Flash News

TREE NEWS update: OpenAI Codex product lead Tibo said on day one of the update that default speeds for GPT-6 Astra and GPT-6.1 Sol have been optimized by about 50%, covering all subscribers and products and partners that log in with ChatGPT accounts, including OpenCode, Pi, Amp and Devin. No user action is required and the boost is expected to take effect within two hours, with throughput rising from 30 TPS to 50 TPS and a top-tier tokenizer.

TREE NEWS reports: OpenAI has been accused of stealing the name “Astra” for its latest AI model release, Bloomberg Law reported. The allegation centers on OpenAI’s use of the Astra name in connection with the newly launched model. No further details on the accuser or the legal claims were provided.

TREE NEWS reports: Nohshad Shah, head of fixed income sales for EMEA at Citadel Securities, said rising US Treasury real yields could test the AI boom, estimating that about one-third of hyperscaler capex this year is debt-financed. Higher funding costs will make future cash generation more important, he said, while reiterating his bullish view on hyperscalers such as Microsoft and Google because their business models extend beyond selling access to AI models.

TREE NEWS update: Ukraine’s air force says AI-equipped turrets have successfully intercepted a new type of Russian jet-powered drone. Air force spokesman Yuriy Ihnat said in a media interview on the 5th that machine vision is being used to shoot down drones and that the AI can even lock onto jet-powered UAVs, adding that such technologies help avoid human error.

TREE NEWS update: OpenAI will introduce text watermarking, or identification, features for ChatGPT users in the European Union market. The move adds a labeling mechanism to ChatGPT-generated text for EU users, marking AI output as such. No timeline or technical details of the rollout were disclosed.

TREE NEWS update: Web3 security firm CertiK released its Intel3D report “The Rise of AI Employees: How Agentic AI Is Reshaping Cybersecurity, AML and Compliance” on Tuesday, saying AI is moving from an analyst aid to a team member that can investigate, judge and act within set permissions. The shift is reshaping security and compliance in traditional finance and Web3, CertiK said, while warning of model misjudgment, prompt injection and attackers’ use of AI. It recommended governance covering permission boundaries, human approval, decision logs and accountability, plus regular red-team testing.

TREE NEWS update: CertiK published its Intel3D report on October 5, finding that agentic AI is moving from assisting analysts to autonomously investigating, judging and acting within defined permissions in cybersecurity, AML and compliance. The report says AI-assisted smart-contract auditing and cross-chain tracing are key Web3 applications, citing CertiK research that 86.29% of stolen ETH after the Bybit attack was swapped into bitcoin within a month. It warns firms must audit AI agents’ own on-chain behavior and assign human accountability.

TREE NEWS reports: Cerebras shares rose 10.8% after OpenAI CEO Sam Altman described the company as a close partner of OpenAI. Altman’s remark was the catalyst for the stock move. No further terms or details of the partnership were disclosed.

TREE NEWS update: Binance announced the launch of Binance Intelligence, an AI product stack comprising three products: free Binance AI for all users, Binance AI Pro, which converts natural language into executable strategies, and Binance Agent OS for developers building AI applications. Binance AI integrates charts, news, social platforms, on-chain data and research tools, with a generative UI that adapts to each user’s knowledge level, interests and activity, plus a “For You” tab, market briefs, smart tooltips and Master Trade. The rollout begins on October 5.

TREE NEWS reports: Calls to slow development of the most powerful AI models are colliding with an $800 billion investment boom, raising the question of whether an AI slowdown could tank the US economy. The tension pits safety advocates against the scale of capital now committed to AI infrastructure.

TREE NEWS reports: US companies are increasingly routing routine AI work to cheaper open-weight models, with mentions of “open-weight” or “open-source” models in earnings and investor calls in August and September running six times the year-earlier level. Vercel said open-source models handled 56% of tokens through its AI Gateway in August, up from 7% in December, while AT&T runs about 40% of AI workloads on open models and targets 70% within a year.

TREE NEWS reports: Bloomberg Intelligence senior analyst Robert Lea said in a report that China’s top AI models now trail US rivals by 3% on benchmark scores, a record low, down from about 9% in May and roughly 15% at the start of the year. The narrowing followed DeepSeek’s September release of V4.1 Flash. Lea expects the performance gains to help Chinese AI vendors expand market share.

TREE NEWS reports: Hong Kong stocks closed higher, with the Hang Seng Index up 0.28% and the Hang Seng Tech Index up 0.62%. Semiconductors and AI application shares were the standout performers. Zhipu (02513.HK) closed up 6.15%.

TREE NEWS reports: DBS Group Chief Investment Officer Hou Wey Fook said Nvidia’s valuation and its projected 70% earnings growth next year show AI-driven tech stocks are still far from bubble territory. Nvidia currently trades at 17 times forward 12-month earnings, he noted.

TREE NEWS update: The United States, Japan, the UK, Germany and 13 other countries on Oct. 4 endorsed the Kyoto Vision for a Golden Age of Science at the 23rd annual Science and Technology in Society forum in Kyoto. The declaration calls for expanding researchers’ access to AI tools, scientific data, computing infrastructure and experimental facilities, and for reforming research funding and institutional organization, including long-term grants, fast funding, prizes and challenge competitions for high-risk, high-potential work.

TREE NEWS reports: DBS Group chief investment officer Hou Wey Fook said Nvidia’s valuation and its projected 70% earnings growth next year show AI-driven tech stocks are far from bubble territory. Nvidia’s forward 12-month price-to-earnings ratio stands at 17 times, which he compared with Cisco Systems’ roughly 100x multiple before the dot-com crash. “If the poster child of AI is Nvidia and it’s trading at only a teens multiple, how can you call this a bubble?” he said, adding semiconductor and AI investment still face tailwinds.

TREE NEWS update: Bank of Japan Deputy Governor Shinichi Uchida said AI adoption could have both positive and negative effects on productivity and the labor market. He said AI is a large positive demand shock that pushes up economic activity and prices, while possibly lifting productivity and capital accumulation on the supply side. AI has boosted share prices and eased financial conditions, but heavy bond issuance by AI-related firms is adding upward pressure to long-term rates, he said, warning of correction risk if profits do not keep pace.

TREE NEWS update: OpenAI CEO Sam Altman has staked out a position on AI risk that diverges from Anthropic PBC’s, arguing that the benefits of AI technology help justify taking on the risk. The disagreement puts the two leading AI labs on different footing over how the dangers of advanced models should be weighed against their potential gains.

TREE NEWS update: SoftBank Group leader Masayoshi Son said he is concerned about the safety risks of artificial intelligence, warning that superintelligence could become “super dangerous” in the wrong hands. Speaking Sunday at an event during the “Technology and Society” forum in Kyoto, Son said AI capabilities are growing exponentially and that countries must build mutual trust to manage the technology together. He said humanity no longer has “the luxury of fighting among ourselves.”

TREE NEWS update: President Donald Trump said US intel chief Jay Clayton will lead the new Super Intelligence Force, reporting to the president and Chief of Staff Susie Wiles. Trump announced the appointment, placing the sitting intelligence chief at the head of the SIF. No further details on the body’s mandate or start date were given.