Algorand Introduces Protocol for Managing AI Agents

cryptonews.ruPublicado a 2026-08-26Actualizado a 2026-08-26

Resumen

The Algorand Foundation has introduced AC2, an open protocol designed to enable secure user interaction with and control over AI agents. The protocol is blockchain-agnostic, establishing an encrypted direct connection between a user's wallet or application and an AI agent. AC2 allows agents to request user permission for various actions, such as executing payments, signing code, or accessing APIs. Users review and approve or deny these requests directly within their wallet interface, eliminating the need to share private keys with the AI. The protocol supports multiple operational scenarios. For instance, a software agent can seek approval before signing, while a payment agent can send transaction details to a user's wallet for verification. Shopping agents can operate within pre-set user-defined limits on permissible items and spending amounts. For messaging, AC2 utilizes DIDComm v2.0, and for authentication, it employs passkeys via Liquid Auth, which is based on FIDO2 and WebAuthn standards. The protocol supports text and voice interaction, signature delegation, and can function without a centralized intermediary or even a blockchain.

The Algorand Foundation has introduced AC2 — an open protocol that allows users to securely interact with and control the actions of AI agents. The protocol is not tied to a specific blockchain and provides an encrypted direct connection between the user's wallet or application and the AI agent.

AC2 allows agents to request permission for various actions: conducting payments, signing code, or gaining access to APIs. The user sees the request in their wallet and decides whether to allow the action or not. There is no need to transfer private keys to the AI agent.

The protocol is designed for various operational scenarios. A software agent can request permission before signing, while a payment agent can send transaction data to the user's wallet for verification. Shopping agents can operate within pre-set limits: the user defines what the agent is allowed to buy and how much it is permitted to spend.

For message exchange, AC2 uses DIDComm v2.0, and for authentication, it uses passkey via Liquid Auth, based on FIDO2 and WebAuthn standards. The protocol supports text and voice interaction, signature right delegation, and can operate without a centralized intermediary and even without a blockchain.

Image: Magnific

end-content

Preguntas relacionadas

QWhat is AC2 and what organization introduced it?

AAC2 is an open protocol introduced by the Algorand Foundation.

QWhat is the primary function of the AC2 protocol according to the article?

AThe primary function of AC2 is to allow users to securely interact with and control the actions of AI agents.

QAccording to the article, what are two types of actions an AI agent might request permission for via the AC2 protocol?

ATwo types of actions an AI agent might request permission for are making payments and signing code.

QWhat communication and authentication standards does the AC2 protocol utilize?

AThe AC2 protocol uses DIDComm v2.0 for message exchange and passkey via Liquid Auth (based on FIDO2 and WebAuthn standards) for authentication.

QAccording to the article, can the AC2 protocol operate without a centralized intermediary and without a blockchain?

AYes, the article states that the protocol can work without a centralized intermediary and even without a blockchain.

Lecturas Relacionadas

Breaking: Opus 5.1 Rumored for Release This Week, Threatening a Major Reshuffle of Top AI Agents

Breaking: Rumors suggest the new Opus 5.1 model from Anthropic will be released this week, potentially reshaping the AI agent landscape. The model reportedly focuses on dramatically improving "intelligence per token," with key upgrades in coding prowess, complex reasoning, and long-running autonomous agent capabilities. This rapid release follows the recent Opus 5 launch and appears to be a response to intense competition. OpenAI is rumored to have completed pre-training of a massive 10-trillion parameter model called "Bel," the foundation for its upcoming Astra and GPT-6. A market prediction site indicates a 74% probability that Anthropic will release its next flagship model (likely Mythos or Fable 5.1) within about two weeks, signaling an unprecedentedly dense wave of major model releases in the coming months. However, a business challenge is emerging. Despite its capabilities, Anthropic's top-tier Fable 5 model reportedly accounts for only about 11% of enterprise AI tool spending, as clients opt for more cost-effective "good enough" alternatives like Opus 5. This highlights a potential crisis where the most advanced models may not be the most commercially viable. The context is Anthropic's impending IPO, with a potential valuation reaching $2 trillion. CEO Dario Amodei has reportedly raised a pointed question in interviews, asking candidates if they would accept the company's stock price falling to zero for AI safety reasons, underscoring the tension between commercial ambition and ethical mission. The stage is set for a major reshuffle in the AGI/ASI race, driven by capital, compute power, and competing visions for the future of AI.

marsbitHace 32 min(s)

Breaking: Opus 5.1 Rumored for Release This Week, Threatening a Major Reshuffle of Top AI Agents

marsbitHace 32 min(s)

Trading

Spot
活动图片