Value capture and appropriability strategy, the case of OpenAIs ChatGPT.

Av /
What happens when Open AI tighten their control launching the webbrowser Atlas, will the World Wide Web become proprietary Infrastructure?
A few thoughts of how OpenAI’s new browser is changing the game for data, competition, and the open web, based on my student essay in AI and Data Strategy at Halmstad University, June 2025.

I now. Its not great to be the one to raise concerns. Being the person in between is not easy. They try to understand different perspectives. They often get smashed between sides. However, it is possible to be positive and see opportunities. One can also dare to look at the risky assumptions. Assess what this will be like, for both the good and the bad. So, I took my essay from this spring out. When I wrote the piece, I got involved in the new area for me, of AI and data strategy. It became clear. As I described in an earlier post of 2024, OpenAI tries to be the layer between original content and user. Different values affect the strategies.

From Open Research to Closed Ecosystem

When OpenAI was founded in 2015, the vision was clear: AI should benefit all of humanity. The organisation operated as a non-profit research entity with values centred on open access to research results and data sharing. But the journey from that vision to the launch of the ChatGPT Atlas browser on 21 October 2025 tells an entirely different story—one about how value capture and control over data have systematically replaced the ideals of openness.

Thanks for reading Ellen’s Substack! Subscribe for free to receive new posts and support my work.

Atlas: The Next Step Towards Total Control

ChatGPT Atlas is an AI-powered browser with ChatGPT built into its core. It features a sidebar where users can chat with their search results, summarise pages, and let the AI perform tasks on their behalf through “agent mode”. The browser can also remember context from the pages you visit and use that information to make responses more personalised.

Thanks for reading EX! Subscribe for free to receive new posts and support my work.

For the user, it sounds convenient. But beneath the surface, something much larger is happening: OpenAI is building itself in as an indispensable layer between people and the internet.

Data as Strategic Resource

OpenAI’s success with ChatGPT is built on access to enormous amounts of data. By scraping the internet, they have trained their models on text, code, and images from social media, blogs, digitised books, reviews, and Wikipedia, often without the data sources even knowing about it or giving consent.

When the New York Times and other media organisations discovered their content had been used without permission, lawsuits for copyright infringement followed. OpenAI defended itself by claiming they had trained their models on “publicly available” content under the principle of “fair use”. But when the company itself began profiting from this free data harvest, several major players such as GitHub and the New York Times suddenly blocked OpenAI from accessing their content.

From Open Web to Controlled Infrastructure

Tim Berners-Lee created the web as an open, neutral infrastructure for sharing information globally. The idea was to democratise access to knowledge. But with Atlas, OpenAI is building a new layer on top of the free web—a layer that curates and commercialises access to information.

Atlas directly challenges Google Chrome, which has long dominated the browser market, by integrating AI assistance throughout the browsing experience. But the difference is crucial: whilst traditional browsers are relatively neutral tools for navigating the web, Atlas is built for OpenAI to control how information is consumed and generated.

Strategic Partnerships That Strengthen Control

OpenAI has developed a sophisticated strategy for value capture through various types of partnerships:

The Microsoft partnership illustrates how OpenAI secures both capital and infrastructure. In January 2023, Microsoft deepened its commitment through a multiyear, multibillion-dollar investment. Recently, ahead of OpenAI’s planned IPO, Microsoft negotiated guarantees for continued access to OpenAI’s future models. Control over data and model access tightens as AI training becomes increasingly capital-intensive.

Media companies and data partnerships reveal another dimension. OpenAI has entered into exclusive licensing agreements with players such as Axel Springer and News Corp. Reports suggest that Axel Springer’s content is even prioritised in ChatGPT’s search results to drive traffic and subscriptions to their brands. This shifts the web from decentralised and open to curated and commercialised.

The Apple integration from 2024 involves deep integration of ChatGPT into iOS, iPadOS, and macOS. Through such partnerships, OpenAI secures control over both data inputs and the infrastructure required to deliver AI services—not merely economic benefits.

The Information Paradox: When Media Companies Dig Their Own Graves

Here, a deeply problematic dynamic emerges. When media companies licence their content to OpenAI, they certainly receive short-term revenue, but risk undermining the long-term value of their own content.

Data is non-rivalrous—it can be used by multiple actors simultaneously without losing value. But when OpenAI’s models are trained on media content, AI-generated content is created that can replicate their style and authority. Media companies are thus trading away their “hard-earned credibility for a little cash from companies that are simultaneously undervaluing them and building products quite clearly intended to replace them”, as critics put it.

Within science, data sharing has traditionally been guided by principles of transparency and collective benefit—researchers share data with clear attribution and traceability. But in the AI ecosystem, media companies’ data is absorbed into opaque, proprietary “black boxes” where contributors have no visibility into how their data is used, no guarantee of attribution, and no ability to evaluate the learning process.

Atlas and Control Over the Internet

With Atlas, users can take ChatGPT everywhere on the web, making the AI a constant companion that sees what the user does. The browser includes a “memory” function that learns more about the user over time to personalise the experience.

But this convenience comes at a price. OpenAI positions itself as a central hub in the AI ecosystem and powers a growing number of applications through its APIs—from Microsoft 365 Copilot to third-party startups that pay for access to OpenAI’s models. By controlling access to both data and models, OpenAI strengthens its grip on the entire chain.

This creates what is called a “kill zone” problem: startups that lack access to comparable data or computing power simply cannot compete. Control is exercised through contracts and infrastructure control.

From Pre-Paradigmatic to Paradigmatic Phase

Initially, ChatGPT existed in a pre-paradigmatic phase where user needs were closely coupled to the market and influenced design. Competition focused on setting the dominant design. But now ChatGPT, with competition from Anthropic’s Claude, Google’s Gemini, and Perplexity, has moved into a paradigmatic phase where volumes increase and opportunities for economies of scale open up.

OpenAI continues to seek strategic partnerships to strengthen control over key resources. Atlas was announced by OpenAI’s CEO Sam Altman as “a rare, once-a-decade opportunity to rethink what a browser can be”. This is not merely a technical innovation—it is a strategic move to control how people access information in the future.

What’s at Stake?

Having once worked at the Swedish Internet Foundation, where “Internet for all” was the accessible goal, this makes me reflect on what it means for infrastructure when a potential new AI layer is created on top of the internet.

OpenAI captures value not only by offering access to vast amounts of data, but by providing its AI models as a service. By positioning itself as a central hub and an underlying infrastructure layer for other companies’ innovation, OpenAI is simultaneously moving away from the principles of an open web and building a more controlled and closed AI ecosystem.

The Risks Are Growing

For businesses building on ChatGPT’s output, value creation becomes increasingly fragile—they rely on AI-generated content whose origins, biases, and limitations remain opaque.

For end users, this opacity raises critical questions about security, fairness, and wellbeing. Without transparency in how answers are generated, it becomes harder to ensure that AI services are trustworthy and equitable.

For the web as a whole, the original democratic vision of open and accessible knowledge is at stake. If key knowledge infrastructures become closed and proprietary, the web’s democratic potential risks being undermined.

What’s Required Now?

As AI ecosystem strategies continue to evolve, policymakers and stakeholders must confront these risks. A key challenge going forward will be to ensure that data governance and AI regulation frameworks promote not only innovation and economic value, but also openness, transparency, fairness, and public trust.

The launch of Atlas is not just a new product; it is a clear signal of where OpenAI is heading. The question is whether we as a society want to go there with them, or whether we need to act now to preserve a more open digital future. What do you think and from which perspective and zoom level?

This analysis is based on an essay written during the course AI and Data Strategy at Halmstad University, 5 credits in June 2025. In this adapted blogpost, with the addition of the 22 October 2025 current reporting on the Atlas launch. The student is essay was so to say, AI-reworked with the help of Claude, using the Prompt 1: Kan du skriva en bloggpost för substack på enkel tydlig svenska av innehållet i texten. Koppla till släppet av Open AIs nya browser, Atlas https://openai.com/index/introducing-chatgpt-atlas/?utm_source=chatgpt.com (+ student essay). Then an additional prompt 2: Kan du skriva samma version, på brittisk engelska? The image is made in Midjourney.

FYI: There may be errors in the text. I want to post it now, since most of my thinking ends up in the “drawer“ at home and this questions is important now, due to the realise of OpenAI’s webrowser Atlas.

I do not have a set, clear opinion myself yet. I do believe we need to research, understand, discuss and explore options to make wiser decisions. By keeping your mind open and curious, you will learn and develop more. / Ellen

References

Added 22 October 2025 for this AI-generated blogpost https://openai.com/index/introducing-chatgpt-atlas/?utm_source=chatgpt.com

Lectures

  • Long, V. (2025). Lectures during the course AI and Data Strategy. Halmstad University.

Course literature

  • Mayer-Schönberger, V. & Ramge, T. (2022). Schumpeter’s Nightmare. In: Access Rules: Freeing Data from Big Tech for a Better Future. University of California Press. Available at: https://www.jstor.org/stable/j.ctv2kx88cp.5
  • Teece, D. J. (1986). Profiting from technological innovation: Implications for integration, collaboration, licensing and public policy. Research Policy.

Internet reports

Internet articles

Rulla till toppen