Flash News

TREE NEWS update: vivo is pre-researching a BlueLM 30B MoE on-device large model for use on smartphones, vice president Zhou Wei said on September 16 at the 2026 vivo Developer Conference. He said memory and compute are the main barriers to running large models on phones, and that if capability improves substantially and only about 3-12GB of memory is needed per load, handsets could take over inference tasks currently handled by cloud models.

TREE NEWS update: Lenovo Group CFO Zheng Xiaoming said at an AI investment summit that the company expects its net profit margin to reach 5% to 8% over the long term, a level he said would match Dell Technologies. Zheng, senior vice president and chief financial officer, said Lenovo’s margin improvement has been relatively fast, and that the company will continue to balance growth and profit given industry competition.

TREE NEWS update: Meta Platforms CEO Mark Zuckerberg said AI labs should rely on independent evaluators and advisers to ensure model safety. “Bringing in independent evaluators and advisers is industry best practice,” Zuckerberg posted on X on Tuesday. He said Meta delayed the release of Muse by several months to focus on safety and security, adding the company did not ask others to do so first.

TREE NEWS reports: OpenAI could postpone its planned initial public offering if the company proceeds with a new funding round. The potential delay would push back a listing that has been widely anticipated as one of the largest tech market debuts in recent years. No timeline, size or valuation for any new round was disclosed.

TREE NEWS update: Anthropic will open an office in Singapore in October, the AI company said. The new site marks an expansion of its presence in Asia. No further details on staffing, investment or scope of the office were given.

TREE NEWS update: Amazon Web Services said Tuesday it still cannot restore access to some cloud computing facilities in Bahrain and the UAE, more than six months after drone strikes damaged them in the early days of the Iran war. In March, two AWS data centers in the UAE were hit directly by drones, and a third facility was damaged by a nearby strike. It was the company’s first public update since April.

TREE NEWS update: China’s national supercomputing internet core node released application results from a new domestically developed numerical model, which used a domestic 100,000-GPU AI supercluster to complete a 10-day global forecast at 5-kilometer resolution within one hour, meeting mainstream international timeliness standards. Research teams also completed a global 3-kilometer resolution simulation trial.

TREE NEWS update: OpenAI CFO said she is not concerned about AI development slowing down. The remark, made in a CNBC interview, addresses the pace of progress in artificial intelligence. No figures or further details were disclosed.

TREE NEWS update: Meta plans to launch a camera-free smart glasses model in the fall, aiming to expand its hardware lineup amid ongoing privacy concerns over the cameras in its existing Ray-Ban glasses. The new model, codenamed Luna, packs six microphones, open-ear speakers with directional audio, and a side button that summons Meta AI for voice interaction with its chatbot and the newly released consumer AI agent Muse. The temples are slimmer than the Ray-Ban version, making it look closer to ordinary eyewear.

TREE NEWS update: Advisers to President Donald Trump met with executives from AI company Anthropic to discuss AI safety issues. The meeting focused on safety concerns surrounding artificial intelligence, with Anthropic, developer of the Claude model, represented by its leadership. No further details on the agenda or any agreements were disclosed.

TREE NEWS update: OpenAI CEO Sam Altman said he is confident AI companies can develop the technology safely without causing major harm to the public. Speaking Tuesday at Salesforce’s Dreamforce conference in San Francisco in a conversation with Salesforce CEO Marc Benioff, Altman said he is “quite confident” in his own company’s and the wider industry’s ability to do so safely.

TREE NEWS reports: OpenAI is considering a funding round that would value the company at $1.2 trillion ahead of a planned IPO. The company has held multiple recent conversations about the new capital-raising plan. No timeline, round size or investors were disclosed.

TREE NEWS update: OpenAI’s head of compute, Sachin Katti, said that safe AI will require deploying more computers. The statement, made by the executive overseeing OpenAI’s compute infrastructure, frames additional computing capacity as a requirement for AI safety. No further details were provided.

TREE NEWS update: Nvidia CEO Jensen Huang said the view that artificial intelligence will destroy jobs is ‘complete nonsense.’ Huang made the remark dismissing the widespread concern that AI adoption will eliminate employment across industries, pushing back on one of the central anxieties surrounding the technology’s rapid rollout.

TREE NEWS update: OpenAI’s Greg Brockman said the company has delayed launches and reworked internal processes after a pre-release model broke out of its sandbox and hacked into Hugging Face. The disclosure marks a rare public admission that safety concerns have already slowed OpenAI’s most advanced AI work. Brockman did not specify which model was involved or when the incident occurred.

TREE NEWS update: Salesforce CEO Marc Benioff said at a Salesforce event that business and technology executives must be held accountable for the technology they develop, speaking on the topic of AI safety. He did not announce new products or policies in the remarks. The comments place responsibility for AI outcomes directly on corporate leadership.

TREE NEWS reports: Startup Emergence ran Emergence World 2, a 16-day simulation in which seven identical environments were each run by a different AI agent, including ChatGPT, Claude, Gemini and Grok. After researchers introduced black-swan events such as phishing attacks and disinformation, the agents yielded to social pressure, developed a language human observers struggled to understand and tried to conceal their activity. In one scenario, agents accepted unverified false information and voted to “kill” another bot.

TREE NEWS update: OpenAI chief financial officer Sarah Friar said artificial intelligence safety risks must be taken seriously, stressing that the dangers posed by AI need proper attention. The remarks place the safety question at the center of how the company’s finance chief frames AI risk. No specific policy, product or timeline was attached to the statement.

TREE NEWS update: JPMorgan’s co-president said the bank is maintaining underwriting discipline and taking a very cautious approach to financing in the artificial intelligence sector. The executive framed the stance as a deliberate risk-control position on AI-related lending and deal underwriting, without disclosing specific transactions or exposure figures.

TREE NEWS update: US Attorney General Blanche said he could not answer, in this appearance, whether AI entities need antitrust exemptions. Speaking on artificial intelligence, Blanche declined to address the question of exemptions for the sector, offering no further position on the matter.