The NSA, CISA, and FBI jointly identified six Chinese AI companies—DeepSeek, Moonshot AI, Alibaba, MiniMax, StepFun, and Z.AI—as running large-scale campaigns since late 2024 to extract the underlying capabilities of leading US AI systems like Claude, GPT, Gemini, and Grok. The agencies said these firms likely operated with the Chinese government's awareness and used tactics such as mass fake-account API abuse and prompt injection to expose hidden reasoning processes.
arstechnica.com
· 2026-09-09
U.S. security agencies have publicly accused six Chinese artificial intelligence companies of systematically using American AI models to train their own systems through a technique known as distillation. The practice allows a smaller model to learn from the outputs of a larger, more advanced one, effectively transferring capabilities without the original research investment.
wsj.com
· 2026-09-09
Anthropic has been ranked No. 1 on Fast Company's Best Workplaces for Innovators 2026 list, with recognition tied partly to Claude Code, a tool originally built internally for engineers to test AI models and assist with coding. The tool spread organically inside the company before Anthropic released it publicly in early 2025, later evolving into the broader Claude Cowork product launched in early 2026.
fastcompany.com
· 2026-09-08
Four leading AI chatbot services experienced service disruptions within the same window, leaving users of ChatGPT, Gemini, Claude and Grok unable to access the tools normally. The near-simultaneous nature of the outages was unusual since these platforms run on separate infrastructure operated by different companies.
tech.slashdot.org
· 2026-09-03
Meta introduced a new pricing tier for its Muse Spark model, built for coding and other agentic tasks, that slashes token costs by roughly 95% for developers who agree to let their prompts and outputs be used to train future versions. Standard pricing runs $1.25 per million input tokens and $4.25 per million output tokens, compared to just 10 cents and 20 cents respectively under the data-sharing plan. The move follows a scrapped internal program at Meta that tracked employee computer activity for training data, which was shelved in June after backlash.
techcrunch.com
· 2026-09-03
Nvidia has agreed to acquire Hugging Face, the AI development platform known for hosting open-weight models, in a deal valued at $13 billion. The acquisition strengthens Nvidia's position in the open-source AI ecosystem beyond its core chip business.
wsj.com
· 2026-09-03
Nvidia has agreed to purchase AI model repository Hugging Face for $13 billion, a huge jump from the $7 billion valuation the startup rejected from Nvidia just last year to stay independent. The chipmaker says the acquisition will accelerate adoption of open-weight AI models, which let developers download and customize systems rather than rely on proprietary offerings from firms like OpenAI and Anthropic. The deal, expected to close by 2027, surpasses Nvidia's 2020 purchase of Mellanox as its biggest acquisition to date.
arstechnica.com
· 2026-09-03
Nvidia announced a deal to purchase Hugging Face, the widely used open-source hub for AI models, datasets and developer tools, for roughly $12.93 billion. CEO Jensen Huang said the platform will stay open to all developers regardless of hardware or cloud provider, and that Nvidia compute won't be a requirement to use it. The acquisition follows weeks of reports that Hugging Face was exploring a sale near this valuation.
theverge.com
· 2026-09-03
At the G20 Innovation Ministerial in Chapel Hill, Commerce Secretary Howard Lutnick argued alongside Anthropic's Tom Brown that AI growth requires more data centers, launching a US effort to persuade global leaders to support them. The pitch comes even as roughly 3,400 data centers already operate in the US with 2,000 more planned, drawing bipartisan opposition over energy demands, water use and rising utility costs. The summit itself drew about 100 protesters objecting to appearances by Sam Altman, Elon Musk and Lutnick.
cnet.com
· 2026-09-02
Russian startup Mostik has developed a method allowing AI models to exchange information through their internal weight values rather than generated text, effectively letting a smaller model absorb capabilities from a larger one. The team demonstrated this by linking a 753-billion-parameter GLM-5.2 model with a 4-billion-parameter Qwen-3.5 model, producing a hybrid system that runs at one-twentieth the cost of the full-size model while performing roughly midway between the two in capability. Mostik also used a related technique to build a model that has topped the ARC-AGI 3 benchmark, though details remain undisclosed while the contest is ongoing.
wired.com
· 2026-09-02
Protect Democracy, a nonpartisan nonprofit, has filed a lawsuit against four federal agencies demanding disclosure of the confidential process the Trump administration uses to vet frontier AI models before release. The group says virtually no details have been shared publicly or with Congress about which companies participate, how they are chosen, or what legal authority underpins the reviews, and alleges OpenAI has struck a private deal limiting access to its top models to government-approved partners. The lawsuit seeks a court order compelling release of unclassified records by September 30 and an injunction against further withholding.
arstechnica.com
· 2026-09-02
OpenAI announced that its forthcoming Astra AI model is the first to cross the company's 'Critical' cybersecurity capability threshold, meaning it can discover unknown vulnerabilities and exploit them autonomously without human step-by-step direction. The company says Astra will still launch soon, but access to its most sensitive cybersecurity abilities will be restricted, with further safety details to come in a system card at release.
cnbc.com
· 2026-09-01
GoKawiil