AI News
Daily Tech Reader - AI Edition
Apple
Apple Siri
iOS 27 to Launch September 14 With Siri AI in English‑Only Beta
CUPERTINO, Calif., Sept 11 — Apple will release its upgraded Siri AI assistant on Sept. 14 featuring on-screen awareness, integrated web search, and personal context capabilities, the company said on Friday. The initial rollout will require an iPhone 15 Pro or newer and will exclude users in the European Union and China at launch.
Sources: Apple · Engadget · 9to5Mac · Techmeme
Apple Intelligence Adds Multi‑Step Automation and Health Guidance Across Devices
CUPERTINO, Calif., Sept 11 — Multi-step automation tools, visual intelligence, and on-device health guidance will debut across iPhone, iPad, Mac, and Apple Watch hardware with the release of iOS 27, iPadOS 27, and macOS 27, Apple announced on Friday.
Sources: Apple Newsroom · Macworld · Techmeme
iPhone 18 Pro Pre‑Orders Begin September 12
CUPERTINO, Calif., Sept 11 — Pre-orders for the iPhone 18 Pro will open Saturday, Sept. 12, ahead of customer deliveries on Sept. 18, Apple said. The new models incorporate upgraded on-device neural processing designed to improve speech recognition and dictation accuracy.
Sources: Apple · Gadget Hacks · Techmeme
Copilot
Microsoft Completes Worldwide Rollout of Unified Copilot App
REDMOND, Wash., Sept 11 — Microsoft has completed the global deployment of a unified Copilot application combining its consumer tools and Microsoft 365 enterprise suite under a single sign-in across Windows and macOS systems, the company said on Friday.
Sources: Windows Central · TechCrunch · Techmeme
Copilot Studio Adds Bring‑Your‑Own‑Model Support Through Azure AI Foundry
REDMOND, Wash., Sept 11 — Microsoft will permit enterprise customers to integrate third-party systems, including Anthropic's Claude and Meta's Llama, into its Copilot Studio workflow builder, the company announced. The integration connects fine-tuned domain models directly via the Azure AI Foundry catalog.
Sources: Dynamics Edge · Microsoft Learn · Techmeme
Copilot in SharePoint Gains Reusable Personal Skills and SharePoint‑Safe HTML
REDMOND, Wash., Sept 11 — Microsoft deployed reusable personal automation skills for Copilot across its SharePoint and OneDrive services on Friday and updated the tool to output tenant-compliant HTML for intranet environments, according to a company bulletin.
Sources: Microsoft Community Hub · Techmeme
ChatGPT
OpenAI Introduces ChatGPT for Financial Services With Live Market Feeds
SAN FRANCISCO, Sept 11 — OpenAI launched a specialized ChatGPT tier for institutional finance on Friday, integrating native data feeds from PitchBook, LSEG, and Crunchbase alongside its GPT-6 Astra reasoning architecture. The tool is designed to parse SEC filings, automate portfolio financial models, and output compliant client pitchbooks without routing corporate queries through external connectors.
Sources: The Wall Street Journal · Financial Times · Bloomberg · Techmeme
OpenAI Rolls Out Images 2.5 With Multi-Turn Consistency and Sketch Input
SAN FRANCISCO, Sept 11 — OpenAI released its Images 2.5 generation engine to ChatGPT tiers and developer APIs, introducing interactive freehand sketch guidance and improved subject-retention controls across multi-turn image edits. The company said the underlying infrastructure now handles more than 3 billion visual prompts weekly.
Sources: TechCrunch · The Verge · VentureBeat · Techmeme
OpenAI Withdraws Sponsorship of Caltech Mathematics Hackathon Over Academic Pushback
PASADENA, Calif., Sept 11 — OpenAI withdrew its corporate backing from the California Institute of Technology’s competitive mathematics marathon following an open letter from faculty criticizing AI model claims on complex theoretical proofs. Organizers confirmed the company's $20,000 computing token grants would be replaced by private departmental endowments.
Sources: Business Insider · Times of India · Inside Higher Ed · Techmeme
Gemini
Google Deploys Agentic Multi-App Workflows Across Workspace Suite
MOUNTAIN VIEW, Calif., Sept 11 — Alphabet introduced agentic automation features across Google Workspace on Friday, allowing users to initiate cross-service actions—such as building Slides presentations directly inside Google Chat and generating financial Sheets from Drive directories—via Gemini background processing.
Sources: Google Cloud Blog · 9to5Google · ZDNet · Techmeme
Gemini Live Integrates In-Chat Multimodal App Actions Globally on iOS and Android
MOUNTAIN VIEW, Calif., Sept 11 — Google began rolling out an updated Gemini Live interface across mobile platforms, allowing users to toggle between conversational voice queries and text chats while displaying real-time navigational maps, visual analysis cards, and calendar updates within a single thread.
Sources: Android Central · Engadget · The Verge · Techmeme
Google Upgrades Core Consumer Chat Default to Gemini 3 Flash Architecture
MOUNTAIN VIEW, Calif., Sept 11 — Google completed the migration of its standard Gemini consumer web and mobile interfaces to Gemini 3 Flash, setting the model as the primary default engine for general reasoning and daily multimodal tasks while reserving Gemini 3 Pro for advanced technical computation.
Sources: Google Release Notes · Reuters · TechCrunch · Techmeme
Claude
Anthropic Reports 200 Million Claude Exchanges Linked to Five Distillation Campaigns
SAN FRANCISCO, Sept 11 — Anthropic published a cybersecurity threat assessment on Friday detailing five Chinese distillation operations that extracted model outputs across 200 million automated interactions. A campaign linked to Alibaba's Qwen project represented the largest share, accounting for 151 million exchanges via thousands of coordinated accounts.
Sources: TechCrunch · Reuters · Techmeme
Departing Anthropic Researcher Estimates Human Extinction Risk Above 10%
SAN FRANCISCO, Sept 11 — Outgoing Anthropic researcher Jack Coxon stated publicly that the probability of human extinction resulting from frontier AI systems exceeds 10 percent. The disclosure prompted safety advocates and market analysts to call for a pause on the company's scheduled October initial public offering.
Sources: CNBC · Wall Street Journal · Techmeme
Anthropic Faces IPO Scrutiny Amid Safety Disclosures and Distillation Reports
NEW YORK, Sept 11 — Anthropic's planned October initial public offering is facing heightened investor review following disclosures regarding model distillation attacks and internal risk evaluations, according to people familiar with the syndicate. The transaction is being underwritten by Morgan Stanley, Goldman Sachs, and JPMorgan.
Sources: CNBC · Yahoo Finance · Barron's · Techmeme
Grok
xAI Expected to Release Grok 4.7 With Expanded Parameter Count
SAN FRANCISCO, Sept 11 — Elon Musk's xAI is preparing to introduce Grok 4.7, a foundation model featuring an expanded parameter architecture trained on SpaceX engineering data, developer disclosures show. Platform documentation currently lists Grok 4.6 as the production flagship.
Sources: CellCog · X (xAI) · Techmeme
xAI Deprecates Grok Image 1.0 in Favor of Image 2.0
SAN FRANCISCO, Sept 11 — xAI retired its Grok Image 1.0 generation tool on Friday, shifting all developer API traffic to Image 2.0 at a reduced unit price without altering interface protocols, the company said.
Sources: Releasebot · xAI · Techmeme
Grok Build Terminal Update Adds Navigation and Rendering Improvements
SAN FRANCISCO, Sept 11 — xAI deployed a software update for its Grok Build Terminal on Friday, adding higher-frequency display rendering, corrected interface navigation, and improved management of autonomous subagents.
Sources: Releasebot · xAI · Techmeme
China AI
U.S. Agencies Accuse Six Chinese AI Firms of Large‑Scale Model Distillation
WASHINGTON, Sept 11 — A joint advisory issued Friday by the National Security Agency, FBI, and CISA accused six Chinese artificial intelligence firms of conducting systematic extraction of training data from U.S. frontier models since 2024.
Sources: Bloomberg · Nextgov · Quartz · Techmeme
Moonshot Releases Kimi K2.5 With Video and Agentic Capabilities
BEIJING, Sept 11 — Chinese startup Moonshot AI released its Kimi K2.5 system on Friday, integrating video generation and agentic task execution. The firm said the system outperformed several competing U.S. models across selected industry benchmarks.
Sources: CNBC · Fortune · Techmeme
DeepSeek Pursues Funding at $71 Billion Valuation
BEIJING, Sept 11 — Chinese artificial intelligence firm DeepSeek is holding discussions with prospective investors to raise its first external funding round at a targeted valuation of $71 billion ahead of a potential Shanghai listing later this year, according to people familiar with the matter.
Sources: Second Talent · AI Weekly · Techmeme
Big Story
Distillation Reports, Safety Estimates, and IPO Planning Shape AI Sector Outlook
NEW YORK, Sept 11 — Scrutiny of the frontier artificial intelligence sector intensified on Friday following U.S. intelligence warnings of widespread Chinese model harvesting, a public existential risk assessment from a departing Anthropic researcher, and ongoing due diligence ahead of Anthropic's planned multibillion-dollar share offering.
Sources: TechCrunch · CNBC · Bloomberg · Reuters · Techmeme