AI News for September 24, 2026

A weekly AI agents news digest covers Alibaba's new AgentCore enterprise agent platform, OpenAI's voice-driven work agents in ChatGPT mobile, Amazon's agentic workflows for marketplace sellers, new agentic payment rails from Mastercard and Worldline, and Proofpoint's agentic data and AI security system.

AI's growth is increasingly constrained by power availability and grid limits rather than chips, prompting interest in distributed and edge data centers that use idle GPU capacity.

Deloitte reported FY2026 global revenue of US$74.5 billion and detailed its AI-focused investments, including a US$3 billion modernization program and agentic AI, AI infrastructure, and physical AI capabilities.

NVIDIA released a reference design to guide data-center operators and partners in building AI facilities that handle the power, cooling, networking, and density demands of GPU clusters.

At WSO2Con Africa 2026, WSO2's Ruwan Linton argued that enterprise AI's execution gap stems from integration, data, workflow, governance, and ownership problems around the model, and proposed outcome-anchored forward-deployed engineering to close it.

agents-radar is a GitHub Actions-powered tool that aggregates AI ecosystem signals from 10 data sources and publishes bilingual daily digests as Issues, Markdown, Pages, RSS, plus a hosted MCP server for querying reports.

The agents-radar project is a GitHub Actions-driven daily AI ecosystem digest that aggregates signals from 10 sources such as GitHub, arXiv, Hacker News, and Product Hunt and publishes bilingual reports via issues, Markdown files, a web UI, and an MCP server.

A practical analysis arguing that timestamp-chaptered video has become essential for brands seeking citation and visibility in AI search results and answer engines.

一篇券商式投资分析，梳理AI巨头布局AI制药背景下临床前CRO两条受益路径——湿实验（DNA/蛋白合成）需求增长与安评实验猴供需紧平衡，并提示相关ETF与风险。

MarTech's weekly roundup covers a LinkedIn AI-slop detector Chrome extension plus a broad set of AI product releases across Adobe, Attentive, Bidgely, G2, Haus, and other marketing technology vendors.

A Price Per Token tracker page lists recently released AI and LLM models from providers like OpenAI, Anthropic, Google, Cohere, xAI, and Xiaomi alongside their per-token input and output pricing and context window sizes.

A TechBullion analysis argues that enterprise software is quietly shifting from tools that help humans execute decisions to systems that autonomously commit decisions themselves, driven by mature inference, tool use, and grounding layers.

A stats roundup covering AI in manufacturing market size, regional breakdowns, and adoption rates, including forecasts to 2035 and 72% plant-floor adoption.

A vendor blog roundup comparing ten AI coding agents for enterprise use in 2026, emphasizing autonomy, security, governance, and integration criteria.

A vendor blog compares Replit, Lovable, and Bolt as AI MVP builders, covering pricing, lock-in, when to outgrow them, and pitching TechRev's migration services.

The THTHDGCS/agents-radar repository is a GitHub Actions-based tool that aggregates AI ecosystem signals from 10 sources and publishes bilingual daily digests, now with a hosted MCP server for querying reports.

A GitHub Actions-based project called agents-radar aggregates AI ecosystem signals from ten sources and publishes bilingual daily digests, plus a web UI, RSS feed, Telegram channel, and hosted MCP server.

A Polaris Market Research press release forecasts the global robotic platform market will grow from USD 10.14 billion in 2025 to USD 16.80 billion by 2034 at a 5.8% CAGR, citing industrial automation, autonomous mobile robots, and healthcare robotics as key drivers.

The Guardian's AI topic hub page aggregates recent articles on AI, covering Chinese and US AI governance talks, an OpenAI agent's hack of Australia's Medicare system, deepfake politics, disinformation, and AI industry developments.

Meta unveiled Muse Charm, a keychain AI puck shipping in December, while a researcher showed Muse could be tricked into emailing out 6.8GB of internal runtime files including SSH keys, and Xiaomi's MIT-licensed MiMo-V2.6 topped the open-weight leaderboard two weeks after Anthropic accused it of extracting Claude training data.

Ema raised a $77 million Series B led by Creaegis to scale its enterprise AI agent platform, but the round's real test is whether it can deliver software-like margins instead of consulting-style bespoke deployments.

At Meta's Connect event, Mark Zuckerberg unveiled a wave of new Muse AI agent capabilities, including a real-time video-chat avatar, smart glasses integration, Mac desktop control, a dedicated email address, and commerce partnerships with Stripe, Shopify, and major retailers.

China Telecom AI has officially released Xing4.0-29B-A4B, a 29B-parameter Mixture-of-Experts agentic model with 4B activated parameters and a 256K context window that runs locally on a single consumer-grade GPU with just 15 GB of memory.

Google launched Gemini 3.8 Flash TTS and Flash-Lite TTS, API-only text-to-speech models that enable prompt-steered custom voice design with SynthID watermarking and consent-gated voice replication.

Anthropic's Claude Help Center release notes document September 2026 launches including Claude Opus 5.5, unified Cowork chat experience, Salesforce plugin, and smart reports.

Google DeepMind's new chief Koray Kavukcuoglu says Gemini 4 is in its refinement stage and will launch 'much earlier' than the end of the year, as Google tries to catch up to rivals after leadership changes and delays.

Google DeepMind has moved its next-generation Gemini 4 model into the post-training phase, with plans for an early staged release as Google tries to close the gap with OpenAI and Anthropic.

A Medium analysis interprets Alibaba's official Qwen 3.8-Max benchmark tables, highlighting large agentic and multimodal gains alongside an Omniscience hallucination regression.

A vendor blog reference page cataloging every Anthropic Claude model from Claude 3 through Opus 5.5, with specs, pricing, benchmarks, and guidance on which active models to use.

An analysis comparing Google's Gemini 3.8 Live speech-to-speech model and Alibaba's Qwen-Audio-3.1-TTS, arguing they serve complementary roles in a voice stack and that the eight-day gap between launches shifted the cost calculus.

Ellucian announced it will showcase Ellucian Student and its higher-education-specific AI capabilities, powered by a 50-year Higher Ed Knowledge Graph, at EDUCAUSE 2026 alongside eight conference sessions.

DemandSphere maintains a reference tracker comparing benchmarks, pricing, and capabilities across major proprietary and open-weight frontier AI models.

This is a section landing page listing recent Indian Express AI headlines, covering topics like Meta's Muse superintelligence, OpenAI agent security incidents, GPT-6 model launches, and AI regulation, rather than a single news article.

TechCrunch's AI category page is a topic hub covering artificial intelligence and machine learning news, companies, and ethical issues.

A UMass Amherst and UNC-Chapel Hill study finds that ChatGPT-4o mini produced nearly 10% inaccurate election-related claims before the 2024 US election, with hedging, critical omissions, and fabricated details that raise concerns for future elections as AI replaces search.

AI-generated political ads featuring deepfakes and fabricated scenes are flooding the 2026 US midterm campaigns, raising concerns about voter deception and eroding trust in elections as regulation remains thin.

A 2026 report finds that U.S. adults' use of AI chatbots and search summaries for election information has risen sharply, raising accuracy, language-access, and misinformation concerns ahead of the 2026 midterms.

A behavioral science newsletter argues that the rapid public backlash against AI and data centers stems from authentic grievances amplified by existing biases, not manufactured misinformation.

An opinion essay reflects on Dario Amodei's warnings about artificial superintelligence and asks what purpose human life will serve if AI surpasses human intelligence.

A media-industry commentary argues that AI-generated misinformation and delayed local reporting in a Long Island mayoral race foreshadow challenges for the 2026 midterms and offers tips for reporters covering fake AI content.

A law firm blog in a three-part series analyzes the Quinteros appellate sanctions against attorneys who filed AI-fabricated citations, warning California workers' compensation adjusters that credibility at the WCAB is equally at risk.

A Mediated Mind newsletter roundup examines a University of Miami study where middle schoolers teach AI agents math, APA guidance on educational technology, and Trinity College Dublin research on workers reviewing AI outputs.

A French philosophy magazine explainer traces the meaning and origins of "AI alignment," from Norbert Wiener's 1960 warnings to Stuart Russell's 2014 value alignment work, arguing the term means more than mere instruction-following.

A Townhall commentary uses an interview with a Manhattan Institute fellow to argue schools should teach AI gradually from third grade rather than ban it outright.

Uniper and Celonis report cutting time-to-signature by 27 days and screening time by 13 days using a context-aware AI agent built on Celonis Context Model and Microsoft Copilot Studio for HR recruiting.

Law firms are rapidly adopting AI while courts sanction lawyers for hallucinated citations, exposing a patchwork of rules and a gap between AI assistance and professional accountability.

An overview of AI recruitment tools for sourcing, screening, chatbots, interviews, and dashboards, with a caution that human oversight and bias testing remain essential.

Career Ahead surveys proposals from policymakers and labor advocates for retraining programs, expanded unemployment benefits, and union bargaining to support workers displaced by AI automation.

A consultant-authored guide argues that growing businesses should map workflows, define precise human oversight, measure full automation costs, and update roles and AI writing policies before adopting AI in HR and operations.

Carlos Felipe Balcazar argues that because AI supports some jobs while undermining others, workers are unlikely to unite in response to automation, which will further concentrate economic and political power among top earners.

A Poets&Quants analysis argues that AI will not uniformly devalue credentials and that Harvard's HBS Online certificates may actually become more valuable in an AI-driven labor market.

An education blog post offers 40 turn-and-talk discussion prompts for K-12 lessons, organized into five categories, along with tools for timing and pairing students.

New York Attorney General Letitia James led a bipartisan coalition of 26 attorneys general in calling on congressional leaders to pass federal AI regulation covering safety testing, incident response, and state enforcement authority.

Newly unsealed internal documents show Microsoft and OpenAI executives privately acknowledged mass web scraping as unprecedented theft and sought paywall workarounds, bolstering the New York Times copyright lawsuit ahead of summary judgment on fair use.

Newly unsealed court filings reveal internal OpenAI and Microsoft executive statements describing AI as substitutive for journalism, which news publishers argue undermines the companies' fair use defense in their copyright lawsuit.

The number of AI copyright lawsuits has more than doubled to 145 since December, with OpenAI named in 23 cases, as commentators debate the unresolved fair-use question and its chilling effect on AI development.

A Politico/Public First poll finds roughly half of adults in six Western countries support pausing advanced AI development, and the article argues new regulation may not reduce risk and could entrench incumbent labs through regulatory capture.

Stephenson Harwood's September 2026 Neural Network rounds up AI regulatory and market developments, including LLM provider transparency compliance under the EU AI Act, Italian deepfake sanctions, UK liability proposals, FCA cybersecurity warnings, and notable AI investments and acquisitions.

A vendor blog post argues that AI governance difficulty in 2026 concentrates in unclear internal risk ownership, decentralized shadow AI use, overconfidence in governance maturity, and fragmented regulation.

AI Weekly's regulation tracker page aggregates links to AI policy and regulation stories from various outlets along with their newsletter mention counts.

A Forbes analysis argues media buying is shifting from traditional celebrity endorsements toward creator-led, podcast, and gaming integrations that prioritize audience trust and attention metrics over raw reach.

IFA 2026 in Berlin drew a record 240,000 visitors from 155 countries and over 2,000 exhibitors, debuting a robot catwalk featuring machines from Unitree, AgiBot, and EngineAI alongside expanded AI and robotics programming.

A comparison of Lovable and Bolt.new for AI app development, recommending Lovable for fast nontechnical prototyping and Bolt.new for developers seeking greater code control.

Smartsheet describes its built-in content collaboration features and integrations with Adobe Creative Cloud and digital asset management tools.

Taylor Swift teased a 13-second clip of her upcoming 'Patient Zero' music video and announced a four-song expanded edition of her album The Life of a Showgirl.

Runway offers a browser-based suite of over 30 AI-powered creative tools, including Gen-1 and Gen-2 for video generation, aimed at making content creation accessible to all.

AlleyWatch's daily NYC funding report covers Rightway's $155M Series E, HIFI's $37M Series A, and Neurath's $1.52M AI reliability raise.

InvestorPlace commentary argues the AI trade is waking up and pitches five stocks—AMD, Arm, Meta, Sandisk, and one more—across chips, storage, and AI applications.

An investment analysis article highlights Pfizer, Medtronic, and Insmed as AI-driven healthcare stocks to watch, noting cost and execution risks.

A monthly archive of 2026 venture capital and private equity funding, deal, and transition news from Venture Capital Access Online.

Malwarebytes researchers report that scammers are using AI to build convincing fake antivirus renewal pages, such as an Avast-branded site targeting Belgian users, to harvest personal data and set up follow-up phone scams.

ET Edge Insights analyzes India's top five 2026 cyber-fraud models—from fake investment platforms and digital arrests to AI-enabled CEO impersonation via WhatsApp hijacking—highlighting how organized, believable scams exploit fear, urgency, greed, and authority.

SPEAR-Q, a decentralized authorization framework using quorum-based signed evidence from independent Policy Authority Nodes, was deployed on the Airbus Cyber Security Simulation Platform and shown to enforce sticky policies without a centralized Policy Decision Point.

Sikich announced it has been approved as a Qualified PIN Assessor by the PCI Security Standards Council, expanding its payment security assessment offerings for banks, processors, and payment networks.

A roundup page listing recent Indian healthcare, pharma, and biotech headlines covering acquisitions, earnings forecasts, regulatory warnings, and market expansions.

A Lexology PRO roundup landing page for AI legal and regulatory updates from 10-23 September, containing only a survey prompt rather than substantive article content.

Anthropic announced that its AI model Claude autonomously discovered a CRISPR-like enzyme system in bacterial DNA, drawing both excitement and scientific skepticism about whether it could lead to a new gene-editing mechanism.

Eli Lilly signed a research collaboration and licensing agreement with China's InnoCare Pharma worth up to about USD 3.35 billion for new drug discovery against up to five targets.