Topic

DeepSeek

5 written stories · 12 tracked

Tech3 outletsover 32h

The NSA, CISA, and FBI issue a joint advisory warning that Chinese AI companies, including DeepSeek, are conducting "industrial-scale" distillation campaigns (Reuters)

U.S. intelligence agencies—the NSA, FBI and CISA—issued a joint advisory warning that six Chinese artificial-intelligence companies are conducting industrial-scale model-distillation campaigns that extract capabilities from American frontier AI models. The firms—DeepSeek, Moonshot AI, Alibaba, MiniMax, StepFun and Z.AI—have reportedly harvested billions of tokens from Anthropic’s Claude, OpenAI’s GPT, Google’s Gemini and xAI’s Grok to train their own systems.

TechMeme·The Next Web·Engadget

Tech3 outletsover 28h

Sources: DeepSeek is in preliminary talks with investors about raising new funds at a ~$71B valuation, after raising ~$7B at a ~$52B valuation at the end of May (Zijing Wu/Financial Times)

DeepSeek is preparing an initial public offering in China, with plans that could be filed as early as this year and potentially debut in 2027. The company is seeking new funding at a valuation of roughly $71 billion, following its recent capital raise. This valuation target reflects its latest fundraising efforts.

TechMeme·TechCrunch·The Next Web

Tech3 outletsover 4h

Reuters: DeepSeek is developing its own AI chips

DeepSeek, a Chinese artificial-intelligence firm, is working on its own AI processor. The effort is aimed at circumventing U.S. export restrictions that limit access to foreign-made chips for advanced models. By creating a proprietary chip, DeepSeek hopes to maintain development momentum despite the curbs.

Engadget·The Next Web·Ars Technica

Tech8 outlets

Three reasons why DeepSeek’s new model V4 matters

DeepSeek has previewed its new flagship models, V4 Pro and V4 Flash, positioning them as highly efficient, million-token context AI systems. The company says V4 Pro delivers performance within three to six months of the current state-of-the-art models, offering near-state-of-the-art intelligence and world-class reasoning at roughly one-sixth the cost of comparable leading models.

Hacker News·TechMeme·The Next Web·The Verge·Engadget+3

Tech6 outletsover 11h

Anthropic says DeepSeek, Moonshot, and MiniMax used 24,000 fake accounts to rip off Claude

Anthropic says three Chinese AI labs—DeepSeek, Moonshot and MiniMax—created roughly 24,000 fraudulent accounts and prompted its Claude model more than 16 million times. Using those interactions, the labs employed a distillation process to extract Claude’s capabilities and train their own products, violating Anthropic’s terms of service.

VentureBeat·TechMeme·Hacker News·TechCrunch·The Verge+1