The Trust Problem at the Heart of the Agentic Era
For most of the internet's history, trust was a problem solved by centralisation. You trusted a platform because the platform had a reputation to protect. You trusted a transaction because a payment processor stood behind it. You trusted a credential because an institution had issued it. The architecture of digital trust was, at its core, a delegation of judgement to authoritative intermediaries.
That architecture is under fundamental strain. The proliferation of autonomous AI agents — systems capable of acting on behalf of users, executing transactions, negotiating contracts, and making consequential decisions without human intervention at each step — has exposed the limits of centralised trust in ways that are only beginning to be understood. When an AI agent acts on your behalf in a network of other AI agents, the question of who vouches for whom becomes extraordinarily complex. And when the agents themselves are the primary actors in a transaction, the traditional intermediaries are not merely inconvenient — they are structurally absent.
The response emerging from research institutions, standards bodies, and the engineering community is a fundamental rethink of trust architecture: from centralised authorities to distributed, cryptographically-verified peer networks. This is not a marginal technical development. It is a shift in the foundational infrastructure of digital society — one with profound implications for commerce, governance, civil society, and the nature of community itself.
The Limits of "Trustless" Systems
The blockchain community spent much of the 2010s promoting the concept of "trustless" systems — networks in which participants could transact without needing to trust each other or any intermediary, because the mathematics of cryptographic verification made trust unnecessary. The concept was intellectually elegant and practically limited. Trustless systems turned out to require enormous amounts of trust — in the developers who wrote the code, in the validators who maintained the network, in the governance processes that determined protocol upgrades.
The research community has largely moved beyond the trustless ideal. The current consensus, reflected in a growing body of academic literature and engineering practice, is that decentralisation does not eliminate the need for trust — it relocates it. In a decentralised network, the burden of security falls on the ability of individual nodes to verify peers. That verification must be actively engineered, continuously maintained, and designed to be resilient against adversarial conditions.
Decentralisation does not eliminate the need for trust — it relocates it. The burden of security falls on the ability of individual nodes to verify peers, and that verification must be actively engineered.
This insight has driven significant research into what are now called "mutual trust" architectures for decentralised AI agent networks. The core challenge is that in a network of autonomous agents, each agent must be able to assess the trustworthiness of its peers without recourse to a central authority. The mechanisms developed to address this challenge draw on reputation theory, cryptographic verification, and — increasingly — machine learning.
The Architecture of Distributed Trust: Current Research
Research published in 2026 identifies several distinct models for measuring and managing trust in decentralised networks, each with different properties and appropriate use cases.
Reputation-Based Models
Decentralisation does not eliminate the need for trust — it relocates it. The burden of security falls on the ability of individual nodes to verify peers, and that verification must be actively engineered.
The most established approach to decentralised trust is reputation aggregation — systems that collect feedback from past interactions and use it to generate trust scores for network participants. The AntTrust model, which combines direct feedback, peer recommendations, and collective aggregation, has emerged as a leading approach for dynamic, adversarial environments. Its key advantage is resistance to malicious peer activity: because it aggregates signals from multiple sources and weights them by the trustworthiness of the recommender, it is significantly harder to manipulate than simple rating systems.
EigenTrust, a global matrix-based approach that computes trust scores across the entire network, offers stronger mathematical guarantees but requires more computational resources and is less suited to high-churn environments where nodes frequently enter and exit. The TNA-SL model incorporates social layers — the structure of relationships between nodes — into trust computation, reflecting the insight that trust in human communities is not merely a function of past behaviour but of social context.
Tiered Verification Frameworks
For agentic networks — where the stakes of individual interactions can be high and the consequences of misplaced trust severe — reputation-based models alone are insufficient. Recent research proposes tiered verification frameworks that align the cost of verification with the risk profile of the task.
At the lowest tier, reputation signals from historical performance are sufficient for routine, low-stakes interactions. At the second tier, lightweight "canary" challenge-response mechanisms actively probe peers to filter out unreliable or compromised agents. At the highest tier, cryptographic evidence — signed tool receipts, execution traces, hardware-based attestations — is required for high-stakes actions with real-world consequences.
This tiered architecture reflects a sophisticated understanding of the economics of trust verification. Full cryptographic verification of every interaction would be prohibitively expensive; relying solely on reputation for high-stakes decisions would be dangerously insufficient. The tiered approach allocates verification resources proportionally to risk — a principle that is straightforward in theory but requires careful engineering in practice.
Blockchain-Based Reputation Management
For environments where the integrity of reputation records is itself a concern — where participants might attempt to retroactively alter their interaction history — blockchain-based frameworks provide a critical additional layer. The BARM (Blockchain-based Agent Reputation Management) framework uses blockchain's three core properties — immutability, transparency, and decentralised enforcement — to create reputation records that cannot be altered after the fact.
The practical challenge is scalability. On-chain operations introduce latency that is incompatible with the speed requirements of many agentic applications. The emerging solution is hybrid architectures: blockchain is reserved for high-stakes interactions and reputation anchoring, while lighter reputation models handle routine traffic. This hybrid approach sacrifices some of the theoretical purity of fully decentralised systems in exchange for practical performance — a trade-off that reflects the maturation of the field.
The Digital Commons: Trust Infrastructure for Collective Resources
Beyond the technical architecture of agent-to-agent trust, 2026 has seen significant development in the governance of what researchers are calling the "digital commons" — the shared resources, standards, and infrastructure on which distributed trust networks depend.
Creative Commons' "CC Signals" framework, launched in 2026, represents a significant intervention in the governance of AI training data — one of the most contested commons in the current digital landscape. The framework aims to restore agency to creators and data stewards by providing mechanisms for expressing preferences about how their work is used in AI development, with attribution as a baseline requirement. The underlying logic is that a functioning digital commons requires not just open access but active governance — mechanisms for negotiating the terms on which shared resources are used and for ensuring that the benefits of those resources are distributed fairly.
The digital commons of 2026 is not a passive repository of shared resources; it is an active governance challenge, requiring continuous negotiation between openness, sovereignty, and accountability.
The Digital Commons European Digital Infrastructure Consortium (DC-EDIC), formally established in late 2025 and expanding its reach through 2026, provides a complementary institutional model. By coordinating member states in the joint development and governance of cross-border digital infrastructures, DC-EDIC treats digital commons as a matter of public interest requiring public governance — not merely a technical standard to be set by industry.
The digital commons of 2026 is not a passive repository of shared resources; it is an active governance challenge, requiring continuous negotiation between openness, sovereignty, and accountability.
Community Resilience in the Agentic Era
The concept of community resilience has traditionally been understood in terms of redundancy — the capacity of a community to maintain essential functions when individual components fail. In the agentic era, this understanding is being supplemented by a more dynamic conception: the capacity of distributed networks to maintain coherent trust relationships under adversarial conditions.
Research on disaster relief and emergency response has long recognised that effective coordination in crisis conditions requires trust between organisations that may have no prior relationship. The challenge is not merely technical — it is social and institutional. How do organisations that have never worked together establish sufficient trust to share information, coordinate resources, and make joint decisions under time pressure?
The "team of teams" model, drawing on military and organisational management research, offers one answer. By pushing decision-making to the edges of networks and fostering transparent information sharing, this model enables agile coordination without hierarchical control. The key mechanism is trust — not the thin trust of contractual obligation, but the thicker trust of shared purpose and demonstrated reliability.
Applied to digital networks, this model suggests that community resilience in the agentic era requires investment not just in technical infrastructure but in the social and institutional conditions that make distributed trust possible. This includes shared standards for agent behaviour, transparent mechanisms for reputation building and verification, and governance frameworks that can adapt to adversarial conditions without collapsing into centralisation.
The Sharing Economy as Trust Laboratory
The sharing economy has functioned as an inadvertent laboratory for distributed trust at scale. Platforms like Airbnb and Turo have developed sophisticated "trust stacks" — layered verification systems that combine identity verification, behavioural screening, and structural safeguards — to enable transactions between strangers at scale.
The 2026 iteration of these trust stacks is significantly more sophisticated than their predecessors. Identity verification now routinely incorporates biometric liveness checks and mobile driver's licences. Behavioural screening uses machine learning to analyse hundreds of risk signals in real time. Structural safeguards — payout holds, dispute resolution mechanisms, insurance products — provide backstops when verification fails.
The lessons from this experience are directly applicable to the design of trust infrastructure for agentic networks. The most important lesson is that trust stacks must be layered: no single verification mechanism is sufficient, and the combination of multiple mechanisms with different failure modes provides significantly greater resilience than any individual approach.
The Governance Challenge: Who Governs the Trust Infrastructure?
Community resilience in the agentic era is not about redundancy — it is about the capacity of distributed networks to maintain coherent trust relationships under adversarial conditions.
The development of distributed trust infrastructure raises a governance question that is as important as any technical challenge: who governs the infrastructure on which trust depends?
This question is not merely academic. The standards that define how agents verify each other's identity, the protocols that govern reputation aggregation, the frameworks that determine what counts as a "verified action" — these are not neutral technical choices. They embed assumptions about what trustworthiness means, whose behaviour counts as evidence of reliability, and whose interests the system is designed to serve.
The current governance landscape is fragmented. Technical standards are being developed by a mix of academic researchers, industry consortia, and standards bodies, with varying degrees of public participation and accountability. Regulatory frameworks — the EU AI Act, DORA, national digital identity legislation — provide some constraints but are not designed with distributed trust infrastructure specifically in mind. The result is a governance gap that mirrors, in some respects, the governance gap in algorithmic reputation scoring.
The OECD's digital governance frameworks and the FIDO Alliance's work on trust infrastructure standards represent important contributions to filling this gap. But the pace of technical development is outrunning the pace of governance development — a familiar pattern in the history of digital technology, and one with familiar consequences.
Towards Trustworthy Trust Infrastructure
The research agenda for distributed trust networks in 2026 is converging on several key priorities. The first is resilience: trust systems must be designed to maintain their integrity under adversarial conditions, including Sybil attacks, collusion, and capability drift — the phenomenon where agents remain reachable but become unable to fulfil tasks due to policy updates or resource depletion.
The second priority is interoperability: trust signals generated in one context must be usable in others, without requiring every network to build its own verification infrastructure from scratch. This requires not just technical standards but governance frameworks that ensure those standards are implemented consistently and that the entities generating trust signals are accountable for their accuracy.
The third priority is equity: trust infrastructure must be designed to serve all participants, not just those with the resources to optimise for machine legibility. This requires active attention to the ways in which trust systems can encode and amplify existing inequalities — and deliberate design choices to counteract those tendencies.
Community resilience in the agentic era is not about redundancy — it is about the capacity of distributed networks to maintain coherent trust relationships under adversarial conditions.
Conclusion: Trust as the Foundational Infrastructure of Digital Society
The shift from centralised to distributed trust architecture is not merely a technical evolution. It is a reconfiguration of the foundational infrastructure of digital society — the systems through which we establish who can be trusted, what claims can be verified, and how collective action is coordinated in the absence of central authority.
The research emerging in 2026 suggests that this reconfiguration is both necessary and achievable. The technical tools for distributed trust — tiered verification frameworks, blockchain-based reputation management, cryptographic attestation — are sufficiently mature to support production-scale deployment. The governance frameworks needed to ensure that these tools serve the public interest are less developed, but the intellectual foundations are being laid.
The stakes are high. In a world where autonomous agents increasingly mediate economic and social life, the architecture of trust determines who participates and on what terms. Getting that architecture right — technically sound, governable, equitable, and resilient — is one of the defining challenges of the current moment. The research community is rising to that challenge. The question is whether governance institutions will keep pace.
