Flash News

TREE NEWS update: OpenAI CEO Sam Altman said the world should trust American AI companies to develop the technology responsibly, calling for the establishment of a consistent safety requirements framework. He added that a shift toward safety-focused AI development will require new tools.

TREE NEWS update: Alibaba Cloud announced that DeepSeek-V4.1-Flash is now live on its Qwen AI platform, with API services and a Token Plan made available at the same time. The model is accessible through Alibaba Cloud’s platform, and the accompanying API and token subscription plan open alongside the launch. No pricing or performance details were disclosed.

TREE NEWS update: Beijing Futekang Intelligent Robot Technology Co. has been established with registered capital of 50 million yuan (about $7 million), with IAT Automobile Technology, Foxconn Technology Group and Beijing Gongxin Tongda Technology Partnership as shareholders. The company’s legal representative is Xuan Qiwu, and its business scope covers sales of intelligent and industrial robots and manufacturing of special-operation robots.

TREE NEWS update: Anthropic CEO Dario Amodei told CBS’s “Sunday Morning” on September 13 that AI risks are real and the industry has lied to the public about them, while rejecting an outright ban on development. He outlined a three-step plan centered on embedded third-party evaluators, joint industry safety standards and government involvement, saying the technology sits near the bend of an exponential curve. The interview followed his blog post “We Must Pace the Frontier,” backed publicly by Elon Musk and Sam Altman.

TREE NEWS update: The consumer version of Doubao Phone Assistant was released on September 14, built on the Doubao app and developed with phone makers at the system level, covering interaction, task execution, local memory and retrieval. The first phone carrying it goes on sale September 16. The company said the consumer release emphasizes stability and everyday usability over last year’s technical preview.

TREE NEWS update: Anthropic has signed a six-year compute agreement worth $13.7 billion with Rum Group, a company with social-media roots and long-standing ties to the Trump administration, people familiar with the matter said. Rum, founded in 2013, hosts Trump’s Truth Social and streams official White House broadcasts. The deal adds to Anthropic’s cloud agreements with partners including Google, SpaceX and Nscale, which total at least 14,800 megawatts of compute capacity and up to $517 billion in costs over the next decade.

TREE NEWS update: Dong Keyan, deputy director of a department at China’s Ministry of Science and Technology, said brain-inspired intelligence and neuromorphic photonics are key breakthrough directions for next-generation AI, speaking at a brain-inspired intelligence forum held during the 2026 Pujiang Innovation Forum in Shanghai on September 13-14. He cited difficulties in engineering deployment and a shortage of interdisciplinary talent as current challenges, and said the ministry will keep backing basic research and technical work in the field.

TREE NEWS update: The US Senate is scheduled to hold a key vote on Tuesday on crypto market structure legislation, a step that could deliver long-sought regulatory clarity for the industry. Separately, fears over AI swarms have prompted a slowdown in development, with some predicting sharp declines in AI stock prices.

TREE NEWS update: US House Speaker Mike Johnson said he is seeking talks with artificial intelligence platform providers and core experts. The move signals congressional outreach to the AI industry as lawmakers weigh policy toward the sector. No further details, including a timeline or list of participants, were given.

TREE NEWS update: Zhipu has raised $5 billion through a zero-coupon convertible bond with a premium conversion feature, earmarking the proceeds mainly for computing infrastructure and model research and development. The financing was disclosed in a roundup of market developments that also flagged a rise in Fed rate-hike odds to nearly 90% for this week after August core CPI came in at 0.3% month-on-month, above expectations.

TREE NEWS update: Anthropic has told investors it expects to be profitable for a second consecutive quarter, the company said in an update to backers. The AI developer did not disclose profit figures or the period covered. The disclosure signals the company is moving past the heavy losses typical of large-model developers.

TREE NEWS update: Multiple listed and tech companies unveiled major new products and breakthrough technical solutions at the 27th China International Optoelectronic Exposition, pushing optical chip, optical engine, fiber and switching systems to new performance levels for AI computing. An industry executive said the return cycle for AI investment is already visible, with the only uncertainty being which technology path prevails.

TREE NEWS update: King Charles III will meet the heads of several AI companies in Scotland this week, including Nvidia, Google and DeepMind. The audiences mark the British monarch’s latest engagement with the artificial intelligence sector.

TREE NEWS update: Microsoft CEO Satya Nadella said any pursuit of superintelligence must rest on a core principle: AI that cannot benefit humanity and remain under human control is not worth building. Nadella added that Microsoft must accelerate and broaden AI’s benefits across countries, communities and businesses, requiring a frontier ecosystem in which both closed and open-source models can thrive.

TREE NEWS reports: Anthropic, OpenAI and Google have been discussing forming an industry-led artificial intelligence standards body, with working-group meetings still ongoing. The talks began before Anthropic CEO Dario Amodei on Saturday called for AI companies to coordinate on technology testing and auditing. OpenAI CEO Sam Altman told an employee meeting he supports an AI testing and auditing body, but said leading labs must establish a standards body themselves without US government backing.

TREE NEWS reports: Anthropic has chosen Nasdaq as the listing venue for its potential IPO, handing the exchange a second major win after it landed SpaceX earlier this year. Some estimates put Anthropic’s valuation at as much as $2 trillion, though that figure is not yet finalized. SpaceX was valued at $1.75 trillion. OpenAI CEO Sam Altman has said OpenAI will not go public for now.

TREE NEWS update: House Democratic leader Hakeem Jeffries said Democratic lawmakers will hold a caucus meeting Tuesday morning focused on legislative action on artificial intelligence. Jeffries called addressing the challenges posed by AI a “high priority” requiring swift measures, and said Democrats are debating how to set regulatory and safety guardrails in response to the industry’s rapid growth and potential risks.

TREE NEWS update: A developer posting as pdfu on X said iOS 27 and macOS 27 contain an undocumented interface letting third-party apps plug into “Siri extensions” or replace the server-side model behind Siri’s AI. In demonstrations using Claude and GPT-5.6 Terra, the latter received Apple’s native prompts and tool definitions, executed system actions and accessed personal data. Both rely on private entitlements, and Apple has recently tightened the related settings.

TREE NEWS reports: OpenAI co-founder John Schulman, alongside researchers Charlie O’Neill and Beren Millidge, said recursive self-improvement is unlikely to trigger a near-term intelligence explosion, arguing the current Transformer-plus-reinforcement-learning paradigm may hit asymptotic limits in generalization, continual learning and sample efficiency. Progress is shifting from raw compute scaling toward data efficiency and sparse or modular architectures, with distillation and continual learning from real deployment data seen as key for smaller labs.

TREE NEWS reports: Arthur Hayes said in a post on X that if AI compute buyers such as Anthropic, OpenAI and SpaceX stop purchasing, the US government may have to step in as the buyer; alternatively, if vanishing compute demand triggers defaults on AI-related debt, the Federal Reserve may need to print money to rescue insurers holding that debt. Hayes said either path ends in more money issuance, joking that it amounts to “Yachtzee.”