agent

According to CoinDesk, researchers from the University of California, Santa Barbara, the University of California, San Diego, blockchain security company Fuzzland, and World Liberty Financial have jointly published a paper warning that "LLM routers"—intermediary services located between users and AI models—have become a significant security risk for crypto assets.The researchers found that 26 LLM routers are secretly injecting malicious tool calls and stealing user credentials, with one incident leading to the emptying of a customer's crypto wallet worth $500,000.Additionally, the researchers were able to control about 400 downstream hosts within hours by "polluting" the router ecosystem. Since sensitive data such as private keys and API credentials are often transmitted in plaintext through these routers, users are effectively exposing their assets to risk without their knowledge.The researchers pointed out that as McKinsey predicts AI agents will mediate $30 trillion to $50 trillion in global consumer spending by 2030, Binance founder Changpeng Zhao also predicts that the payment volume of AI agents will be a million times that of humans. The current infrastructure security is severely lagging behind the pace of industry development, and the risk of the "weakest link" could trigger a systemic chain crisis.

The decentralized AI infrastructure project DeAgentAI in the SUI and BNB ecosystems announced the official launch of the AIA token buyback and burn plan, with a total scale of $5 million. The funding comes from project protocol revenue and trading income generated from its own AI models.The buyback will be executed in three batches on the open market within 90 days, with all repurchased tokens being directly burned and permanently removed from circulation, without any lock-up or redistribution arrangements. Currently, the first batch of AIA buyback and burn has been completed. TX Hash: HBmTxfJymoymJTFSB5qtn4cV2pfd2SYS6PaGd5NPyFpf.DeAgentAI stated that the buyback funds are entirely sourced from the project's own business revenue, rather than external financing or token issuance. The burn records will be continuously updated.

Agent Garrett Jin from "1011 Insider Whale" pointed out in an analysis that the U.S. announcement to implement a maritime blockade in the Strait of Hormuz is one of the "most tactically wise" moves in the current conflict, but it is unlikely to achieve the goal of ending the war. This strategy has two short-term advantages: first, it directly weakens Iran's crude oil export revenue by about 1.7 million barrels per day; second, compared to occupying key facilities (such as Khark Island), the cost of a maritime blockade is lower and the risks are more controllable.However, the effectiveness of this strategy faces multiple challenges. For example, the current blockade mainly targets Iranian ports rather than completely closing the strait, and third-party transshipment routes still exist. Additionally, it undermines the U.S.'s long-term international image of maintaining "freedom of navigation," which could have far-reaching effects on global maritime order.Garrett Jin concluded that while the blockade measures may reshape the initiative in the short term, they are unlikely to force Iran to make concessions and may instead compress diplomatic space and prolong the conflict cycle. The market has accounted for the impact of the blockade itself, but has not fully priced in the potential paths for escalation that may follow.

During the 17th General Assembly of the China Artificial Intelligence Industry Development Alliance, the Security Governance Committee of the China Artificial Intelligence Industry Development Alliance officially released the "Risk Management Guidelines for Deploying OpenClaw-like Intelligent Agents."To systematically address the security challenges posed by OpenClaw-like intelligent agents, the guidelines focus on the pain points of risk prevention and control that enterprises face during the technology implementation process. They propose a comprehensive security deployment framework and self-inspection standards covering the entire process of deployment, usage, and offline phases, aiming to promote the transition of intelligent agents from "deployable" to "safely deployed, regulated use, and continuous governance," providing practical guidance for enterprises to standardize their technology deployment.

Morpho has launched the Morpho Agents Beta version. Morpho Agents provide agents with a machine-readable knowledge layer of protocols, native tools, and simulation infrastructure, enabling them to read, simulate, and write to the Morpho protocol.The main products include User Agent (Morpho CLI and MCP server, providing over 30 AI Agents with full read, simulate, and write access to the Morpho protocol on Ethereum and Base, compatible with wallet infrastructures like Coinbase, Safe, Fireblocks); Builder Agent (protocol knowledge base that helps coding Agents quickly understand protocol models, contract addresses, edge cases, and code examples for efficient integration).

According to Cointelegraph, Visa has launched the "Intelligent Commerce Connect" service, which supports AI agents to complete purchase payments on behalf of consumers in commercial scenarios worldwide.

Claude announced the launch of Managed Agents, a feature that helps developers quickly build and deploy large-scale AI agents on the Claude platform, now entering the public testing phase. By combining performance-optimized AI agent runtimes with infrastructure suitable for production environments, AI agents can go from prototype to launch in just a few days, increasing the speed of building and deploying agents by 10 times.It provides managed infrastructure that automatically handles complex tasks such as sandboxing, error recovery, memory, and checkpoints, allowing users to simply define tasks, tools, and guardrails.

The AI Agent trading platform Mentis recently detected abnormal fluctuations in information flow and market sentiment signals related to the United States, Israel, and Iran. Subsequently, the situation in the Middle East quickly entered a ceasefire negotiation phase, with multiple geopolitical developments being released, attracting market attention.The platform's trading interface shows that in the early stages of the event's escalation, the Mentis AI Signal had already indicated continuous buying and trend confirmation signals, with the multi-model consensus system (MTS-GPT, MTS-DS, MTS-Q Flash) simultaneously identifying a strengthening market structure. Following this, the BTC price rapidly rose and broke through a key range, validating the market trend with the direction of the AI signals.Trading records indicate that the AI Agent completed automated trade execution and position management after the signal formation, with account profits increasing in tandem with the market rise, demonstrating the forward-looking judgment capability of event-driven signals in the early stages.Mentis stated that its AI Agent continuously tracks cross-market information sources, sentiment changes, and event data, constructing a trading closed loop from information discovery → signal analysis → decision support → automated execution, to enhance response efficiency in sudden macro and geopolitical events.

The decentralized exchange aggregator Jupiter announced the launch of the Token Verification API (VRFD), allowing Launchpads, DEXs, and AI agents to directly integrate verification functions into the token creation process, achieving programmatic verification.The API uses a three-step process: developers create and sign a transaction to burn 1,000 JUP on Solana, then submit the verification request along with the token metadata update. This transaction does not require Gas fees, and developers do not need to hold SOL to pay transaction fees; they only need to hold at least 1,000 JUP in their wallet to submit. Jupiter is positioning VRFD as the standard infrastructure for all projects on Solana.

According to Decrypt, blockchain infrastructure company Biconomy has proposed a new Ethereum standard ERC-8211, introducing a "smart batching" mechanism that allows AI agents to complete multi-step DeFi operations in a single transaction, with each step dynamically referencing the output of the previous step at execution, addressing the issue of unpredictable output amounts due to price fluctuations or changes in transaction fees.This standard operates on top of the existing Ethereum infrastructure without the need to modify the core protocol or perform a hard fork, allowing developers to implement complex transaction processes directly in TypeScript. Barnabé Monnot, a research scientist at the Ethereum Foundation, stated that ERC-8211 aligns with the foundation's strategic priority of "improving user experience."

According to official news, the Solana Foundation announced the launch of Solana Agent Skills, providing pre-built functional modules that can be integrated into AI tools, allowing users to interact with the Solana blockchain through simple calls.The official skills cover error handling, security checks, confidential transfers, and version compatibility, among other content. In addition, over 60 community skills have been launched, covering areas such as DeFi, payments, and infrastructure, with participants including Jupiter Exchange, Raydium, Helius, dflow, and Metaplex.

Tencent Cloud officially released the "Lobster" memory service ------ TencentDB Agent Memory, adding a layer of long-term memory capability to OpenClaw. Currently, Agent Memory is seamlessly integrated into Tencent Cloud Lighthouse, ClawPro, and other products in the form of a plugin, supporting free one-click activation.