B2C model at first. In every case, I left quite quickly post-acquisition to embrace a new challenge.

My last gig was nothing like the previous ones. In February 2020, I became Scaleway’s CEO, an existing but below the radar cloud company based in Paris, founded by entrepreneur-extraordinaire Xavier Niel some twenty years ago. Xavier had been an angel investor in my previous two startups and asked me to help develop Scaleway. This time, I joined as a “regular” employee within the Iliad Group, Scaleway being one of its many subsidiaries. I will not elaborate here on my journey at Scaleway, but let me tell you: what a ride! Three intensive years that started with a global pandemic and lock-down. Not to be deterred, I "hyperscaled" the company threefold to 600+ Scalers, fixed the general ambiance, values and organization, redefined the core target segment, accelerated a market-beating double-digit growth 3 years in a row in the Public Cloud, and ultimately positioned Scaleway as a credible regional alternative to the major 3 cloud hyperscalers. And lobbying, a lot of it! Today, Scaleway is stronger than ever, ready to face its next growth challenge. The news has been covered here, here and here.

And so I find myself at a crossroad, again. Which is massively exciting given the perspectives: so many problems needing solutions (and by default, I’m biased towards technology), so many solutions looking for problems (bad idea in general, except for deeptech). And yet, there lies the conundrum: where should I really invest my time, energy and passion, and with whom? This is a universal question that I dare anyone to answer for themselves. It is a recurring question in one’s career. As an entrepreneur or as an employee.

So I need your help, I welcome it with open arms. I already resorted to this technique 7 years ago which had led me to be headhunted for the position of GM Twitter France (for some reasons, I still like Twitter very much), but instead joined Snips.ai as COO (Snips was acquired by Sonos in November 2019). I have a number of meetings lined up starting Monday January 2 2023, but I'm anxious to deepen the breadth and depth of the possibilities. In a nutshell and to be clear: I’m not looking for a job. I’m just impatiently looking forward to discover the one mission to embark on, to meet my next professional challenge.

This form will allow you to pitch in, up to 3 ideas. It takes no time to fill-in, but you might want to revisit my LinkedIn page as well as wrap your head around the following unsorted list that captures some of the key concepts I'm interested in going forward:

• Technology for good, or technology advancement for mankind
• Entrepreneurship & value creation; seed, growth or hyper growth
• General management, with or without business development
• Corporate Social Responsibility: social, environmental, corporate citizenship
• Computing, cloud computing, tooling, open-source & software ecosystems
• Deeptech & moonshots
• European tech sovereignty & excellence

So, what do you think? Any ideas come to mind?

Either way, thank you for reading this far. You'll hear from me soon!

Following recent EU-U.S. initiatives announced by president Joe Biden dealing with tech and transatlantic trade, the status of these two parties can be called into question. Are they friends, or are they foes? Moreover, which will prevail: domination, or ‘coopetition’? When it comes to technology in particular, it is not a zero-sum game, and both sides stand to lose to Xi’s China unless a joint vision is adopted.

The U.S. that we know today is very different from the U.S. that attracted people from around the world as a new country full of vitality and promise. The words “liberty”, “entrepreneurial freedom”, and “equality” may sound familiar to us now, but they were what inspired Europeans especially to leave everything behind. Men and women crossed the ocean knowing they would be part of a new era — they realized that they had the power to change their own lives.

Two hundred and forty-five years later, the U.S. is not a new country anymore. Over the centuries, it has taken on the role of a superpower. Today’s digital era was built there, yet the very same values that were once sacred are now at risk. The indisputable technological leadership of the U.S. is now putting its historical partners in an intolerable position of subordination. The renewal of the concepts of “independence”, “strategic autonomy”, and “freedom” may now be coming from old Europe.

To gain its independence, the U.S. had a revolution. Post-war Europe, on the other hand, resorted to regulation.

Europe needs to answer pragmatically, while keeping in mind the ideals at stake here. Adding a new law to all those already in existence will only lead to confusion. We have to stay focused on a humanist vision of what progress is and always should be.

Technological progress cannot exist without social and political progress, and entrepreneurial freedom cannot exist without democracy. If fair trade is not protected, there will be no market at all. It is time to end the legal loopholes that new Information and Communication Technologies have benefited from until now.

By voting for the General Data Protection Regulation (GDPR), the Digital Markets Act (DMA), and the Digital Services Act (DSA), Europe sought to point out that it was time to renew privacy standards, readjust the rules of competition, and give power back to its citizens. The intention was never to descend into hostilities. U.S. lawmakers will surely be swift to follow suit.

Until now, American legislators never let the private sector endanger state prerogatives. The Tech Antitrust Reforms recently pushed by a bipartisan coalition prove that American Congress members still have the will to protect free competition, even if it took them time for the new technologies sector. Day by day, the growth of digital giants increases the challenge of writing legislation that addresses our underlying concerns about the power they hold. Those are heated debates, but the need for rules has become obvious. How else could it work? Antitrust regulation, and the abolition of all monopolies, are necessary to create an environment where competition is synonymous with fair-play, innovation, and wealth creation. No-one is more aware of this than American economic and legislative stakeholders. We Europeans share this point of view, and we are acting accordingly to regain the status of equal partners, and technological allies with the same ethical priorities.

The recent American-European summit and the joint EU-U.S. statement about a renewed transatlantic partnership put us — officially — on this path.

Undeniably, some rules have to be updated to address the multifaceted challenges ahead of us, ahead of humankind. More importantly however, we need to act together and promote shared ambitious objectives in this crucial time. At the same time, on both sides of the Atlantic, the defenders of state and technological sovereignty will be watching and making sure that future legislation is based on transparent and reliable information, guaranteeing digital free will to users and protecting the common good.

The creation of an EU-U.S. Trade and Technology Council whose mission is to “remove trade barriers, set global standards, and promote joint innovation in key technologies” can be seen as an encouraging step.

This may be our last chance to build a balanced and fair digital society and market. We must act now, and we must act quickly. We still have the opportunity to choose the partners we want to work with. In this globalized world, where the next technological breakthroughs may give definitive competitive advantage and absolute ideological supremacy, there is no time for Americans and Europeans to quarrel, especially with Xi’s China abiding only by its own rules. We need to remember that we are historic partners, united by a common past and shared humanist values, and to act accordingly in this paradigm-shifting world order.

1. https://www.platformer.news/p/the-antitrust-bills-get-their-hearing
2. https://www.consilium.europa.eu/fr/press/press-releases/2021/06/15/eu-us-summit-statement-towards-a-renewed-transatlantic-partnership/
3. https://www.cnbc.com/2021/06/15/america-and-eu-will-join-forces-to-craft-new-rules-on-trade-and-tech.html

Independence is a two-way street, or why the USA needs a strong technological Europe was originally published in Scaleway on Medium, where people are continuing the conversation by highlighting and responding to this story.