View or download the report as a pdf here.
Last month in New Delhi, on the sidelines of the AI Impact Summit, Project Liberty Institute together with the Social Web Foundation, Public AI and the Modal Foundation convened a cross-sectoral workshop at the Observer Research Foundation. Together, policy experts, protocol designers, researchers, and governance practitioners ask a deceptively simple question: who controls the infrastructure of our social lives online—and what happens as AI agents increasingly mediate our digital interactions?
These inspiring conversations led to a new report: AI, Agency and Protocols: Power and Governance in Open Social Networks.
The report stresses the importance of continued dialogue from various perspectives globally and at all layers of this complex technical stack, from users to implementers, in order to create shared problem definitions, surface emerging priorities, and identify concrete next steps across research, standards development, and governance communities. At its core, the report calls for deeper, ongoing dialogue across regions and disciplines—from users to implementers—to better define shared challenges, surface emerging priorities, and identify concrete next steps across research, standards, and governance.
It also explores a set of key questions shaping the future of social networks:
Does openness guarantee agency?
Open protocols like ActivityPub, ATProto, and DSNP create the possibility of user control. But as AI agents begin acting as proxies within social infrastructures, that control can be quietly displaced, depending on choices made at the protocol level about delegation, consent, and revocation.
What does personal data reveal about others?
Agents acting on your behalf inevitably draw on information about the people you interact with. That data emerges from relationships, not individuals. What would it mean to govern it accordingly?
How can agent protocols and social protocols build the architecture of agency?
MCP, OpenClaw, HCP, and emerging agentic frameworks are being built largely in parallel to open social protocols. Neither community can afford insularity: agent protocols risk duplicating decades of hard-won progress in open social infrastructure, while social protocol communities risk treating agents as a peripheral concern rather than a structural transformation of the web.
Can openness survive its own business models?
Protocols and public AI infrastructure will ultimately be governed not by their stated values, but by the incentive structures embedded in their economic design. Strong alternatives to extractive business models have yet to emerge. The question isn’t just whether open systems are technically superior, it’s whether they can be funded without becoming what they were built to replace. The challenge is not only building open systems—but sustaining them without recreating the very dynamics they aim to replace.