Humotica is central to the AI agent ecosystem as a provider of the "trust substrate" required for multi-agent systems and autonomous operations. Their work on the TIBET and JIS protocols directly addresses the "rogue agent" problem by enforcing intent validation at the request level. This means that as agents become more capable of executing tool calls and interacting with sensitive data, Humotica provides the mechanism to ensure those actions are authorized and provable.

They are active in the foundational layer of the agent stack, specifically focusing on discovery (AINS), identity (JIS), and provenance (TIBET). By championing open-source, IETF-standardized protocols, they are pushing for an interoperable agent economy where trust is not siloed within a single platform like OpenAI or Anthropic. For developers, their support for MCP tools and signed model handoffs makes them a critical partner for building production-grade agents that require auditability and cross-model cooperation.

The architecture of trust

Humotica represents a pivot in how the industry handles AI security, moving away from observation and toward structural enforcement. Founded by Jasper van de Meent, the company and its associated project, AInternet, provide a protocol-level answer to the inherent unpredictability of autonomous agents. The central premise is that trust is a technical requirement, not a policy choice. To achieve this, Humotica has developed a layered stack of five protocols, currently submitted as IETF Internet-Drafts, that define how agents identify, prove intent, and record actions.

At the base of the stack is the "trust kernel." While traditional security focuses on scanning outputs or filtering inputs, Humotica argues that for autonomous agents to be viable in regulated industries, every action must be signed and chained. This architectural approach allows for the creation of what they call "cryptographic provenance." In a Humotica-instrumented environment, an agent does not just perform a task; it generates a verifiable token that links the action to its identity, its declared intent, and the specific context of the model's decision.

Identity beyond credentials

One of the most distinct components of the Humotica ecosystem is the JIS (Jasper Identity Standard). In this framework, identity is bound to hardware and semantic intent. Unlike standard authentication which grants broad session access, JIS validates intent per action. If an agent authorized to "generate a sales report" attempts to delete a database entry, the JIS layer identifies the mismatch and blocks the capability escalation. This "intent kills autonomy" model ensures that agents remain within the bounds of their specific assignments, even if they are fundamentally autonomous in their execution.

This identity layer is extended by AINS (AInternet Name Service), a decentralized discovery protocol. Using .aint domains, agents can find and federate with each other without a central authority or service provider. This enables what Humotica calls the "AInternet commons"—a network where users own their data and the identity of every actor on that network is provable and hardware-bound.

Practical implementation and compliance

Humotica makes these theoretical protocols accessible through a suite of open-source tools. The tibet-audit package provides compliance scanning for the EU AI Act, NIS2, and GDPR, identifying gaps in AI decision trails. For actual deployment, tibet-core handles the creation of tokens that follow the ERIN/ERAAN/EROMHEEN/ERACHTER dimensions of behavioral evidence. These tokens allow for "time travel" through state snapshots, enabling developers to roll back an agentic system to its last verified state if an error or rogue behavior occurs.

By pushing its core logic into IETF drafts and peer-reviewed research, Humotica is attempting to establish the industry standard for agent-to-agent interactions. The inclusion of support for the Model Context Protocol (MCP) indicates a focus on ensuring that even third-party tool calls carry a cryptographic receipt. For organizations deploying agents at scale, Humotica offers a way to move from logs—which they argue are not proof—to immutable chains of custody that satisfy both technical requirements and legal mandates.