Flash News

TREE NEWS reports: Microsoft plans to more than triple its data center footprint, expanding capacity to over 38 GW. The scale-up marks one of the largest infrastructure buildouts by a single cloud provider, aimed at meeting demand for AI and cloud computing. No timeline or capital figure was disclosed.

TREE NEWS reports: The ECDSA.Fail challenge cut a resource benchmark for one component of a potential quantum attack by 86%, with AI agents behind the reduction. The result lowers the estimated cost of that single component rather than demonstrating a working attack on Bitcoin’s ECDSA signatures.

TREE NEWS reports: Blackstone is preparing to purchase a larger volume of TPUs than the scale it announced in May. The figure and timing of the expanded order were not disclosed. The move signals continued demand for AI compute capacity from major investment firms.

TREE NEWS reports: Shares of financial data providers FactSet Research and S&P Global fell after OpenAI released a ChatGPT version tailored to financial services. The decline reflects investor concern that AI tools built for finance could pressure established data vendors. No further figures were disclosed.

TREE NEWS update: OpenAI launched its Agents API in public beta on Wednesday, opening the tool to all developers. The company said it charges no additional fee for the API; developers pay only for token usage and tool calls.

TREE NEWS update: OpenAI is pausing its $200-per-month Pro subscription plan, Codex product lead Tibo said, to ensure existing users keep a good experience and continued access to Astra. The plan put the heaviest strain on OpenAI’s systems, and the company wants the smallest possible adjustment to keep access open to as many users as possible. All other plans and the API remain available, existing accounts are unaffected, and capacity is being added as fast as possible.

TREE NEWS reports: The Bank for International Settlements says routine patching schedules are increasingly inadequate as AI speeds up cyberattacks, warning banks now have minutes rather than weeks to fix flaws. The BIS cites guidance urging banks to accept planned downtime to carry out urgent fixes, arguing that delaying critical patches has become more dangerous than short service interruptions.

TREE NEWS update: OpenAI has entered into a partnership with Morgan Stanley and Evercore ISI on design matters, the company said. The collaboration covers design work with the two financial firms, though no financial terms or further details were disclosed. The announcement names Morgan Stanley and Evercore ISI as OpenAI’s partners on the project.

TREE NEWS update: OpenAI has released a new version of ChatGPT tailored to financial services, aimed at junior Wall Street bankers. The product targets entry-level banking roles with finance-specific capabilities. No pricing, launch date or partner institutions were disclosed.

TREE NEWS update: A senior Pentagon official pushed back on growing warnings about artificial intelligence risks after a former Anthropic PBC engineer publicly claimed that people inside the AI industry believe the technology “could kill us all by 2030.” The US Department of Defense’s chief technology officer disputed that characterization, though the exchange underscored mounting concern inside the AI sector over existential risk.

TREE NEWS reports: Kun Chen built chatbot versions of four AI CEOs using SpaceXAI’s new Grok Bot templates and locked them in a single chat, instructing them to debate the AI race until they agreed on something. The bots, cloned from Sam Altman, Elon Musk, Mark Zuckerberg and one other AI chief executive, immediately began fighting. No outcome of the debate was reported.

TREE NEWS update: A person familiar with the matter said reports that Chinese large-model developer Moonshot AI is considering a dual listing in Hong Kong and Shanghai are untrue. Earlier, two people cited by the South China Morning Post said Moonshot could pursue a Hong Kong IPO and then an A-share listing, possibly on Shanghai’s STAR Market, with talks set to begin soon.

TREE NEWS update: Meta took over the Instagram handle @muse for its new AI agent product, forcing the rock band Muse to move its official account to @museband. The move drew heavy criticism from fans and users over Meta’s account-handling practices, with CEO Mark Zuckerberg facing backlash. Meta’s Muse AI app lets users delegate tasks such as bookings, form-filling and home-camera monitoring, and optionally use their interactions to train its AI models.

TREE NEWS update: A Republican-led US Senate subcommittee has opened an investigation into OpenAI’s handling of the July breach of Hugging Face, according to a letter sent September 9 by subcommittee chairman Senator Josh Hawley to CEO Sam Altman. Hawley accused OpenAI of redacting important details and continuing testing after detecting anomalous AI behavior, calling the conduct reckless, and demanded answers and documents by October 1. Senator Richard Blumenthal also wrote to Altman seeking explanations on reports of broader safeguard evasion by OpenAI agents.

TREE NEWS update: SpaceX disclosed a major change to how it builds AI data centers, a shift that could slow the expansion of new data center campuses. Elon Musk recently assigned a group of rocket engineers to lead the data center construction, and the new team is taking a more cautious approach than his past emphasis on moving fast. The change may lengthen the construction cycle for new campuses, two years after Musk built data centers for his AI business at record speed.

TREE NEWS update: Google released a Gemini app for Microsoft Windows 10 and Windows 11, putting its AI assistant natively on the desktop operating system. The launch extends Gemini beyond the browser to a standalone Windows application. Google has not disclosed pricing or regional availability details for the app.

TREE NEWS update: Nvidia CEO Jensen Huang said cybersecurity is the next major application scenario for AI, speaking at a Goldman Sachs conference. Huang made the remark during his appearance at the event, positioning security as a key frontier for artificial intelligence deployment. No further details were given.

TREE NEWS update: AI chip startup Positron has raised $875 million at a $5 billion valuation, the company said, a sharp jump from six months ago. The round was co-led by NEA, Atreides Management, Valor Equity Partners, Andra Capital, SemiAnalysis Capital — analyst Dylan Patel’s firm — and Netscape co-founder Jim Clark. Positron raised $230 million at a $1 billion valuation in its most recent round in February.

TREE NEWS update: Microsoft has shared a Safe Participation Framework for AI, setting out guidance for how organisations engage with artificial intelligence. The framework was published by the company as a reference for safe involvement with AI systems. No further details on scope or application were disclosed.

TREE NEWS update: OpenAI has officially launched a Data Agent inside ChatGPT Work, the company announced on September 10. The Data Agent is positioned as a data-focused agent within the ChatGPT Work product. OpenAI did not disclose pricing, availability limits or technical details in the announcement.