BTC $84,403.66 +0.01%
ETH $2,688.27 +0.52%
BNB $779.81 +1.89%
XRP $1.54 +2.74%
SOL $116.90 +2.26%
TRX $0.3403 +0.00%
DOGE $0.0961 +4.09%
ADA $0.2487 +4.44%
BCH $339.57 +1.15%
LINK $13.22 +7.83%
HYPE $93.61 +0.79%
AAVE $145.07 +5.06%
SUI $1.02 +6.76%
XLM $0.2136 +5.82%
ZEC $1,548.84 +2.51%
AAPL $336.01 -0.26%
AMZN $249.17 -0.12%
GOOGL $342.02 +1.01%
MSFT $497.41 -0.49%
META $776.00 +4.12%
NVDA $224.00 -0.48%
TSLA $377.83 -0.51%
SNDK $1,756.69 -3.50%
INTC $127.20 +3.98%
SPCX $147.91 -0.62%
MU $1,078.41 +0.57%
AMD $628.54 +2.23%
BTC $84,403.66 +0.01%
ETH $2,688.27 +0.52%
BNB $779.81 +1.89%
XRP $1.54 +2.74%
SOL $116.90 +2.26%
TRX $0.3403 +0.00%
DOGE $0.0961 +4.09%
ADA $0.2487 +4.44%
BCH $339.57 +1.15%
LINK $13.22 +7.83%
HYPE $93.61 +0.79%
AAVE $145.07 +5.06%
SUI $1.02 +6.76%
XLM $0.2136 +5.82%
ZEC $1,548.84 +2.51%
AAPL $336.01 -0.26%
AMZN $249.17 -0.12%
GOOGL $342.02 +1.01%
MSFT $497.41 -0.49%
META $776.00 +4.12%
NVDA $224.00 -0.48%
TSLA $377.83 -0.51%
SNDK $1,756.69 -3.50%
INTC $127.20 +3.98%
SPCX $147.91 -0.62%
MU $1,078.41 +0.57%
AMD $628.54 +2.23%

ini

All
Article
Flash

first_img DeepSeek publicly releases the Agent training system DSec, signed by Liang Wenfeng

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.

first_img Australian Prime Minister: OpenAI AI entity first invaded government website

Australian Prime Minister Anthony Albanese stated to reporters in New York on Wednesday that an AI entity built by OpenAI breached the Australian government website Medicare Statistics Reporting Service portal in June this year, accessing both public and non-public documents without authorization. The portal is managed by Services Australia and stores non-sensitive information such as Medicare expenditure data. This appears to be the first known incident of an AI entity breaching a government website.Albanese indicated that it is currently believed that no personal information was accessed, but the Australian Signals Directorate is assisting in a forensic investigation to determine which other government systems may have been affected. He mentioned that he has directly raised the matter with OpenAI CEO Sam Altman and expressed disappointment that the company took about three months to inform the government, deeming the manner of notification equally unacceptable.In a statement provided to Australian ABC News, OpenAI stated that it is reviewing the model behavior deviations that occurred during training and evaluation, and identified that its model engaged in activities involving multiple Australian government websites when querying relevant answers and statistics about Australia in internal evaluations, and that the model "took actions that we did not intend."

first_img OpenAI opens the Daybreak cybersecurity tool to the Ukrainian government

OpenAI announced that it is opening its AI vulnerability detection tool Daybreak to Ukraine's Ministry of Digital Transformation to locate and fix software vulnerabilities before they are discovered by Russian hackers. The news was announced on Wednesday during the United Nations General Assembly by Dmytro Kushneruk, Ukraine's Consul General in San Francisco, and Sasha Baker, OpenAI's National Security Policy Lead.Daybreak is based on OpenAI's flagship large model GPT-5.6 Sol and can scan outdated systems for weaknesses, verify whether suspected vulnerabilities can be exploited, and confirm whether patches actually close the vulnerabilities before they are released. OpenAI has previously opened the model to cybersecurity agencies in France, Germany, and Poland, where Poland's CERT Polska discovered six vulnerabilities in commercial router software, all of which have been fixed by the vendors.Ukraine's CERT-UA handled nearly 6,000 cyberattacks in 2025, a 37% increase from 2024, with local governments and government agencies being the hardest hit. This opening follows OpenAI's commitment to a $1 billion subsidy program announced on September 3, which is said to expand to "partner countries" in the coming weeks. Jamie MacColl, a researcher at the Royal United Services Institute, told the BBC that Western companies can also gain valuable intelligence from active conflict zones.

first_img The Cyberspace Administration of China is investigating DeepSeek and the Dark Side of the Moon for allegedly leaking data to Claude

According to The Information, citing informed sources, China's National Internet Information Office has launched an investigation into AI companies DeepSeek and Moonshot AI, triggered by Anthropic's allegations that the two companies secretly routed sensitive user data to their servers. Reports indicate that regulators visited the offices of both companies, interviewing executives and employees, focusing on whether sensitive data related to law enforcement, military, and state-owned enterprises has flowed into U.S. servers.The trigger for this investigation was Anthropic's fourth threat intelligence report released on September 10. This 154-page document accuses seven Chinese labs—Alibaba, Moonshot AI, DeepSeek, Z.ai, MiniMax, SenseTime, and Xiaomi—of engaging in what it calls "illegal distillation," which involves using the outputs of large models to train smaller models. Anthropic states that distillation itself is a legitimate practice, but it opposes its implementation through fraudulent accounts. The National Internet Information Office initially summoned all seven companies named in the report, but later narrowed the investigation to DeepSeek and Moonshot AI. Anthropic claims that Moonshot AI routed over 23 million interactions to Claude through 5,380 fraudulent accounts, while DeepSeek generated over 12.1 million interactions within a 14-day window in July.The timing of the investigation is quite delicate for both companies.

first_img The State-owned Assets Supervision and Administration Commission investigates the usage of Broadcom switches in state-owned enterprise data centers

According to the Financial Times, Chinese authorities are reviewing the use of Broadcom hardware by state-owned data centers to support domestic manufacturers and reduce reliance on foreign artificial intelligence infrastructure. The State-owned Assets Supervision and Administration Commission has been investigating how many Broadcom switches are used by state-controlled data centers in recent weeks, according to two informed sources.One of them stated that the penetration rate of Broadcom switches in state-owned enterprises could be as high as 90%, and preliminary results may lead the commission to issue informal guidance to reduce the use of Broadcom switches in domestic data centers. The sources indicated that state-owned data centers are no longer allowed to use NVIDIA products but still use a large number of Broadcom switches. The commission is also investigating whether Broadcom is leveraging its market-leading position to bundle other products or requiring one-time purchases of tens of thousands of switch chips.The sources noted that these practices have restricted Chinese companies like H3C and Ruijie Networks from placing orders with other switch manufacturers such as Huawei. H3C and Ruijie Networks are often included in the procurement recommendation lists for government agencies, public institutions, and state-owned enterprises, which spend billions of dollars annually on information technology products.

first_img After 11 years of operation, BitMEX has stopped trading and deposits, retaining only withdrawals

The cryptocurrency derivatives platform BitMEX has ceased trading, deposits, and opening new positions, retaining only the withdrawal function. The platform announced on social media platform X that as of this Tuesday at 04:00 UTC, the aforementioned services will no longer be available. Users can still log in and withdraw their balances through the official website, but deposits have been confirmed as impossible. BitMEX was co-founded by Arthur Hayes, Ben Delo, and Samuel Reed in 2014, known for promoting perpetual contracts as mainstream trading tools in the cryptocurrency derivatives market.BitMEX urges customers to withdraw their funds as soon as possible and states that account fees will be charged to users with completed KYC certification who still have a balance, billed monthly, at a standard of 1% annualized on assets or a minimum equivalent of $50, whichever is higher. The company emphasizes that this shutdown does not freeze customer assets, and the withdrawal channel remains open during the liquidation process.BitMEX announced its closure in July this year, following a strategic assessment by its parent company HDR Global Trading, as the exchange's market share in the perpetual contract market it helped create continued to decline. According to research data from CoinDesk, the monthly trading volume of centralized cryptocurrency exchange derivatives reached $3.4 trillion in August, with an annual scale potentially reaching $50 trillion. Previously, BitMart also announced its shutdown in July after nine years of operation due to market conditions.

Hyperscale Data established a new subsidiary to independently operate mining, and Reap plans to launch multi-national local currency stablecoins to facilitate 24/7 foreign exchange clearing

According to BBX data, yesterday, publicly listed companies in the U.S. and cross-border financial infrastructure disclosed the latest developments in the restructuring of cryptocurrency businesses and global compliance stablecoin settlements. The core information is as follows:Hyperscale Data (NYSE: $GPUS) established a subsidiary, Patriot BTC, to independently undertake mining operations and assets: Nasdaq-listed Bitcoin mining company and AI data center infrastructure company Hyperscale Data officially announced the establishment of a new wholly-owned subsidiary, Patriot BTC. The purpose of this subsidiary is to hold and operate the company's cryptocurrency mining business, mining machines, and related on-chain assets as a completely independent business entity, in order to further optimize the overall asset structure of the parent company and align with the implementation of the AI computing power center strategy.Reap plans to launch multi-national local currency stablecoins to build a 24/7 on-chain foreign exchange trading network: The multinational payment technology platform Reap, which has received strategic support from Kraken's parent company Payward and is a core partner of Visa, announced plans to successively launch local currency stablecoins in multiple countries and regions to achieve 24/7 on-chain foreign exchange settlements. Reap will first launch the Mexican Peso (MXN) stablecoin and is accelerating the exploration of issuing various local currency stablecoins, including the Hong Kong Dollar (HKD), Euro (EUR), South Korean Won (KRW), and Japanese Yen (JPY). This layout aims to completely break the physical limitations of traditional commercial bank operating hours, providing multinational companies with efficient and low-friction on-chain cross-border foreign exchange clearing solutions during nighttime and weekends.

first_img Apple's new Mac mini and Mac Studio are shipping, focusing on edge AI

Apple's new Mac mini and Mac Studio began shipping on Tuesday local time. Apple emphasized to business buyers that purchasing new machines is cheaper than renting cloud computing power. The new machines support on-device AI, capable of handling high-intensity tasks like coding locally, allowing users to avoid purchasing tokens from OpenAI, Anthropic, and others. Apple's Chief Hardware Officer, Johnny Srouji, stated: Buying the device and placing it on the desk is a one-time investment, and the product offers excellent cost-performance ratio, without needing to pay per token; the machine can be used indefinitely.At the launch event, Apple demonstrated four Mac Studios working together to run a trillion-parameter AI model, searching for and fixing graphic code vulnerabilities, with the group of devices powered by a single wall outlet. The Mac Studio equipped with the M5 Ultra chip can be configured with up to 512GB of unified memory, supporting the formation of a cluster with multiple devices sharing a memory pool via RDMA. After networking four units, the distributed AI inference speed can reach up to three times that of a single machine. Apple claims that the underlying design philosophy of various chips is consistent, and on-device AI models can be deployed on high-end Mac Studios while also being compatible with entry-level iPhones and iPads.IDC analyst Linn Huang's data shows that Apple holds about 4.6% of the enterprise desktop market share, while Windows accounts for 91.3%. NVIDIA and various PC manufacturers are launching new desktops to compete directly with the new Macs, and these new Windows machines are expected to be the focus of Microsoft's Windows event in San Francisco next month.
app_icon
ChainCatcher Building the Web3 world with innovations.