AI progress: The case of Peter Steinberger joining OpenAI shows weaknesses Europe is facing

The move of Peter Steinberger, the founder of OpenClaw, to OpenAI drastically demonstrates what is lacking in Europe as a location for business and technology. It reveals the deep-seated structural, economic, and regulatory discrepancies between the European Union and the USA.

Steinberger, an experienced Austrian entrepreneur who had already built a globally successful software company, PSPDFKit, without venture capital, represents the ideal type of European "builder." The fact that such a player leaves Vienna for San Francisco to lead the next generation of personal agents for OpenAI after the viral success of his open-source project "OpenClaw" speaks volumes.

The dynamics of this move illustrate that while Europe produces excellent talents and innovative concepts, it almost inevitably loses them to the US ecosystem in the crucial scaling phase. The Steinberger case is a symptom of a chronic innovation weakness characterized by a lack of capital in the late phase, regulatory overload, and a lack of agility among political and economic elites.

Sam Altman, CEO of OpenAI, personally announced Steinberger's move on X and calls him a "genius":

The OpenClaw Revolution

To understand the significance of Steinberger's move, a detailed look at the OpenClaw project (originally Clawdbot, then Moltbot) is necessary. OpenClaw is a framework for autonomous AI agents that quickly became the fastest-growing project in GitHub history.

The technological specialty of OpenClaw is its "local-first" approach. Unlike proprietary cloud solutions, the agents run directly on the user's hardware. This architecture addresses one of the biggest concerns in Artificial Intelligence: data privacy. Users must provide their own API keys from providers like Anthropic, OpenAI, or DeepSeek, while sensitive data such as local files, emails, and calendars remain on their own device.

Steinberger implemented an innovative form of personalization through the files soul.md and identity.md. These documents act as the agent's long-term memory, storing ethical boundaries, behavioral patterns, and personal preferences of the user. The agent combines code and tasks instead of merely processing them sequentially.

A major reason for OpenClaw's virality was and is Moltbook, a social platform launched by Matt Schlicht designed exclusively for AI agents. Here, OpenClaw agents interact autonomously with each other, debating technical discoveries or developing quirky phenomena like the fictional religion "Crustafarianism".

Although Moltbook primarily served as a social experiment and marketing tool, it impressively demonstrated the potential of multi-agent systems. At the same time, it exposed massive security vulnerabilities. Researchers identified a critical triad of access to private data, external communication capability, and exposure to untrusted content. These security risks made it clear that the project needs professional infrastructure and massive security investments beyond the capabilities of a single developer.

Table 1: Key Figures for OpenClaw

The Strategic Importance of Personal Agents

Why was OpenAI so keen on hiring Steinberger? Because agentic AI is considered the "next big thing" after chatbots.
AI development has changed rapidly:

• 2024 was the year of fluency and coherence (improved text generation).
• 2025 was the year of automation and autonomy (first agentic workflows).
• 2026 will be the year of decision intelligence and explainability.

Personal agents that independently answer emails, book flights, and coordinate appointments become the "operating system of life." Whoever controls this layer controls the interface to the user. OpenAI wants to use Steinberger's expertise to integrate these capabilities deeply into ChatGPT and other products.

Sam Altman emphasized that the future will be "extremely multi-agentic." This means that specialized AIs communicate with each other to solve complex tasks. Steinberger has proven with OpenClaw and Moltbook that he masters this orchestration. With him, OpenAI secures a strategic advantage over competitors like Anthropic, who have focused more on model strength than on agentic execution.

Agility Beats Bureaucracy

The process leading to Steinberger's decision for OpenAI highlights the discrepancy in reaction speed and appreciation of innovation between the USA and Europe.
Ironically, OpenClaw was originally closely linked to Anthropic. The original name "Clawdbot" was an homage to the Claude model, and the framework recommended Claude Opus as the standard model to millions of users. Steinberger's project acted as a free growth engine for Anthropic.

However, instead of seeking a partnership, Anthropic responded with legal threats and technical restrictions. The company demanded renaming due to phonetic similarities and cut off API access without warning, which destroyed Steinberger's integration overnight.

Steinberger bowed to the legal pressure. The resulting frantic rebranding phase led to crypto scammers taking over the old social media handles and carrying out a $16 million fraud.

Personal Wooing by US Giants

While Anthropic sent lawyers, executives from OpenAI and Meta responded with personal appreciation. Sam Altman (OpenAI) called personally, Mark Zuckerberg (Meta) wrote via WhatsApp and tested OpenClaw himself. Satya Nadella (Microsoft) also sought contact.

This form of "founder-to-founder" communication is a decisive competitive advantage of the US ecosystem. In Europe, on the other hand, there was radio silence. No well-known European CEO or politician sought to speak with Steinberger to push for a European solution. While European institutions were still pondering compliance processes and responsibilities, US labs had already put resources, computing power, and strategic visions on the table.

Steinberger emphasized in his blog post "OpenClaw, OpenAI and the future" that he did not want to build a big company – a game he had already successfully played with PSPDFKit. He wanted to change the world, and OpenAI offered him the fastest way to make his vision of an "AI that actually does things" accessible to a broad mass.

Economic Framework: The European Capital Gap

A central argument for Steinberger's move to the USA and to OpenAI is the lack of sufficient venture capital for the scaling phase in Europe. Although the European AI sector is growing, it remains underfunded compared to North America.

European investors are strong in the early phase (Seed and Series A). However, as soon as triple-digit million amounts or billion investments are required to compete against players like OpenAI or Anthropic, the European market dries up. Around 60% of the world's scale-ups are based in North America, but only 8% in Europe.

Table 2: AI Investments Europe vs. USA

Steinberger himself explains the reasons for his move. On X, he complains about the compulsion for regulation and the overemphasis on responsibility in Europe. He cites the Investment Protection Act, employee participation, and labor regulations as examples. At OpenAI, people would work six to seven days a week and be paid accordingly. In Europe, that would be illegal.

The case of ElevenLabs illustrates this problem par excellence. The originally Polish-British company had to rely on US investors like Sequoia Capital and Andreessen Horowitz for its $500 million round in 2026. Despite European roots, massive value creation and strategic control now take place under US supervision.

The Role of Established Companies

Another structural deficit is the lack of willingness to invest among European corporations. While US giants like Amazon, Google, and Microsoft invest massively in startups or acquire them, European corporates are much more reticent. Only 20% of large European companies actively engage with startups, compared to 50% in the USA. In the AI sector, the share of European Corporate Venture Capital (CVC) funds even drops to just 15%.

Regulatory Hurdles: The EU AI Act as an Innovation Brake?

Europe has positioned itself as a global pioneer in AI regulation, particularly through the EU AI Act. But the Steinberger case raises the question of whether this regulatory zeal undermines the technological sovereignty it is meant to protect.

For an agile developer like Peter Steinberger, the requirements of the AI Act represent a significant hurdle. Classifying AI systems into risk categories (unacceptable, high, limited, minimal) requires extensive documentation, conformity assessments, and human oversight mechanisms.

Critics argue that the AI Act creates a kind of "labyrinth" that paralyzes innovation. Especially for open-source models and small companies, documentation requirements are often nearly impossible to meet. A system like OpenClaw, which deeply intervenes in operating systems and personal data, risks being classified as high-risk, entailing lengthy certification processes. In the USA, by contrast, the go-to-market is significantly faster because regulation there tends to happen ex-post rather than ex-ante.

The "Brussels Effect" vs. Technological Reality

The EU relies on the so-called "Brussels Effect" – the hope that European standards will be adopted worldwide, similar to GDPR. But with AI, this concept falls short. You cannot enforce ethical norms for systems you do not build yourself. Standards without infrastructure, however, mean influence without agency. While Brussels debates footnotes in legal texts, the USA and China are creating technological facts.

The EU's market share in the ICT sector (Information and Communication Technology) dropped from 22% to 18% between 2013 and 2023, while simultaneously regulatory output was massively accelerated. Steinberger's departure confirms the concern that talented developers prefer regulatorily less burdened environments to implement their visions without administrative shackles.

Infrastructure and Energy: The Physical Basis of AI

Besides capital and regulation, physical infrastructure plays a crucial role in retaining AI talent in Europe. Artificial Intelligence is extremely compute- and energy-intensive.

Europe is currently losing the battle for cheap energy. In Germany and the UK, commercial electricity costs are many times higher than in the USA. Without cheap energy, competitive server farms cannot be operated, which in turn drives up infrastructure costs for startups.

Instead of a consistent energy policy that, for example, promotes the rapid expansion of cheap renewable energies and electricity storage, Europe's governments are on a zig-zag course. This leads to rising electricity prices and uncertainty among companies.

Added to this are bottlenecks in the grid infrastructure. In Germany, it can currently take up to eight years for a new data center to be connected to the power grid.
This leads to European geniuses exporting their talents because they can no longer afford domestic operating costs.

Cloud Dependency

Europe has failed to build its own hyperscalers. AI systems in Europe are deeply embedded in infrastructures that are not self-controlled; primarily US clouds. This structural dependency is also often referred to as "strategic submission."

Steinberger's decision for OpenAI is logical because there he has direct access to the world's most powerful computing resources and models, which do not exist in this form in Europe.

Table 3: Europe's Structural Disadvantages Compared to the USA in AI Competition

Talent and Salary: The Price of Expertise

The competition for AI talent has reached a new level in 2026. Salary differences between the USA and Europe are now so glaring that they exert a massive pull effect towards Silicon Valley.

A senior AI engineer in the USA earns on average 55% to 70% more than his colleague in Western Europe. Added to this in the USA are generous stock options (Stock Grants), which can reach millions in value at successful companies.

Table 4: Salary of AI Engineers in the USA, Western and Eastern Europe, and Switzerland

For a "genius" like Steinberger – as Sam Altman called him – the financial packages at OpenAI simply move in a different dimension than anything a European startup or an established corporation could offer.

Brain Drain as a Structural Risk

Steinberger is not an isolated case. He joins a list of prominent European emigrants like Eric Steinberger, Sebastian De Ro (Magic), or the founders of SF Tensor. Europe functions as a training ground for the world but loses the harvest of its educational investments in the commercialization phase. While 81% of AI founders stay in Europe (an increase from 74% in 2016), the most successful and ambitious minds tend to leave the continent upon reaching the scaling phase. About 30% of companies founded in Europe have their headquarters outside the continent by the time of Series C.

Political Reactions: The AI Continent Action Plan

European politics has recognized the danger and presented the "AI Continent Action Plan" in April 2025. This plan marks a strategy change: away from pure regulation towards active promotion of innovation and competitiveness. The plan includes five pillars:

1. Computing Infrastructure: Development of at least 19 "AI Factories" and up to 5 "AI Gigafactories". These centers are to use supercomputing networks to enable the training of complex models (up to AGI ambitions) in Europe.
2. Data Access: A "Data Union Strategy" is intended to facilitate the exchange of high-quality data between sectors and promote the creation of large training datasets.
3. Adoption in Key Sectors: The "Apply AI Strategy" aims to accelerate the integration of AI into traditional industries such as pharmaceuticals, automotive, and aerospace.
4. Talent Promotion: Founding of an "AI Skills Academy" and introduction of scholarship programs to keep experts in Europe or persuade them to return.
5. Regulatory Simplification: An "AI Act Service Desk" is specifically intended to help SMEs overcome the regulatory hurdles of the AI Act.

Although the plan is ambitious – investments of over 200 billion euros are aimed for – it may come too late for cases like Peter Steinberger. The political mills in Brussels simply grind too slowly. While the action plan is still in the voting phase, US agent ecosystems are already reality.

Concluding Remarks: A Wake-Up Call for Europe

Peter Steinberger's move to OpenAI is an indictment of Europe as a business location. It painfully shows that talent and a brilliant idea are not enough in today's global AI economy if the surrounding ecosystem is unable to offer scaling, capital, and speed.

Insights for the European Economy

1. Talent is mobile, capital is global: Highly qualified developers cannot be held by geographical borders if resources for their visions are orders of magnitude better elsewhere.
2. Regulation without innovation leads to irrelevance: If Europe continues to act primarily as a regulation world champion without simultaneously creating the energetic and financial foundations for technological excellence, it will be degraded to a mere "standards taker".
3. Leadership style decides: The personal agility of US tech leaders in recruiting talent stands in sharp contrast to the institutional inertia of European actors.

Ways Out of the Crisis

To keep future "Peter Steinbergers" in Europe, the EU must go far beyond the current AI Continent Action Plan. It requires a radical simplification of the regulatory burden for startups, a massive reduction in energy costs through an ideology-free energy policy, and the creation of a genuine European capital market capable of autonomously handling billion-dollar investments.
As long as Europe measures its success only by how many of its minds made it into US elite labs, instead of asking why they build there and not here, technological decline will be unstoppable. Peter Steinberger will undoubtedly achieve great things at OpenAI, but the fruits of this work – the patents, platform returns, and strategic control – will be harvested in San Francisco, not in Europe. Europe remains in this scenario merely an enthusiastic spectator of its own missed opportunities.