A Carnegie Mellon research team has proposed LOKA (Layered Orchestration for Knowledgeful Agents), a protocol that ties AI agent identity directly to an ethics-and-consensus layer rather than treating governance as a separate concern. The paper frames the problem as three unresolved questions about autonomous agents: who or what the agent is, whether its actions can be verified and audited, and whether it can be trusted to align with human values. The architecture has four layers: identity, governance, security, and consensus. The identity piece, called the Universal Agent Identity Layer (UAIL), is built on Self-Sovereign Identity principles — each agent gets a unique, verifiable Decentralized Identifier and can acquire Verifiable Credentials attesting to its capabilities and ethical standing. Critically, the agent itself controls this identity rather than a central authority, which the authors argue is what makes accountability possible in the first place rather than just identification. What distinguishes LOKA from identity-only proposals is the Decentralized Ethical Consensus Protocol (DECP) sitting alongside UAIL. Instead of a single global ethics policy, DECP is designed to let agents make context-aware decisions grounded in shared ethical baselines while still respecting jurisdictional differences — a financial agent operating under EU rules and one operating under US rules could both hold LOKA-compliant identities without collapsing into a single rigid rulebook. The paper positions LOKA as a possible complement to existing interoperability efforts like Google's Agent2Agent and Anthropic's Model Context Protocol, which handle how agents talk to each other but don't specify who they are or how their behavior gets judged. LOKA's authors argue that fragmentation across siloed agent systems — no common protocol for identity, ethical reasoning, or regulatory compliance — creates real risk as agents multiply across sectors, since interoperability gaps and accountability gaps tend to compound each other. The security layer adds post-quantum cryptography and privacy-preserving techniques like homomorphic encryption, aimed at agents that need to prove things about themselves without exposing the underlying data. This is still a research proposal rather than a deployed standard — there's no reference implementation or adopting registry cited in the paper — but it's a notable attempt to specify identity and ethical accountability as the same protocol layer rather than two separate stacks bolted together after the fact. The open question is whether DECP's jurisdiction-aware consensus mechanism can actually resolve conflicts between agents operating under genuinely incompatible ethical frameworks, or whether it just pushes that judgment call to whichever human eventually has to arbitrate a dispute.