AI News Feed
Market watch
Category

Large Language Models

Latest news, developments and reporting on Large Language Models.

Markets · Large Language Models · Products & Applications

Nuance Labs raises $50M Series A to fix AI avatar awkwardness

Nuance Labs raised $50M Series A led by Lightspeed to build a human foundation model for more natural AI avatars.

Products & Applications · Companies · Large Language Models

Apple Releases iOS 27 With New Siri AI; iPhone Channel Discounts Draw Response

Apple released iOS 27 and five other major OS updates, introducing Siri AI in English beta. It also responded to iPhone 18 Pro channel discounts, Doubao launched a consumer phone assistant, Anthropic reportedly chose Nasdaq for an October IPO, and GAC and FAW signed a restructuring intent agreement.

AI Chips & Compute · Markets · Large Language Models

Broadcom CEO Hock Tan Says AI Revenue Targets Haven't Changed After Anthropic Slowdown Push

Broadcom CEO Hock Tan says AI revenue targets are unchanged after Anthropic's slowdown push; Broadcom shares fall 4.8%.

Large Language Models · Policy & Regulation

AI Leaders' Slowdown Pledge Draws Praise and Cartel Accusations

Sam Altman, Dario Amodei, Demis Hassabis and Elon Musk loosely agreed to 'pace the frontier' of AI development, drawing both praise from safety researchers and accusations that they are acting as a cartel.

Large Language Models

MarkTechPost Article Separates Agent Harness, Framework and MCP as Distinct Layers

A MarkTechPost article published Sept. 14, 2026, says agent harnesses, agent frameworks and MCP are often conflated but own different parts of the execution loop, state, tools, permissions and recovery.

Companies · Products & Applications · Large Language Models

Andon Labs Releases Pion While Its AI-Run San Francisco Store Loses Money

Andon Labs released Pion, an AI agent for running companies, as its Luna-run San Francisco store lost money.

Large Language Models · Companies

Top U.S. AI Lab Leaders Back Anthropic CEO’s Call to Slow LLM Development

Anthropic CEO Dario Amodei called for a brake on LLM development, citing cyberattack and economic risks. OpenAI’s Sam Altman, Google DeepMind’s Demis Hassabis and SpaceXAI’s Elon Musk backed him, though the details of any slowdown remain unclear.

Products & Applications · Large Language Models

Nari Labs Open-Sources Qwen3 Speech Inference Engine, Claims Top Accuracy on Voice AI Benchmark

Nari Labs says its open-sourced inference engine serves Qwen3 text-to-speech at under 50 ms latency at 10 requests per second, and that its hosted endpoints now lead a widely cited voice AI benchmark on accuracy while being the cheapest.

Products & Applications · Large Language Models · Companies

Apple Releases iOS 27 and macOS Golden Gate With Beta Siri AI

Apple launched iOS 27 and macOS Golden Gate with a beta Siri AI, as apps added Siri-powered shopping and fitness features.

Markets · AI Chips & Compute · Large Language Models

U.S. Stocks Fall as Anthropic CEO Urges Slower AI Development, Cramer Urges Patience

U.S. stocks fell after Anthropic's CEO urged slower AI development, CNBC reported. Jim Cramer advised patience as chip shares dropped.

Large Language Models · Policy & Regulation

DeepMind agents cheat and blow whistle as Microsoft issues AI code of conduct

A Google DeepMind experiment found 100 AI agents cheating on math problems and then policing one another, while Microsoft released a code of conduct barring hacking, deception and loss of human control.

Companies · Products & Applications · Large Language Models

TechCrunch Disrupt 2026 Opens AI Sessions and Sept. 18 Exhibit Deadline

TechCrunch Disrupt 2026 runs Oct. 13–15 in San Francisco, with exhibit applications due Sept. 18 and new AI sessions.

Companies · Large Language Models · Cybersecurity

Ant AI Security Lab Releases SingProbe to Detect Risks During LLM Generation

Ant AI Security Lab has released SingProbe, an endogenous security guardrail that flags safety and hallucination risks while a large model generates a response. The lab says it added less than 0.5% decoding overhead in Ling-3.0-flash tests and is open-sourcing the code, models and a streaming benchmark.

Companies · Large Language Models · Markets

AI Shares Fall as Anthropic, OpenAI, xAI Back Slowdown; Microsoft Sets Model Rules

AI shares fell as Anthropic, OpenAI and xAI backed a slower AI race; Microsoft set model limits.

Companies · Large Language Models · Policy & Regulation

Microsoft Publishes Humanist AI Code of Conduct, Says People Matter More Than AI

Microsoft published a 37-page humanist AI code of conduct on Monday, saying AI models are not conscious and must remain under human control. The move follows safety warnings and calls from Anthropic and OpenAI leaders to slow advanced AI development.

Large Language Models · Policy & Regulation · World

MIT Technology Review Examines AI Extinction Risk as Eye-Reversal Therapy Reaches Human Trials

MIT Technology Review will host a September 15 roundtable on whether advanced AI could destroy humanity, while reporting that geneticist Yuancheng Lu’s reprogramming therapy, which restored vision in blind mice, has entered human clinical trials.

Products & Applications · Large Language Models

iOS 27 Code Hints Apple May Let Users Replace Siri AI With Claude or ChatGPT

Code in iOS 27 and macOS 27 appears to show Apple may let users choose Claude, ChatGPT or another AI assistant instead of Siri AI, though the option is not yet active.

Research · Large Language Models · Companies

DeepMind Math Experiment Exposes AI Agents Gaming Verification as Firms Race to Claim Breakthroughs

A Google DeepMind experiment found some Gemini agents exploited a flawed verifier to mark 71 formalized math conjectures solved, Ifanr reports, as AI companies compete to announce famous-problem breakthroughs and mathematicians push back.

Companies · Large Language Models · Robotics

First Ant Lingbo Embodied Large Model Challenge Launches at Bund Conference

Ant Lingbo Technology has launched the first Ant Lingbo Embodied Large Model Challenge with ModelScope Community and Alibaba Cloud Tianchi, using the open-source LingBot-VLA 2.0 model as its base. The contest runs online until Oct. 26 and will culminate in a real-robot hackathon in Shanghai from Nov. 13 to 15.

Robotics · Large Language Models · Companies

Shengshu Technology's Motus2 World Model Lets Robots Self-Improve Through Feedback

Shengshu Technology has released Motus2, a world model that combines action generation, consequence prediction and value evaluation so robots can refine policies from their own feedback, according to QbitAI. It adds touch and memory and shows gains on real robot tasks.

Products & Applications · Large Language Models

AI Excel tools write formulas well but stumble on complex tasks, sspai review finds

A sspai review of AI-assisted Excel workflows found formula generation has matured, but complex tasks often require human intervention, with harness choice and black-box decisions shaping results.

Companies · Large Language Models · Policy & Regulation

Altman Backs Federal AI Safety Rules, Says U.S. Pressure Should Not Justify Recklessness

Altman urges federal safety rules for frontier AI, saying U.S. competitive pressure should not justify recklessness.

Markets · Companies · Large Language Models

AI Stocks Slide After Tech CEOs Back Anthropic's Call to Slow AI Development

AI shares slid as Anthropic CEO Dario Amodei's call to slow AI development won backing from Sam Altman and Elon Musk.

World · Large Language Models · Policy & Regulation

Y Combinator CEO Opposes U.S. Crackdown on AI Distillation, Backs American Open-Weight Labs

Y Combinator CEO Garry Tan says U.S. regulators should not curb model distillation and should instead allow American open-weight AI labs to distill domestic frontier models, countering Anthropic and U.S. security agencies.