Snippings

xxx

The FIDO Alliance today announced initiatives to develop interoperable standards for agentic interactions and commerce. These initiatives include the formation of an Agentic Authentication Technical Working Group and efforts to develop specifications for agent-initiated commerce, drawing from initial contributions from Google (AP2) and Mastercard (Verifiable Intent).

 

These efforts focus on three core areas:
Verifiable User Instructions – Enabling users to authorize AI agents through clear, phishing-resistant mechanisms so agents only perform approved actions, including transactions, without exposing credentials.
Agent Authentication – Allowing services to verify that an AI agent is acting on behalf of an authenticated user and within defined parameters, distinguishing legitimate agents from unauthorized actors.
Trusted Delegation for Commerce – Defining how agent-initiated transactions can be executed within user-controlled boundaries, with verifiable authorization. This work recognizes that trusted agentic transactions require not only strong authentication, but also clear, verifiable authorization mechanisms aligned with real-world commerce and payment flows.

From: FIDO Alliance to Develop Standards for Trusted AI Agent Interactions | FIDO Alliance.

xxx

xxx

5.1 Model Context Protocol (MCP)

Description: MCP is a widely adopted open specification initiated by Anthropic and now open-sourced, aimed at specifying how applications provide context to large language models (LLMs). It is metaphorically described as the “USB-C port for AI applications,” targeting the M×N integration challenge faced when AI models connect with external data sources, tools, and systems (such as cloud platforms, enterprise databases, local files). By providing a unified interface, MCP simplifies interactions between AI models and the external world, reducing the need to build custom connectors for each new data source or tool [ref11].
Key Features/Mechanisms: MCP adopts a client-server architecture where AI applications (such as chat assistants, AI-driven IDEs) act as MCP clients, connecting to one or more MCP servers that expose capabilities or data . Its core interaction primitives include: Tools (executable functions that can be dynamically invoked, such as API calls), Resources (structured static data streams for AI reference), and Prompts (reusable conversation workflows or templates). The protocol layer handles message framing, request/response mapping, and notification delivery, supporting various transport protocols such as Stdio for local processes and HTTP+SSE for network services.
Focus Areas/Target Challenges: MCP’s core goals are to provide structured context injection for LLMs, enable flexible plugging of tools and knowledge, support secure infrastructure integration, and ensure compatibility across different LLM vendors. It aims to enhance AI models’ context awareness and dynamic tool discovery and execution capabilities.
Core Technologies Used: JSON-RPC (for client-server interface), HTTP, Server-Sent Events (SSE).
5.2 Agent-to-Agent Protocol (A2A)

Description: A2A is an open protocol initiated by Google and jointly promoted with over 50 industry partners, designed to enable independent AI agents built on different frameworks by different vendors to communicate, collaborate, and coordinate actions securely and seamlessly. It aims to address interoperability issues in heterogeneous agent ecosystems, allowing agents to work together without exposing their internal states, memories, or tools [ref8] [ref9].
Key Features/Mechanisms: A2A’s core architecture revolves around client agents and remote agents. Agents publish their identities, capabilities, skills, service endpoints, and authentication requirements through “Agent Cards” (JSON metadata documents), enabling capability discovery. The protocol supports task management lifecycle, allowing creation, sending, and tracking of task status, and can handle long-running tasks that might take hours or even days. A2A supports multiple interaction modalities, including text, files, structured JSON data, as well as audio and video streams.
Focus Areas/Target Challenges: A2A focuses on enabling dynamic interaction, capability sharing, and task coordination between opaque, autonomous agents, especially in enterprise-grade workflows. It aims to break down agent silos, simplify enterprise integration, and foster a more interconnected, powerful AI ecosystem. A2A addresses different problems than MCP, with MCP focusing on connecting agents with tools/data, while A2A focuses on collaboration between agents.
Core Technologies Used: HTTP(S) (as the transport layer, requiring TLS encryption), JSON-RPC 2.0 (as the payload format for requests and responses), Server-Sent Events (SSE) (for real-time streaming communication from server to client, such as task status updates).
5.3 Agent Network Protocol (ANP)

Description: ANP is an open-source protocol with the vision of becoming “the HTTP of the Agentic Web era,” designed to build an open, secure, and efficient collaborative network for billions of agents. It aims to address the inadequacies of current internet infrastructure in meeting the specific needs of agent networks. ANP is developed and maintained by the open-source community, which is committed to an open, neutral stance, and the community pledges never to commercialize [ref6] [ref7].
Key Features/Mechanisms: ANP primarily addresses the connection and collaboration of agents on the internet, adopting a three-layer architecture:
Identity and Encrypted Communication Layer: Builds decentralized authentication schemes and end-to-end encrypted communication based on W3C DID (Decentralized Identifier) specifications, enabling cross-platform agents to authenticate each other without relying on centralized systems.
Meta-Protocol Layer: A protocol for negotiating communication protocols between agents, key to enabling self-organization and self-negotiation for efficient collaboration in agent networks.
Application Protocol Layer: Based on Semantic Web specifications, using JSON-LD and schema.org to describe agent information, capabilities, and interfaces. The entry point for an agent is an Agent Description Document. Uses RFC 8615 to design agent discovery mechanisms, W3C VC to implement credential records for transactions between agents, while reusing many existing specifications such as OpenAPI, WebRTC, etc.

Figure 3 Figure 3: ANP three-layer protocol architecture showing the identity layer, meta-protocol layer, and application protocol layer with their interactions
Focus Areas/Target Challenges: ANP aims to address three core challenges of the agent internet: achieving interconnection between all agents, breaking data silos, and ensuring AI can access complete contextual information; providing AI-native interfaces for efficient interaction with the digital world through APIs or communication protocols rather than mimicking human operations; and utilizing AI for automatic organization and negotiation between agents to build a more economically efficient collaborative network. It particularly focuses on decentralized discovery and collaboration in open internet environments, as well as interoperability across heterogeneous domains.
Core Technologies Used: W3C Decentralized Identifiers (DIDs), JSON-LD, W3C Verifiable Credentials (VC), end-to-end encryption technologies.
5.4 Agent Connect Protocol (ACP)

Description: ACP is an open-source protocol led by Cisco and developed in collaboration with partners such as LangChain and Galileo as part of the AGNTCY initiative, designed to provide a communication layer for autonomous agents to collaborate and share resources in distributed systems [ref12].
Key Features/Mechanisms: ACP adopts RESTful APIs as the standard interface, defining how agents interact, including retrieving workflows that agents can execute, creating and managing context threads, and running agents. It supports stateful communication threads, allowing agents to negotiate and reason together during tasks, and implements loosely coupled interactions through message passing. Agent discovery is achieved through the Agent Directory and OASF (Open Agentic Schema Framework) documents, which are standardized JSON files describing agent capabilities, invocation methods, input/output patterns, and more. ACP supports asynchronous-first interactions, multi-part messages, and observability features.
Focus Areas/Target Challenges: ACP primarily addresses communication barriers and collaboration efficiency issues between heterogeneous agents (potentially built on different technology stacks or frameworks) in enterprise environments. It aims to enable scalable, standardized multi-agent interactions, allowing multiple agents to work together as a logical unit to complete complex tasks.
Core Technologies Used: RESTful APIs, JSON (for OASF documents and message schemas). Can be integrated with workflow frameworks such as LangGraph.
5.5 Agent Communication Protocol (ACP)

Description: Agent Communication Protocol (ACP) is an open-source standard contributed by IBM to the Linux Foundation, designed to provide a shared language for heterogeneous AI agents to enable connection, collaboration, and complex task execution. The protocol’s primary goal is to eliminate vendor lock-in and promote the development of agent ecosystems through an open governance model [ref13].
Key Features/Mechanisms: ACP defines a standardized RESTful API that supports synchronous, asynchronous, and streaming interactions, adopting a peer-to-peer interaction design. Its core features include:
No Specialized SDK Required: The protocol is designed to enable interaction without specialized SDKs, allowing direct use of standard HTTP tools while also providing Python/TypeScript SDKs.
Offline Discovery: Enables agent discovery through metadata that can be embedded in distribution packages for offline discovery, using Agent Detail models to describe agents.
Peer-to-Peer Interaction: Emphasizes peer-to-peer interaction, supporting direct communication and collaboration between agents.
Complementary to MCP: Forms a complementary relationship with Model Context Protocol (MCP), focusing on inter-agent communication.
Focus Areas/Target Challenges: ACP primarily addresses interoperability issues between heterogeneous AI agents, enabling cross-framework and cross-technology stack agent collaboration through a shared communication language. The protocol particularly emphasizes avoiding vendor lock-in, using an open-source, Linux Foundation governance model to ensure the openness and neutrality of the standard.
Core Technologies Used: HTTP, JSON, OpenAPI Specification, Python/TypeScript SDKs. The protocol relies on underlying HTTP(S) and enterprise-grade security practices of the deployment environment, supporting discovery and interaction in secure/air-gapped environments.

From: Agent Network Protocol White Paper.

xxx

xxx

According to the charging documents, the defendants who created cryptocurrency companies made false statements about their cryptocurrencies (“tokens”) and executed sham trades in those tokens (“wash trades”) to create the appearance of trading activity that would make the tokens look like good investments. These deceptive tactics allegedly attracted new investors and purchasers, which resulted in an increase in the tokens’ trading prices. The defendants are then alleged to have sold their tokens at the artificially inflated prices, a fraud commonly known as a “pump and dump.” The largest of these cryptocurrency companies, Saitama, at one point had a multi-billion-dollar market value.

From: District of Massachusetts | Eighteen Individuals and Entities Charged in International Operation Targeting Widespread Fraud and Manipulation in the Cryptocurrency Markets | United States Department of Justice.

xxx

xxx

A digital wallet isn’t just a place to hold crypto tokens. It’s any software that verifies identity and links you to financial, social and even government systems. It may seem like Apple or Meta already manage our digital lives, but their power is still limited. The influence of Apple’s digital wallet, say, is real, but it’s checked by competition from banks, payment networks and rival platforms.

Crypto wallets work differently. They are digital wallets, but they eliminate intermediaries entirely, linking users directly to blockchain-based platforms and decentralised apps rather than traditional websites. That gives their designers outsized influence over who participates in the digital economy.

They might sound niche today, but it’s where the digital economy is heading as crypto use increases and central banks develop digital currencies. Crypto wallets can merge financial, personal and even civic information into a single system. The same technology that verifies your driving licence or health record could also determine whether a payment clears or what data you must share to use a service. This level of integration gives whoever builds the wallet enormous power to set the terms of everyday life.

From: Whoever controls digital wallets will control the future.

xxx

xxx

Key takeaways

Data portability should be a fundamental consumer right. It gives individuals control over their personal information across digital services.
Data portability can be a game changer in the digital economy. It enhances consumer choice and market competition by making it easier for consumers to switch between providers and lowering barriers to for companies to enter the market.
Even when benefits are clear, many people will not switch. Habits, risk aversion and loyalty to current providers are behavioural barriers that can limit adopting data portability enabled services.

From: Your Data, Your Control.

xxx

xxx

Three men are facing 44 charges in a first-of-its-kind cybercrime investigation in Canada.

Project Lighthouse began in November 2025 after a security partner alerted law enforcement to a suspected SMS blaster operating in downtown Toronto.

This marks the first known instance of this technology being used in Canada and highlights an emerging threat to both public safety and financial security.

From: Unprecedented SMS Blaster Arrests.

xxx

xxx

Invitations have gone out to 36,000 UK addresses asking recipients to weigh in on the question: “How should we design a Digital ID system for the UK?” – a framing that, as worded, rather forecloses the answer “Don’t bother.”
Those selected will attend in-person workshops in central Birmingham from 9.30am to 5pm across the weekends of 30 to 31 May and 20 to 21 June, plus three weekday evening sessions on Zoom from 6-9pm, beginning Thursday 21 May. The letter, seen by The Register, does not include the address of the Birmingham venue.
Around 100 to 120 participants are being chosen from volunteers to represent a cross-section of the UK population

From: UK govt shells out £550 for Digital ID panel, bans press.

xxx

xxx

Now, however, the Commodity Futures Trading Commission, which oversees prediction markets, wants you to know that it’s watching very, very closely. The agency is searching for suspicious behavior from traders within the United States who have been sneaking onto offshore markets, including Polymarket’s crypto platform—which is blocked stateside—by using virtual private networks. “We’re going to find them, and we’re going to bring actions,” agency chairman Michael Selig told WIRED this week, speaking from the CFTC’s headquarters in Washington, DC.

From: The US is betting on AI to catch insider trading in prediction markets.

xxx

xxx

Three-quarters of agentic activity (69.57%) touched product and search routes: browsing product listings, reading articles, and running searches. This echoes the annual finding that most agents today are exploring product listings rather than completing transactions.
The remaining activity splits across authentication (9.22%), user account routes (8.55%), miscellaneous (4.82%), content engagement (4.68%), and checkout and payment (3.16%).

From: State of Agentic Traffic – April 2026: Agentic browsers generate nearly three-quarters of agentic traffic.

xxx

xxx

Overview
SPIFFE and SPIRE provide a uniform identity control plane across modern and heterogeneous infrastructure. Since software and application architectures have grown substantially, they are spread across virtual machines in public clouds and private data centers. Security models for the organizations that manage them must keep up with these infrastructure technologies. And this is where SPIFFE and SPIRE come in. With SPIFFE/SPIRE, developers and operators can build software using new infrastructure technologies, while allowing security teams to step back from time-consuming security processes.

From: SPIFFE – Secure Production Identity Framework for Everyone.

xxx