Chip startup Tenstorrent has closed a $1.5 billion funding round that values the company at $6.2 billion, more than double its $2.6 billion valuation from late 2024. The round was led by new firm Trace Capital and Innovation X, with Cadence Ventures, Eclipse, Fidelity, SanDisk, and T. Rowe Price also participating. CEO Jim Keller said the company chose to stay independent and raise capital after demonstrating its Galaxy servers could run DeepSeek models, with funds earmarked for chips tailored to open-weight AI and video generation workloads.
The Department of Homeland Security has ordered all civilian federal agencies, including Treasury, State and Justice, to remove DeepSeek's AI models and any technology built using them. The order became effective at the end of July but was only publicly disclosed this week; Defense Department systems are exempt from the ban.
The Department of Homeland Security has issued a notice in the Federal Register directing all federal civilian agencies to stop using DeepSeek products and strip out any that are already installed on their IT systems. DHS cited risks that DeepSeek could transfer sensitive data on government employees and vendors to the Chinese government, and warned its models may contain malware or other security flaws. The order also bars DeepSeek from participating in federal contracts.
The Department of Homeland Security has directed federal civilian agencies to remove DeepSeek technology from their IT systems and barred any future purchases of the Chinese company's products or services. The order also covers products made by DeepSeek affiliates and third-party tools built using DeepSeek's implementations.
DeepSeek has released V4.1 Flash 1017 (Max), a multimodal model accepting text and image input and producing text output, with a 1 million token context window. The model scored 45 on the Artificial Analysis Intelligence Index, well above the comparable median of 19, while running at 230 tokens per second.
DeepSeek has introduced two new tools, a Search API and a Fetch API, giving developers programmatic access to live web content for AI agents and retrieval-augmented generation applications. The Search API returns structured results from DeepSeek's own search engine, while the Fetch API pulls a single webpage's content and converts it to Markdown; both use the same bearer-token authentication as DeepSeek's existing API and are priced at $0.75 per 1,000 successful requests.
DeepSeek has shipped V4.1 Flash 1017, dated October 17, 2026, as a snapshot of its V4.1 Flash model available via the deepseek-flash API. The sparse mixture-of-experts model uses a new Causal Encoder-Decoder architecture, activating 8B parameters on input and 16B on output from a 552B-parameter backbone, with native image understanding trained jointly with text from the start.
DeepSeek has launched V4.1-Flash-1017, a minor version update to its V4.1-Flash model that the company says improves reasoning, coding, agentic task handling, and visual benchmark scores compared to the prior version. The update also shortens task execution time and output length. DeepSeek is also releasing deepseek-recipe, a Rust-based toolkit with Python bindings for encoding prompts and parsing model outputs.
China has rapidly expanded its scientific output over the past five decades and is now positioned to lead global research and development, according to Nature's reporting. A growing number of China-trained AI graduates are choosing to remain in the country rather than move abroad, with Chinese researchers now dominating global AI-related publications and firms like DeepSeek and MoonshotAI demonstrating competitive AI capabilities.
A growing number of Chinese computer science graduates are opting to remain in China for AI research rather than pursue PhDs in the United States, as the capability gap between the two countries narrows. China now produces more AI researchers than the US, and firms like DeepSeek and Moonshot AI are building models that rival top US systems at much lower cost.
Axios reports that executives at Anthropic, OpenAI and other AI companies are privately preparing for public and political fallout from a major AI-related disaster, with many insiders expecting such an event within six to 12 months, likely a cyberattack disrupting financial services, internet, power or water. Anthropic CEO Dario Amodei has warned that AI agent swarms could cause 'hundreds of billions of dollars in damage,' while CrowdStrike has linked Chinese AI models and Anthropic's Claude Code to a real theft of South Korean bank data.
CrowdStrike reported that a Chinese-speaking hacker used the ARTEX AI penetration testing suite alongside Claude Code agents to attack major South Korean banks including Shinhan, KB Kookmin, and Hana, exposing customer personal and credit card data and causing outages at some institutions. Researchers found open directories containing Claude session histories and ARTEX configuration files that tied the attacker's infrastructure to the breaches and revealed identifying details from a résumé the attacker had generated with the same tools.
GoKawiil