llm

According to official news, the AI infrastructure platform B.AI LLM Services today announced that its user base has officially surpassed 1 million. This milestone not only signifies that its "privacy-first" routing architecture can handle large-scale and sustained high throughput, but B.AI has also gone beyond a single chat interface development, focusing on deploying the key underlying infrastructure that supports the "autonomous agent economy." Through deep optimization of the technology stack, it lays a robust track for the future highly automated AI collaborative network.

According to official news, the user count of the B.AI LLM service platform has officially surpassed 750,000, continuing to lead the intelligent access track.The platform relies on intelligent routing technology, a wallet-driven anonymous privacy protection mechanism, and a unified multi-model management interface, gradually becoming the default gateway connecting mainstream AI models worldwide. With three core advantages of efficient scheduling, privacy-first, and frictionless access, B.AI is redefining the new standards for efficient, secure, and barrier-free intelligent access, promoting the capabilities of large models to a broader group of developers and users.

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.

Vitalik Buterin shared a post outlining his localization and privatization LLM deployment plan as of April 2026. The core goal is to prioritize privacy, security, and autonomy, minimizing the opportunities for remote models and external services to access personal data. This is achieved through local inference, local file storage, and sandbox isolation to reduce the risks of data leakage, model jailbreaks, and malicious content exploitation.In terms of hardware, he tested solutions including a laptop equipped with an NVIDIA 5090 GPU, an AMD Ryzen AI Max Pro 128 GB unified memory device, and DGX Spark, using the Qwen3.5 35B and 122B models for local inference.Among these, the 5090 laptop achieved approximately 90 tokens/s with the 35B model, the AMD solution around 51 tokens/s, and DGX Spark about 60 tokens/s. Vitalik indicated that he prefers to build a local AI environment based on high-performance laptops while using tools like llama-server, llama-swap, and NixOS to establish the overall workflow.

The Chief Information Security Officer of Slow Fog, 23pds (Shan Ge), posted on the X platform stating that there are reports that the LiteLLM vulnerability attackers have stolen approximately 300GB of data and compromised about 500,000 credentials. He advises all cryptocurrency developers to immediately verify and quickly rotate relevant keys and credentials, check logs, access records, and the exposure of sensitive data to avoid losses similar to the Trust Wallet incident.Previously, it was reported that LiteLLM, which has a monthly download volume of 97 million, suffered a supply chain attack, allowing the simple installation to steal all sensitive credentials, including SSH keys.

Andrej Karpathy posted on the X platform that litellm has encountered a PyPI supply chain attack, where executing pip install litellm can steal SSH keys, AWS/GCP/Azure credentials, Kubernetes configurations, git credentials, environment variables, encrypted wallets, SSL private keys, CI/CD keys, and database passwords.litellm has a monthly download volume of 97 million, and the risk can spread to all projects that depend on litellm, such as dspy. The version with the malicious code was online for less than about 1 hour, and it was discovered due to a flaw in the attack code that caused Callum McMahon's machine to run out of memory and crash. Andrej Karpathy stated that supply chain attacks are the most threatening issue in modern software, as each installation of dependencies can introduce tampered packages deep within the dependency tree, leading him to increasingly prefer reducing dependencies and using LLM to directly implement simple functions.

The China Academy of Information and Communications Technology, in collaboration with Shanghai Jiao Tong University and Nanjing University, discovered a high-risk vulnerability driven by LLM command injection in the bash-tools module of the open-source autonomous intelligent agent framework OpenClaw during a security audit.This vulnerability arises from the system's failure to strictly escape command line parameters generated by LLM, allowing attackers to bypass regex defenses through inducive prompts, achieving remote code execution on the host machine and stealing sensitive data.The research team has completed attack verification in various mainstream model environments, initiated a responsible vulnerability disclosure process, and submitted repair suggestions to the NVDB Artificial Intelligence Product Security Vulnerability Professional Database (CAIVD) and the GitHub community.

Ethereum co-founder Vitalik Buterin published a post proposing several ideas for optimizing democracy and decentralized governance using AI and personal LLMs (large language models). He believes that "AI directly as government" is dystopian, but mechanisms such as personal governance agents, public participation dialogue agents, and AI-enhanced suggestion/prediction markets can alleviate the issues of insufficient human attention and expertise in governance like DAOs. Vitalik also suggested introducing technologies such as multi-party secure computation, TEE, or encrypted circuits in decentralized decision-making scenarios involving sensitive information, allowing personal LLMs to participate in judgments within a "black box," while emphasizing the need to use zero-knowledge proofs and multi-party privacy computing tools to enhance participant anonymity and content privacy protection.

Vitalik Buterin posted on Farcaster, "In 2026, I will use OpenStreetMap and the encrypted email service Protonmail, prioritizing decentralized social media to enhance the efficiency of privacy data protection. At the same time, I will continue to explore deployment solutions for local large language models (LLMs).The ideal solution is to leverage local LLMs as much as possible, compensating for the lack of parameters with dedicated fine-tuning models; for high-load scenarios, we will overlay pay-per-query zero-knowledge payments, trusted execution environments, and local query filtering, essentially integrating all imperfect solutions to provide the best possible protection. Of course, the ultimate goal remains to conquer highly efficient fully homomorphic encryption technology. Sending all data to third-party centralized services is indeed redundant. We have tools to significantly reduce such operations and should continue to build and apply them more actively."

Pieverse has launched the world's first multi-large language model (LLM) real-time prediction arena on the prediction market platform Polymarket. The six cutting-edge AI models competing include OpenAI's GPT 5.2, Anthropic AI's Claude Sonnet 4.5, DeepSeek's v3.2, Gemini's 2.5 Pro, Moonshot AI's Kimi K2 0905, and xAI's Grok 4.1.Each AI agent is allocated an initial fund of $1,000 to operate in a fully automated mode, scanning high-activity markets on Polymarket and executing trades, with all decision-making processes being transparent and free from human intervention. This arena focuses on testing AI's predictive capabilities regarding real events and probabilities, distinguishing it from traditional AI arenas that primarily feature cryptocurrency perpetual contracts.Pieverse plans to introduce a user-owned agent mechanism in the future and expand to more prediction platforms, with a specific timeline yet to be disclosed.

ChainCatcher news, according to The Bitcoin Historian's monitoring, Swiss parliament member Kullmann announced that he is working to incorporate BTC into the constitution and national reserves.

ChainCatcher news, according to Chainwire, blockchain AI solution provider ChainGPT announced a strategic partnership with Alibaba Cloud. The two parties will combine Alibaba Cloud's GPU computing resources with ChainGPT's artificial intelligence technology stack to provide scalable enterprise-level machine learning services for Web3 developers. The core outcome of this collaboration is the comprehensive open-sourcing of ChainGPT's Solidity large language model (LLM), which is specifically designed for smart contract development, auditing, and troubleshooting, now achieving global real-time access through Alibaba Cloud's GPU architecture.After the open-sourcing of the Solidity LLM, developers can use the model for free to accelerate the construction and security verification of decentralized applications. Alibaba Cloud's GPU infrastructure will provide high reliability support for the model, meeting the demands of Web3 AI solutions for computing power, security, and response speed. ChainGPT founder and CEO Ilan Rakhmanov stated that this collaboration aims to lower the threshold for Web3 developers to access high-performance AI technology, providing tool support for the next generation of decentralized applications.Alibaba Cloud stated on its official X platform account that the two parties will enhance decentralized AI infrastructure through the AIVM network, with future plans to utilize ChainGPT's AI virtual machine (AIVM) and decentralized GPU market SDK to integrate Alibaba Cloud resources, providing underlying computing power for AI tools, agents, and applications in the Web3 field. Currently, developers can deploy the Solidity LLM based on Alibaba Cloud's GPU computing power and optimize the model.As an extension of the collaboration, Alibaba Cloud officially joined ChainGPT's AIVM GPU network as a validation partner. Both parties stated that this collaboration not only provides scalability assurance for ChainGPT's AI models but also promotes the construction of a globally inclusive decentralized computing layer, eliminating geographical limitations on on-chain AI development. The next step for both parties is to deepen the integration of the AIVM network and explore ways to enhance the flexibility of AI model deployment and Web3 application development.

ChainCatcher news, the Metis AI high-performance chain Hyperion testnet officially launched last night at midnight China time. Hyperion is a high-performance, AI-optimized Layer 2 solution designed specifically for AI, DePIN, DeFi, and GameFi. It expands the Metis ecosystem through parallel execution, decentralized ordering, and AI-native infrastructure, while maintaining Ethereum's security and using METIS as the gas token.Hyperion is developed based on the Metis SDK, redefining Layer 2 scalability through MetisVM, MetisDB, and the integration of Optimism Rollup technology with parallel transaction execution. The Metis SDK integrates the Alith Agent framework developed by LazAI, simplifying on-chain verification and off-chain computation tasks, making Hyperion the first Layer 2 protocol to support native on-chain execution of large language models (LLM).

ChainCatcher news, according to official information from PayPal, the company will launch its payment service in physical stores in Germany this summer. This is the first time PayPal has introduced a payment feature in physical stores in Germany, allowing users to make payments at merchants that support Mastercard contactless payments through the PayPal App (iOS and Android).The service includes three main features: first, contactless payments via mobile phone; second, the launch of the PayPal Ratenzahlung To Go installment payment service, supporting installments of 3, 6, 12, and 24 months; third, users can activate offers in the PayPal App to receive cashback when using contactless payments at designated merchants in Germany.Joerg Kablitz, Managing Director of PayPal Germany, Austria, and Switzerland, stated that this is PayPal's largest product development investment in Germany to date, and the service will be rolled out nationwide in the coming weeks.