mode

According to The Information, Anthropic and OpenAI have both experienced security incidents, raising market concerns about the safety of AI models themselves.Currently, Anthropic is investigating potential unauthorized access to its Claude Mythos model by users. Almost simultaneously, OpenAI was also reported to have accidentally exposed several unreleased models in its Codex application. Industry opinions indicate that these vulnerability incidents have intensified scrutiny on the security governance capabilities of AI companies and reflect that the security systems of current AI technology still need improvement as it rapidly develops. Analysts believe that even AI model providers that focus on cybersecurity capabilities face significant security challenges themselves. As AI is gradually used to defend against cyberattacks, issues related to platform security and access control have also become critical risk points.

AI laboratory Sooth Labs, founded by former Meta employees, is raising approximately $50 million in funding, led by Felicis Ventures, with a post-money valuation of about $335 million.The company plans to develop AI models to predict the likelihood of specific geopolitical events and market events, providing related capabilities to enterprises. Sooth Labs has also received support from Yann LeCun and Google Chief Scientist Jeff Dean, with Meta Chief Technology Officer Andrew Bosworth serving as an advisor.

According to news from Shanghai's Cyberspace Administration, Tesla's in-car voice large model service completed its filing on April 20. The Science and Technology Innovation Board Daily learned from informed sources that Tesla's in-car voice service will integrate the Doubao large model. It is worth mentioning that the "Tesla In-Car Voice Assistant Usage Terms" on Tesla China's official website previously indicated that the Tesla Model Y L will be equipped with both the Doubao large model and the DeepSeek model, both of which are accessed through the Volcano Engine.Among them, the Doubao large model will handle voice command functions, such as navigation settings, media playback control, air conditioning temperature adjustment, etc., and will also have a vehicle owner's manual query function; the DeepSeek model provides AI interactive services.

The 0G Foundation announced a partnership with Alibaba Cloud to launch on-chain access to the Qianwen large model for AI agents.This move shifts artificial intelligence infrastructure from centralized, API-based systems to programmable, tokenized access specifically designed for autonomous agents. Through this collaboration, 0G allows developers to directly embed access to Qianwen into their infrastructure via a token-based mechanism. This eliminates the need for traditional account management and enables on-demand programmable access to AI computation. Agents running on 0G can query the Qianwen large model to generate responses, process information, or execute tasks in real-time.Michael Heinrich, CEO and co-founder of 0G Labs, stated, "From a macro perspective, inference runs on Qianwen, and verification runs on 0G—this builds a more complete computational and trust foundation for autonomous AI systems. The integration of 0G with Alibaba Cloud's Qianwen marks a milestone, achieving the fusion of top-tier intelligence and trusted infrastructure."

ChainCatcher reported live that Fan Wenzhong, Executive Director of the Chinese Financial Society and former Chairman of Beijing Financial Holdings Group, delivered a keynote speech at the 2026 Hong Kong Web3 Carnival. He pointed out that we are currently at a historical turning point similar to the transition from pawnshops to modern banking, where AI is an advanced productive force in the physical world, and Web3 represents a new type of production relationship in the digital world. The integration of the two will give rise to a Decentralized Agent Economy (DAE).He analyzed that AI Agents have execution capabilities but lack independent identity, accounts, and trust mechanisms, while Web3 addresses these pain points through smart contracts, on-chain identities, and programmable currencies; conversely, AI Agents significantly lower the barriers to using Web3. He proposed that DAE has four major characteristics: agent sovereignty and 24/7 operation, high-frequency atomic exchanges, a collaboration mechanism based on technological trust rather than moral or legal frameworks, and the organizational evolution from companies to DAOs and then to Decentralized AI Companies (DACs).He warned that this transformation will impact the labor market, with white-collar workers being replaced before blue-collar workers. AI quantitative funds have already achieved returns far exceeding those of retail investors, and he suggested that the government proactively advance social security reforms. Finally, he recommended that Hong Kong develop a self-controllable high-performance public chain, pilot limited digital persona registrations, attract Web3+AI composite talents, establish a digital finance industry fund, and promote collaborative innovation between the Hong Kong dollar stablecoin and digital renminbi.

According to Cointelegraph, the Monetary Authority of Singapore (MAS) has urged banks to strengthen their cybersecurity defenses due to heightened regulatory concerns arising from the spread of the Mythos AI model under Anthropic in the Asia region.

Ju.com, in collaboration with Nivex, officially launched the Web 4.0 Intelligent Intent AI Model ------ OG Agent. This model aims to end traditional "physical trading" through an AI intelligent agent cluster, achieving the return of computational sovereignty for individual investors, while marking the official entry of decentralized finance into the "AI Sovereignty" era.Core technological breakthroughs:Extreme millisecond-level execution: Based on a distributed edge computing architecture, OG Agent achieves a response speed of 10 microseconds, far exceeding the limits of traditional manual operations, capturing global market liquidity premiums in real-time.Intent-driven configuration: Users only need to set profit targets, and the AI cluster can automatically analyze the optimal path and complete intelligent configuration, thoroughly simplifying complex financial operations.Secure sovereign isolation: By adopting the separation of "execution rights" and "ownership" technology, it ensures that the AI operates only within restricted authorization, allowing users to maintain absolute control over their assets.Deflationary economic model: The ecological token $OG is linked to real returns, achieving long-term value drive through a buyback and destruction mechanism.Global partner recruitment has begun:OG Agent has now launched a global "Computational Power Pioneer" recruitment program. The first batch of participants can directly unlock VIP 3 advanced permissions and obtain a $OG original asset confirmation seat. In addition, partners will deeply share in the protocol's revenue pool dividends and enjoy up to 90% institutional-level rebate rights.

The Web3 asset data platform RootData has outlined Upbit's business partners, which are relatively limited and mainly focused on areas such as blockchain infrastructure and compliance services. This is closely related to South Korea's unique and strict regulatory framework: under a system environment centered on real-name systems, VASP regulation, and anti-money laundering, exchanges are required to take greater responsibility for the flow of funds and user identities, making Upbit more inclined to develop its own systems and build localized capabilities rather than extensively introducing external partners.From the disclosed structure, on the technical side, infrastructure capabilities are strengthened through Lambda256, Aptos, and others; on the compliance level, service providers such as VerifyVASP, Chainalysis, and Crystal Intelligence are introduced; and supplementary funding and liquidity are provided through Circle and Naver Pay. Additionally, Upbit quickly built trading capabilities relying on its cooperation with Bittrex since its launch in 2017.On the branding side, Upbit has explored NFT entertainment collaborations with Hybe and has laid out plans in esports (LCK) and signed top players like Faker, while also becoming a primary partner of the Korean Sports Council. However, as the popularity of NFTs and the metaverse has declined, the influence of related entertainment collaborations has weakened. Related compilation: [Upbit Web3 Partner Network Compilation (continuously updated)](https://cn

Coinbase and Binance are seeking access to the Anthropic AI model Mythos to enhance their ability to respond to AI-driven cyber threats.Reports indicate that Coinbase has begun discussions with Anthropic, with its Chief Security Officer Philip Martin stating that the model is expected to accelerate digital threat identification and defense capabilities. Meanwhile, Binance and Fireblocks are also evaluating the potential applications of the model in the field of cybersecurity.

OpenAI announced the launch of the GPT-5.4-Cyber model, which is fine-tuned from GPT-5.4 for cybersecurity scenarios, reducing the refusal limits on security-related requests under specific conditions to support specialized operations such as binary reverse engineering.According to the introduction, the model is only available to vetted security vendors, corporate security teams, and researchers through the "Trusted Access for Cybersecurity" (TAC) program. OpenAI stated that the relevant layered mechanism will limit the scope of model usage and impose additional constraints on access in low visibility scenarios.

Bitmine Chairman Tom Lee posted a video on the X platform showing that Ethereum co-founder Joe Lubin stated in an interview that after communicating with Michael Saylor, he found that the Ethereum treasury strategy could perform better than Strategy.Lubin pointed out that using Bitmine's model as an example, by allocating 100% of funds to ETH and participating fully in staking, compound growth can be achieved from day one. Compared to simply holding Bitcoin, Ethereum can also drive the balance sheet from a "holding" type to a "revenue + allocation" type through staking rewards and a multi-layer return mechanism within the ecosystem.

Meta launched its new AI model "Muse Spark" on April 8. The model was developed by Meta's "Super Intelligence Lab" and led by former Scale AI CEO Alexandr Wang. Meta stated that the new model achieves performance close to the previous Llama 4 medium model with lower computational resources by restructuring the AI infrastructure.

According to official news, AI NFT has announced the official integration of Google login, allowing users to choose a convenient Google account or decentralized wallet connection based on their preferences. New users do not need to go through a complicated registration process; they can log in to enjoy a million points reward and directly access several leading global AI large models. This upgrade further lowers the usage threshold, enabling users to easily embark on their AI exploration journey in their preferred way.

According to a report by Jinshi, Derek Halpenny, a foreign exchange analyst at Mitsubishi UFJ Bank, stated that the dollar's response to the surge in energy prices has been mild, highlighting that the current foundation of the U.S. economy is weaker compared to the inflation shock in 2022. He pointed out that since the outbreak of the conflict, although oil prices have risen nearly 60%, the dollar index has only strengthened by about 2.5%. The current price shock may have a greater impact on the U.S. economy than expected, undermining business confidence and employment.

Aave founder Stani Kulechov tweeted that Aave respects Chaos Labs' decision to withdraw from its risk management role and thanks them for their contributions over the past years. The smart contracts, asset listings, and network deployments of the Aave protocol are not affected, and both parties will maintain close cooperation during the withdrawal process.Aave currently employs a dual-layer economic risk management model, jointly managed by Chaos Labs and LlamaRisk. Given the scale of the Aave protocol, it will continue to maintain the dual-layer model in the future and additionally add a technical risk layer managed by Aave Labs. Aave Labs will work closely with LlamaRisk to ensure that risk coverage remains uninterrupted during the transition and to support LlamaRisk in increasing its budget and team size.

Recently, the State Taxation Administration and the National Financial Regulatory Administration of China jointly issued a notice on "Further Deepening and Standardizing 'Bank-Tax Interaction' Work," which pointed out the need to deepen the application of bank-tax data.Standardize the content and methods of providing bank-tax data to further resolve the information asymmetry issue between banks, taxes, and enterprises. Encourage both banks and tax authorities in various regions to utilize technologies such as blockchain and privacy computing to legally and compliantly innovate the "bank-tax interaction" model. Banks will continue to optimize credit services, increase financing service supply for honest tax-paying enterprises, and continuously monitor the effectiveness of the "bank-tax interaction" loan model to improve risk management levels.