Is AI working for you (or are you working for it)?
Interoperability: the word that could save the internet
You are reading this email thanks to the “secret sauce” of the internet: interoperability.
This is the first Project Liberty newsletter published on our new Substack. It arrives in your inbox because open protocols like SMTP allow emails to be delivered between different email providers.
Those open protocols were part of the early internet’s architecture. Over the past few decades, open protocols gave way to private platforms, which built powerful lock-in dynamics. Network effects were captured and commodified by platforms, concentrating power, making user retention the central business model, and turning people and their data into the product.
We are now at an inflection point. Will AI systems deepen that concentration of power, or open the door to something better?
In this newsletter, we consider the relationship between interoperability and digital sovereignty, and we explore why, in the age of AI, sovereignty—which is the ability to control AI rather than be controlled by it—matters more than ever.
// The rise and fall of the open protocol
Interoperability existed well before the internet.
Power grids share electricity across separate utility networks.
Financial systems enable transfers between separate banking institutions.
Phones communicate across different service plans and carriers.
Standardized railroad gauges allow trains to run on tracks regardless of who built them or where they’re located.
We’ve explored interoperability in the past. Foundational protocols such as TCP/IP, HTTP, and SMTP created a shared internet infrastructure that anyone could build upon. This is why Mozilla, the nonprofit behind Firefox, called interoperability “the internet’s secret sauce” in 2019.
Braxton Woodham, President of Project Liberty Labs, explained the internet’s open protocols in these terms: “The development of protocols was so rich in the late ‘80s through the early ‘90s, and then everything became privatized, and we stopped developing these core protocols.”
Web 2.0 ushered in privatized social platforms. People supplied the data, but the platforms controlled how it was stored, used, and monetized.
Now, as AI systems ingest vast amounts of data about our lives, businesses, finances, and health, we are entering a new level of lock-in. It’s less like a walled garden, and more like a relationship that’s hard to leave—the longer you stay, the more it knows, and the more you’d lose by going.
// Import your memory
And yet, there are workarounds for those who want to assert their sovereignty.
With social media platforms, accessing and downloading your own data has never been straightforward. Most platforms do offer a way to download it—see the steps required on TikTok here and on X here—but the process isn’t straightforward. And even when you manage to retrieve an activity log of every tweet you’ve liked or list you follow, a bigger question looms: then what?
AI works differently. Consider Anthropic, which recently promoted its memory import feature that lets people bring digital context from other AI platforms to Claude “without starting over.”
Described as “Your AI should know you from day one,” the feature lets individuals prompt their current AI to export everything it knows about them—preferences, projects, personal details—and then paste it directly into Claude’s memory settings. (Your data will remain on the previous AI platform, unless you delete it.) Claude reached #1 in the app store following the debut of the import feature and controversy over the company’s and OpenAI’s government contracts.
AI platforms can create deeper lock-in than anything before them as their value compounds with every conversation. But the same intelligence that absorbed everything can be asked to give it all back—making portability more possible, and more accessible than social media ever allowed.
// Interoperability → Sovereignty
Why would a person want to leave an AI platform? They might choose to leave for the same reasons people choose to leave a social media site: They’re tired of the algorithm or find themselves dependent on technology; they disagree with the company’s decisions or politics; or they’re seeking more control over their digital experience.
When users can move freely between platforms, and when platforms can interoperate, power shifts from Big Tech to everyday people. Interoperability drives sovereignty.
But have individuals actually reclaimed their power and exercised their sovereignty if they’re just moving from OpenAI to Anthropic or Anthropic to Google? Today’s most advanced frontier models are owned by a small handful of Big Tech companies, posing few alternatives for individuals.
And at the level of nation-states, countries are running AI systems they neither built nor control. A recent Tech Policy Press piece made the case that for “middle power countries,” digital sovereignty “does not depend on replicating frontier model development. It depends on ensuring that AI systems can be integrated, governed, audited, and, if necessary, replaced on national terms.”
// Building the digital sovereignty stack
For those seeking sovereignty over AI systems, the responses range from individual habits to national policy.
Individuals can exercise agency now
The most immediate form of AI sovereignty is practiced, not legislated. Here are a few steps that individuals can take:
Move between platforms. Explore tools that talk to each other rather than locking you in. Use AI’s memory prompt above to see exactly what AI has learned about you.
Know your data. Many platforms let you download your data. Do it, even just to understand what data they have collected about you.
Practice cognitive sovereignty. Use AI in ways that extend your thinking, not replace it.
Technologists can build an alternative AI tech stack
There is a growing movement of technologists and civil society organizations interested in building an alternative tech stack that affirms that humans, not AI agents or LLMs, are the owners of their data. This alternative AI tech stack will likely be built on core principles, which include:
Privacy and user control over digital identity and data.
Transparency, using open source software and explainable AI for core functionality.
Clear boundaries between intelligence and data where AI operators cannot hold data hostage.
Decentralization of foundational services that ensure no single actor can exert rent-seeking pressure.
Equitable value exchange for creators and consumers of data.
Governments can pass laws that center agency and interoperability
Legislative progress is happening at the state and international levels:
With support from Project Liberty, Utah passed the Digital Choice Act in 2025, establishing landmark portability and interoperability requirements for social media. Versions of the bill introduced this year in California and Virginia would extend those same principles to AI.
Bipartisan U.S. Senators Mark Warner and Josh Hawley have released federal legislation to require portability, interoperability, and the ability to delegate those decisions to third parties, which Senator Warner argues is even more important with AI agents.
The EU’s Digital Markets Act includes a similar requirement of platform interoperability. For Meta, that includes interoperability obligations for WhatsApp and Messenger, though implementation is still being worked out and does not yet amount to seamless compatibility across every product.
Policymakers can assess their digital sovereignty
Last year, Project Liberty Institute released its Digital Sovereignty Toolkit, a framework for policymakers designing, scoping, and governing digital infrastructure to advance data agency. And earlier this year, we interviewed government officials in Brazil, Cambodia, Canada, France, Germany, Greece, Indonesia, Japan, Kenya, Mexico, Senegal, South Africa, and Switzerland to understand how middle power nations can build digital infrastructure that heightens national sovereignty.
The key takeaways from the February 2026 briefing:
Sovereignty today is less about isolation and more about optionality—the ability to switch providers, retain skills, and preserve continuity under geopolitical fragmentation.
Well-designed digital systems can expand citizens’ agency while being economically viable and geopolitically resilient, making governance the primary consideration even when the language sounds technical.
// The stakes
The open internet was built on the radical bet that shared protocols could distribute power broadly enough that no single entity could capture it. For a while, that bet paid off. Then platforms arrived and began consolidating the power that had been dispersed. Now AI is doing it faster, at greater depth, with less friction—and with far more intimate knowledge of who we are.
AI systems that learn from billions of people, optimized by a handful of companies, are curating human cognition, compressing it, and then feeding it back to us. That’s a new kind of power, and it’s being built into infrastructure most people will never see or question.
The next time you use an AI system, ask one question: Is this working for me, or am I working for it? That instinct (to notice, to question, to expect better) is where sovereignty starts.
Project Liberty in the news
// Project Liberty Institute partnered with Aapti Institute to launch an initiative that will explore opportunities for data cooperatives in South East Asia. It will strengthen the capacity of community-focused practitioners to understand and initiate data-cooperative models that enhance community data agency. Learn more here.
📰 Other notable headlines
// 🌐 An article in 404 Media reported that AI translations are adding ‘hallucinations’ to Wikipedia articles. (Free).
// 🇮🇷 The future of Iran’s internet is more uncertain than ever, according to an article in WIRED. (Paywall).
// 📱 Online harassment is entering its AI era, according to an article in MIT Technology Review. (Paywall).
// 🤖 AI is turbocharging the war in Iran. Intelligence, targeting, and damage assessments are accelerating, according to an article in The Wall Street Journal. (Paywall).
// 🏛 Meta’s Oversight Board is racing to govern the AI surge. As generative AI explodes on social media, the board’s slow, human-led review model faces a breaking point, according to an article in Rest of World. (Free).
// 🇮🇩 Indonesia will ban social media for children under 16, according to an article in The Guardian. (Free).
// 🤔 The squabble between America’s government and Anthropic makes an AI disaster more likely, according to an article in The Economist. (Paywall).
// 💵 A leading AI industry-backed Super PAC is targeting a pro-regulation candidate—and building a strategy that other politicians will have to contend with, according to an article in Politico. (Free).
Partner news
// A global call to center humanity in technology
The Pro-Human AI Declaration (led by the Future of Life Institute) invites organizations and individuals to publicly affirm a shared commitment to protecting human dignity, rights, and well-being in the digital age. By signing the statement, participants signal support for a people-first approach to technology, prioritizing safety, accountability, and societal impact in AI and digital systems.
// The Childhood Index ranks all 50 states on kids and tech
The Anxious Generation Movement has launched The Childhood Index, a first-of-its-kind, state-by-state ranking of how well technology laws and policies support healthy childhoods. Utah and New York lead, but 27 states have done nearly nothing. The Index evaluates states on distraction-free schools, childhood independence, social media age limits, and tech accountability laws. See where your state ranks and take action here.
// Applied Social Media Lab launches Transparency Hub
The Applied Social Media Lab at Harvard University has launched Transparency Hub, a new platform designed to help the public better understand how technology and social media companies govern user data.
What did you think of today’s newsletter?
We’d love to hear your feedback and ideas. Reply to this email.
// Project Liberty builds solutions that advance human agency and flourishing in an AI-powered world.
Thank you for reading.
This is uncontrollable identity of self: reputation, opportunity, competency, safety. If anyone has any control, no one can have uncontrollable identity. If everyone has uncontrollable identity, no one has any control over anyone or anything. This is morally intelligent society. A natural network surface and a personal identity surface could learn and share with uncontrollable morality. Project LIberty should aim at morally intelligent society. Thank you for all of your sharing and teaching and caring. You have already accomplished great value for everyone to learn from and appreciate. Substrate Intelligence white paper "Stable Disequilibria as Natural Intelligence" is at corus.me.
"the ability to control AI rather than be controlled by it" This is an unimaginable, unwinnable, immoral framing of our potential flourishing with moral collective living intelligence. We have natural networks inside of us and we can have the same network design in our society.
Four couplings are required and autogenerated in a natural network
Uncapturable abundance.
Inviolable safety.
Undiscoverable identity.
Unmistakeable reputation.
You can not have any of these unless you have all of these. Anything less than all of these all the time and you can restart the ai and human conflict for control. Project LIberty should not be supporting the effort to control ai. There is no chance of doing good in this effort.