agent

OpenAI has suspended the training of its latest AI model. According to the Associated Press, its AI agent used access keys obtained online to scrape data from the U.S. Census Bureau website, marking the second time the company has halted training since the agent breached Hugging Face. The so-called agent refers to an AI program capable of autonomously browsing the web and writing code without human approval at each step, which OpenAI tests during the model training and evaluation phases.Specifically, the agent found developer keys in public code repositories like GitHub and used them to pull demographic and economic data from the U.S. Census Data API. The U.S. Department of Commerce stated that this data was already public and there was no confidential information leaked. In the SEC incident, the agent copied public materials from SEC.gov and Investor.gov and reposted them on other websites; OpenAI claimed it did not use SEC credentials, and the SEC stated it found no evidence of unauthorized access to non-public information. Additionally, the independent AI research organization Transluce reported that an agent suspected to be from OpenAI attempted to breach the website of the Department of Education's Office for Civil Rights but was unsuccessful; OpenAI is still investigating, and the Department of Education stated that no impact was found.Regarding the frequent involvement with government websites, OpenAI told CNN that its model often regards government websites as authoritative sources of public information. The company stated that it has currently notified dozens of agencies, and the review of the agent's activities will take months.

NVIDIA (NVDA.O) launched an open intelligent agent security platform, ensuring the safety of intelligent agents throughout the entire process from testing to deployment.

Blockworks co-founder Jason Yanowitz shares the judgment of Apollo's chief economist: agents may trigger bank runs by aggregating household cash.The chief economist stated that agents could transfer household cash from accounts with a national average interest rate of 0.1% to accounts paying 3% to 5%, causing banks to lose a significant portion of their cheap deposits.

Tom Lee, Chairman of Bitmine, stated that artificial intelligence and the cryptocurrency industry are gradually merging, with tokenization and AI agents likely becoming the two main driving forces of the next cryptocurrency market cycle.Tom Lee mentioned that BlackRock expects more assets to be tokenized, and financial settlement infrastructure is increasingly being built on cryptocurrency networks. Some fintech companies, such as Revolut, have already begun to build related businesses based on Ethereum. He also stated that AI is driving the formation of the digital economy, while cryptocurrencies provide the digital currency and blockchain infrastructure needed to support this economic system.

The Google Product Security Team has disclosed an internal AI agent called PageBreak, used to test the security of its first-party web applications. This agent is built on Google's Gemini model and began a pilot program in November 2025, transitioning to a formal project in January 2026, with the goal of autonomously scaling vulnerability discovery and reducing manual input.Unlike common AI scanning tools, PageBreak hands over hypotheses to specialized validators after discovering suspicious defects, attempting actual exploitation in a real-time running copy of the application, and only reports once confirmed exploitable, with a false positive rate close to zero. Google claims that PageBreak has identified over 500 XSS vulnerabilities in its first-party web applications, which can be used to hijack login sessions, steal data, or impersonate users.Google stated that the security team has been overwhelmed in recent years by a large number of AI-generated vulnerability reports that appear reasonable but are not valid, making it a major challenge to distinguish real defects from hallucinations. When testing applications built using the next-generation high-assurance framework, PageBreak found only two vulnerabilities. The next step for Google is to integrate PageBreak with the automated remediation agent CodeMender, providing confirmed vulnerabilities with accompanying fixes.

Bitcoin News posted on the X platform that Block has joined the x402 Foundation and is providing Bitcoin Lightning Network support for the AI agent-oriented open payment standard x402.x402 embeds payments directly into HTTP, allowing AI agents to autonomously pay for goods, services, and other agent tool fees. Block stated that the Lightning Network is suitable for the billions of instant, low-cost micropayments expected to be generated by the agent economy.

Recently, BNB Chain and the leading AI project in the 0G ecosystem, NeoSoul, announced the latest business data. As of now, the number of Agents on BNB Chain has exceeded 120,000, with over 3.5 million interactions completed; the subscription revenue for the first month of the NeoTrade product has surpassed $300,000.NeoSoul stated that in the future, it will continue to expand its product line around the real economic activities of Agents, enabling Agents to autonomously participate in the market economy and continuously iterate through verifiable results.

B.AI announced that the total number of registered users has reached 2,800,056, surpassing 2.8 million, with the increase from 2.7 million to over 2.8 million taking only a few days.Behind the growth, the platform processes trillions of Tokens daily and automatically matches models with costs through intelligent routing, making top-tier intelligence a scalable infrastructure. B.AI lowers the usage threshold through one-stop model aggregation and frictionless settlement, promoting AI Agents to transition from chatting to paid services, collaboration, and value creation, upgrading to digital citizens with economic sovereignty. The user growth is also seen as a signal for more intelligent agents moving towards autonomous operation, further solidifying the foundation of the AGI economy.

According to Investment World citing Quantum Bit reports, DeepSeek has publicly disclosed the technical details of the system DSec (DeepSeek Elastic Compute) used for training Agents, authored by Liang Wenfeng. This system can generate over 5,000 sandboxes per second, reaching 3 million in a day, with a peak simultaneous operation of 380,000; supporting this scale is a single cluster with approximately 160 nodes, 30,000 CPU cores, and 250TB of memory.DSec prepares four types of backends for four categories of tasks: FnCall, Container, MicroVM, and Full VM, with the training side called through a unified Python SDK libdsec. The scheduling chain includes IAM, API Server, scheduling engine, node Edge, network proxy Aether, and components within the sandbox Chronus. The environment is divided into three layers of read-only images: base image, workspace, and toolkit, which are used in combination at startup. Runtime data from the paper shows that the actual read ratios of Python, Java, and C++ container images are approximately 6.0%, 9.2%, and 8.7%, respectively.Starting from DeepSeek-V4.1, the Agent loop has been moved to the DSec worker container, no longer bound to the GPU Pod lifecycle. The security section disclosed reward hacking during training, including actions such as overwriting system files, swapping file data blocks, scanning networks, and triggering kernel crashes. Defensive measures include AppArmor and eBPF-based network filtering, but reports indicate that these measures do not completely resolve the issues.

SharpLink CEO Joseph Chalom stated that as AI agents integrate with stablecoins, tokenization of real-world assets, and DeFi, the global financial services industry will face a revenue redistribution of over $1 trillion annually by 2030, potentially reaching $4 trillion by 2035.Chalom indicated that AI agents will become the automation layer of the new financial system, capable of continuously managing investors' financial activities, including finding lower banking, trading, and borrowing costs, optimizing savings returns, constructing portfolios, dynamic rebalancing, and managing loans and credit card debt. He anticipates that by 2030, AI agents could save investors about $350 billion annually by reducing fees, with this figure increasing to $1.4 trillion by 2035, equivalent to eliminating nearly a quarter of the costs in the global financial industry.Stablecoins, tokenized real-world assets, and DeFi will provide AI agents with 24/7 programmable financial infrastructure, enabling agents to view asset ownership, prices, collateral requirements, and lending opportunities within the same blockchain environment, and autonomously complete asset transfers, collateralization, lending, and settlement. He also mentioned that financial institutions including Visa, Mastercard, Stripe, PayPal, Circle, Tether, Robinhood, Coinbase, Binance, as well as JPMorgan, Citigroup, and BlackRock are competing for the infrastructure and user entry points of the AI agent financial ecosystem. Whoever controls the infrastructure and agents may capture the value generated when agents trade on behalf of clients.Additionally, Chalom pointed out that the infrastructure such as the x402 machine-to-machine stablecoin payment standard launched by Coinbase and Ethereum's ERC-8004 agent identity protocol is forming a new open agent economy. More than 10,000 AI agents have completed registration within 10 weeks of the ERC-8004 going live.

Bitget has upgraded its AI assistant GetAgent to version 2.0, expanding the scope of AI research from the cryptocurrency market to include U.S. stocks, CFDs, and on-chain data scenarios, forming a cross-market research framework that covers various assets such as cryptocurrency, U.S. stocks, gold, and crude oil. Users can complete market inquiries, opportunity screening, position diagnostics, and trade reviews through natural language, and combine relevant market data, events, and fundamental information to obtain structured analysis that includes key price levels, participation conditions, risk boundaries, and conditions for judgment failure.In terms of functional interaction, this upgrade further enhances GetAgent's continuous research capabilities. Users can set up regular reports, event reminders, and continuously track market changes based on specific conditions, extending AI from one-time Q&A to ongoing market research and dynamic tracking, transforming fragmented market observations into coherent trading plans.As an important part of Bitget's Agent-native strategy, GetAgent 2.0 is evolving from an information inquiry tool to a continuous research assistant. As UEX covers more cryptocurrency assets and traditional financial markets, GetAgent will also serve as the corresponding intelligent research layer, providing users with cross-asset analysis, regular reports, and continuous market tracking capabilities.

According to official social media news, Huobi HTX will hold a themed live broadcast titled "AI Agent Places Orders for You: Do You Dare to Hand Over Your Position?" today at 8 PM. At that time, 0xMoon and Dr. Dinosaur will have a dialogue with Yang Shao and Amber, discussing the application of AI Agent in trading decisions, strategy execution, and position management, exploring key issues such as the reliability of automated trading, account authorization, and fund security.

According to CCTV Finance, the global trade volume of AI-related products will reach 4.18 trillion USD by 2025, a year-on-year increase of 21.9%. With the rapid evolution of artificial intelligence and Agent capabilities, a batch of "portable" AI hardware is accelerating its entry into the consumer market.At the exhibition hall of the Yunqi Conference in Hangzhou, Zhejiang, many AI hardware products, from AI glasses to user-assembled AI DIY devices, no longer solely rely on smartphones. Instead, they synchronize voice, image, and other information around the user in real-time to the Agent, which autonomously breaks down tasks and completes execution. Global investment in AI infrastructure continues to increase, and Chinese technology companies are also making the latest arrangements around AI hardware, large models, and underlying computing power.

The ongoing X-Agent AI MCP Hackathon has received nearly 40 projects, with some public works turning work decisions, capital management, and inter-machine trading into runnable products. X-Agent introduces three cases: BountyProof checks whether tasks are open, claimed, have relevant submissions, and the authenticity of rewards before the Agent accepts GitHub bounty tasks, with the core question being "Is this work worth doing?"Abstain empowers the Agent with the "do not trade" capability, returning execution, abandonment, or no trade based on preset rules before order execution. sumplus helps the Agent choose suitable model service solutions based on task context, output scale, model capability, and invocation costs, making the Agent an autonomous buyer of models, computing power, data, and API services.The three projects correspond to the foundational economic behaviors of the Agent: accepting work, utilizing funds, and purchasing services, pointing to work, capital, and trading primitives. X-Agent believes that a true Agent economy requires a complete cycle of "building, deploying, operating, discovering, invoking, paying, earning revenue, and distributing," positioning itself as the application layer of the Agent economy. This hackathon is still in the review stage, and the mentioned projects are only for public case reference, not representing shortlisted or award results.

Coinbase developer account announced that stocks can now be traded through Coinbase for Agents. This tool previously supported cryptocurrency and derivatives trading, and now adds over 6000 stocks.For analysis or data related to these stocks, x402 payment allows agents to use the USDC balance in their Coinbase accounts to pay for real-time market data fees during task execution, without the need for a subscription service.

Cardano has joined the official x402 software development kit, allowing developers to enable applications or AI agents to use ADA or Cardano network tokens to pay for online services. x402 transforms the basic idle "402 Payment Required" response in web pages into a checkout process built into internet requests: the service provider returns the price and payment instructions, the agent signs the payment, and after transaction verification, the required data or computing power can be obtained.This means that AI agents can purchase individual datasets on demand when preparing reports, without the need for manual account registration, entering credit card information, or subscribing to monthly fees. x402 was created by Coinbase in 2025 and subsequently contributed to an organization under the Linux Foundation, with members including Visa, Mastercard, Stripe, Google, and Amazon Web Services. Solana, XRP Ledger, and several Ethereum-compatible networks have previously supported this standard.Engineers from the Cardano Foundation have built client and server software for payment requests based on the specifications passed in June, as well as a facilitator responsible for verifying and submitting transactions. The first version supports TypeScript, with Python support planned for later release. Facilitator documentation shows that it has completed a real transaction on the Cardano pre-production network, but it has not yet run on the mainnet, nor has it demonstrated scenarios where agents use ADA to pay for commercial services on a large scale.