Welcome to the Decentralization Research Center report, a monthly briefing on events and research relevant to decentralization, DAOs and governance.

This Month’s Updates

Key Notes

“Alternatives to Concentrated AI” was released this month, our landmark report on AI based on 36 interviews, 3 case studies, and 30 working examples. In it, we show that decentralization is a choice that many organizations are making across the stack — from federated learning to fully open models — and that policy will be required to incentivize and support those choices. Our policy recommendations to support this vision will be coming in 2027. A huge thank you to everyone involved in this project, and the case study authors from OpenMined, the Ethereum Foundation, and Civic AI.

Kyle Bligen published a guest article on the Ethereum Foundation’s X account on how DRC works with policymakers, developers, and industry to advance a safer, more competitive, and decentralized technological future. The article makes the case for policy engagement before legislative and regulatory frameworks harden, describes DRC’s work organizing Ethereum Layer 2 developers around shared regulatory priorities, and argues that policymakers must distinguish decentralized infrastructure from traditional financial intermediaries.

The DRC also participated in an Open Source Policy Strategy Session hosted by GitHub, UC Law San Francisco, and Proteus Strategies, which brought together technology, legal, and policy leaders to identify shared priorities for open-source development. We presented findings from Alternatives to Concentrated AI: A Layer-by-Layer Map of What Is Being Built and argued for preserving open-source AI development, protecting software developers, and regulatory frameworks that recognize the difference between centralized frontier systems and open, decentralized alternatives.

Kyle spoke on the Digital Energy Council’s Powering the Digital Economy panel at D.C. Fintech Week on the energy and compute demand driven by AI and digital assets. Drawing on DRC’s AI research, he argued for policies that expand access to the infrastructure AI development requires while letting decentralized alternatives compete: lower barriers to entry, wider access to compute and energy, and regulatory frameworks that do not entrench market concentration or disadvantage open and decentralized systems.

We co-hosted a closed-door roundtable at the D.C. Privacy Summit with the Security Alliance (SEAL), as part of Project Glitch’s work to strengthen coordination between industry, law enforcement, and regulators. Around 30 stakeholders from government agencies, cybersecurity organizations, and industry took part under the Chatham House Rule. The roundtable featured a tabletop simulation developed by SEAL to examine practical challenges associated with responding to cyber incidents, sharing actionable intelligence, and coordinating efforts to freeze, seize, and recover stolen funds. These discussions helped identify existing gaps in public-private coordination and opportunities to improve incident response capabilities, strengthen consumer protection, and combat sophisticated cybercriminal and state-sponsored threats without imposing unnecessary restrictions on decentralized infrastructure.

Finally, we submitted two comment letters to the U.S. Securities and Exchange Commission on its proposed Regulation Crypto Assets framework and March interpretive guidance. We welcome the Commission’s push for regulatory clarity but warn that the proposal could let issuers exit investment contract obligations without demonstrating any real reduction in control over the underlying network. We recommend objective, verifiable standards for evaluating decentralization, alongside a non-exclusive safe harbor grounded in demonstrable loss of control, so that the rules distinguish decentralized networks from centrally controlled enterprises, protect investors, and give projects a reason to decentralize over time. Sarah Brennan walks you through our submissions in this month’s “The Deep Dive” below.

The Deep Dive

In “DRC's SEC comment letters, explained: Who decides when a network is decentralized?” Sarah Brennan explains the DRC’s two comment letters on the proposal this month, calling the proposed exemptions and safe harbor meaningful progress and asking the SEC to add a bright-line safe harbor for when essential managerial efforts have ceased, built on an objective, control-based test:

"We want to limit the footprint of the promises test. We would confine the promises that count to those made before the network is functional and to covenants disclosed in public filings, so the promises element is not overweighted. The main line in the sand should be control. Once no person or group controls the network, promised efforts left unperformed cannot be essential, because no one is positioned to perform them."

Sarah Brennan
Continue reading here…

The Full Rundown

Further reading from the space:

• The Ethereum Foundation’s Vittorio Rivabella introduces zkAPI, built with the Open Anonymity Project. Users prepay into a vault and spend through zero-knowledge proofs. Neither AI provider nor payment server can link a request to a payer.

• Gupta, Fuller and Speer at the Manhattan Institute propose a safe harbor for open weights.

• Spain’s International Council on AI, with Castells, Bria, Véliz, Crawford and Cerf among its members, publishes the Madrid Manifesto for Responsible AI: “Self-regulation cannot ensure that AI develops in ways that advance the public interest. Democratic societies therefore have a legitimate role in shaping, governing and overseeing AI so that its benefits, opportunities and capabilities are shared broadly and aligned with human rights and democratic values.”

• The Allen Lab’s Kristen Busch and Charlie Covit explain how California writes public deliberation into law.

• SpruceID reads Utah’s SEDI statute as digital identity without surveillance: credentials signed once, verified locally, and revoked through status lists.

• The ACLU’s Jay Stanley looks at how technologies move from empowering to imprisoning.

• EFF’s Jillian York and Sheila Lalwani ask what happens when age verification means that no ID means no internet — even zero-knowledge age proofs rely on a person having a digital ID in the first place.

• Open Future’s Paul Keller and Sebastian Posth propose moving from a registry to a registry ecosystem: “[A]utomated systems now make decisions relevant to copyright about billions of works, such as filtering uploads, assembling training data for artificial intelligence, and resolving reuse permissions. Those systems depend on reliable, machine-readable rights information. But currently there is no public infrastructure that can provide it at scale.”

• Thirty researchers led by Ziyue Luo and Jiaxuan Cai argue for democratizing LLM training.

• Harvard’s Guowei Xu, Yilun Du and colleagues train agents toward self-orchestration in decentralized swarms. Each agent is trained to bid alongside its actions; bids decide who acts and who gets credit, so coordination emerges without an orchestrator and survives agents dropping out.

• Andrew Lewis-Pye and Ehud Shapiro take grassroots platforms beyond decentralization.

• Brookings’ Landry Signé and WEF’s Karla Yee Amezaga set out how to shape global governance of AI agents.

  
  

If you’re working on related research or would like to get involved in our work, please reach out to us via info@thedrcenter.org. We’d love to hear from you!