This Week in NLP #357
Keep up with what happened in NLP in the week ending Friday 19th September 2025.
Above the Fold
Overwhelmed? Here’s our pick for five things to know about this week.
Anthropic has refused to allow its AI models to be used by law enforcement agencies for surveillance, creating tension with the Trump administration. [Gizmodo]
Google has unveiled the Agent Payments Protocol, an open system enabling AI agents to make purchases across platforms with merchant and financial institution support. [TechCrunch]
Google has integrated Gemini AI across Chrome, offering features like contextual web browsing, task completion, and cross-app functionality to deepen user engagement with its ecosystem. [Gizmodo]
Microsoft and OpenAI have signed a nonbinding agreement to restructure their partnership as OpenAI transitions to a for-profit entity with continued nonprofit oversight. [ITPro]
Nvidia invested US$5B in Intel to develop custom x86 CPUs for AI infrastructure and future PC platforms, marking a surprising collaboration between the two tech giants. [TechRadar]
Now read on for everything else that happened in NLP this week.
Making News:
What makes this newsletter different?
This Week’s Topics:
If you’re reading this in a desktop web browser, you can access a navigation menu at the left-hand edge of this window.
The Generative AI Wars
Alphabet surged past US$3T in market value, driven by AI investments, favorable legal outcomes, and robust cloud growth amid tech sector enthusiasm. [WebProNews]
Anthropic acknowledged technical bugs affecting Claude’s code generation performance and launched an investigation into reported quality issues across its AI models. [The Decoder]
Google has strategically expanded Gemini’s AI capabilities through partnerships with Nvidia, Oracle, Salesforce, and others, enabling secure, flexible enterprise AI deployment across various industries and platforms. [WebProNews]
Google’s Gemini achieved gold-medal performance at the International Collegiate Programming Contest by solving 10 out of 12 complex coding problems. [ZDNet]
And Google’s Gemini mobile app rocketed to the top of US iPhone app charts, likely due to its new Nano Banana AI image editing feature. [ZDNet]
Microsoft has integrated Copilot Chat into its 365 business apps, offering AI assistance in Word, Excel, PowerPoint, Outlook, and OneNote for subscribers. [ZDNet]
And Microsoft is introducing Copilot AI to US House staff as part of a broader effort to integrate AI into congressional operations. [Gizmodo]
OpenAI projected reducing revenue sharing with Microsoft to 8% by decade’s end, potentially retaining over US$50B in additional revenue. [DealStreetAsia]
OpenAI has been recruiting robotics researchers and exploring humanoid systems as part of its efforts to develop more advanced AI algorithms for physical world interaction. [Wired]
OpenAI’s preliminary agreement with Microsoft could pave the way for restructuring and a potential future public offering. [Fortune]
And OpenAI's AI system solved all 12 problems at the ICPC World Finals 2025, outperforming human teams and Google's Deepmind model. [The Decoder]
AI Supremacy
Alibaba and Baidu began training AI models on domestically developed chips, signaling China’s strategic push to reduce dependence on US technology amid escalating geopolitical tensions. [Firstpost]
China has launched a series of regulatory actions targeting American chipmakers, potentially using semiconductor market leverage in negotiations over TikTok and trade restrictions. [Wired]
And China’s Ant Group criticized US tech giants for using open-source tools to lock developers into closed-source AI ecosystems, highlighting market dominance concerns. [TechRadar]
Futurewei, allegedly a Huawei subsidiary, shared office space with Nvidia, sparking concerns about potential Chinese technological espionage. [TechRadar]
Huawei is expanding its AI and cloud businesses in developing countries after being barred from Western markets by US-led restrictions. [Rest of World]
Malaysia is slowing data centre expansion, potentially hindering China’s access to advanced AI chips and impacting its technology ambitions in Southeast Asia. [DealStreetAsia]
Nvidia experiences challenges in China after a reported ban on its AI chips, with CEO Jensen Huang expressing disappointment while maintaining a diplomatic stance. [CNBC]
Nvidia's CEO Jensen Huang argues that the US tech stack should become a global standard, comparable to the dominance of the US dollar in international finance. [Yahoo Finance]
And Nvidia faces an antitrust investigation by China over its 2020 Mellanox acquisition, further straining US-China semiconductor trade tensions. [TechCrunch]
US Interior Secretary Doug Burgum argued that winning the AI arms race is more critical than addressing climate change, advocating for natural gas-powered datacenters to ensure US technological dominance. [The Register]
Sovereign AI
Governments worldwide are pursuing sovereign AI, but face significant challenges in achieving true independence due to the dominance of American and Chinese tech firms in AI infrastructure. [Rest of World]
Cohere has expanded its Paris office to 40 people, aiming to capitalize on growing European enterprise AI opportunities and compete with local players like Mistral AI. [Sifted]
Carbon3.ai aims to build a UK-wide sovereign AI infrastructure of over 30 datacentres powered by renewable energy, supporting the nation’s AI ambitions. [Computer Weekly]
CommonAI launched a collaborative AI platform to help UK and European startups access critical resources, IP, and strategic investment for global competitiveness. [Tech.eu]
The Cortex Hub launched the MCP Hackathon Africa 2025, challenging developers across 40 cities to embed local languages and cultural contexts into AI infrastructure. [News Ghana]
The European Union is pursuing a third AI technology stack through strategic partnerships with like-minded countries, aiming to create a democratic alternative to US and Chinese AI dominance. [The Brookings Institution]
Google is investing £5B in the UK over two years, focusing on AI, infrastructure, and scientific research with a new Hertfordshire data center. [TechRadar]
Google is investing in connectivity, AI tools, and skills training to empower Africa’s youth and bridge the digital divide, with plans to reach 500 million Africans by 2030. [Google]
IndiaAI Mission has selected eight firms, including BharatGen and Tech Mahindra, to advance India’s sovereign AI model development for the second phase. [Analytics India Magazine]
Lebanon has established a new Ministry of Technology and Artificial Intelligence to transform governance, attract investors, and leverage diaspora innovation through digital modernization. [TahawulTech.com]
Nigeria and other developing nations are challenging Big Tech’s data dominance by demanding local data storage and seeking economic benefits from their citizens’ digital information. [Scroll]
Indian think tank NITI Aayog is strategically exploring 20 frontier technologies to position India as a global leader in AI, biology, energy, and quantum computing convergence. [Mint]
The UK government has secured a £30B tech investment deal with US companies, raising questions about potential concessions on digital services tax and copyright. [The Daily Mail]
The UK government is encouraging public sector organizations to use US-based cloud services, contrary to Europe’s push for homegrown cloud providers, raising concerns about data sovereignty and domestic tech innovation. [Computer Weekly]
The UK government signed a US partnership to develop Europe’s largest AI factory, committing to attract talent and deploy 120,000 GPUs over the next 12 months. [Computer Weekly]
CoreWeave committed £1.5B to develop sustainable AI infrastructure in the UK, partnering with Nvidia and DataVita to advance technological innovation. [Business Wire]
Microsoft and Nvidia announced a US$45B investment in the UK’s AI infrastructure and research, aimed at bolstering the country’s technological capabilities. [Wired]
OpenAI is preparing a multibillion-dollar investment in UK AI infrastructure with Nvidia and Nscale Global Holdings, targeting significant data center expansion. [The Decoder]
Nvidia announced a £11B investment to deploy 120,000 Blackwell GPUs across the UK by 2026, establishing Europe’s largest GPU cluster. [ITPro]
Brent Hoberman, co-founder of lastminute.com and executive chairman of Founders Forum Group, welcomed US tech giants’ AI investments in the UK but warned against becoming mere technology users rather than innovators. [Tech.eu]
Protesters gathered in London to denounce Trump’s state visit, criticizing the UK-US AI deal and expressing anger over perceived government capitulation to tech giants and far-right politics. [Wired]
And ex-Meta exec Nick Clegg criticized the US-UK tech deals as ‘sloppy seconds from Silicon Valley’ during Trump’s state visit. [The Guardian]
Feature Creeps
Anthropic introduced an automatic memory feature for Claude AI, allowing Team and Enterprise users to have conversations with context retention without manual prompting. [The Verge]
And Anthropic has released a free incognito mode for Claude, allowing users to have private, unsaved conversations without impacting the AI’s memory features. [TechRadar]
Google Meet introduced Ask Gemini, an AI assistant providing real-time meeting summaries, notes, and key takeaways for Workspace users. [ZDNet]
Google is enabling users to share custom AI assistants called Gems with others, similar to sharing files on Google Drive. [TechCrunch]
Microsoft has added MAI-Voice-1 audio modes to Copilot, offering emotive, story, and scripted voice options with diverse performance styles. [The Decoder]
Microsoft will automatically install the Copilot app on Windows devices with Microsoft 365 desktop client apps in October, with exceptions for European Economic Area users. [TechRadar]
Microsoft is adding AI features like Summarize, Write, and Rewrite to Notepad, with Copilot+ PC users able to run local models without a subscription. [The Register]
Mozilla's Firefox 143 update added Microsoft Copilot to its AI sidebar, introduced Google Lens image search, and included several web app and accessibility improvements. [PCWorld]
OpenAI updated ChatGPT’s personalization hub with personality options and memory toggles, sparking mixed reactions from users seeking a more intuitive AI experience. [ZDNet]
And OpenAI has introduced a Developer Mode for ChatGPT, enabling Plus and Pro users to connect remote servers with powerful but potentially risky MCP tools. [The Decoder]
Zoom unveiled AI-powered features at Zoomtopia, including cross-platform meeting tools, AI avatars, scheduling assistance, and enhanced productivity capabilities. [TechCrunch]
Hype Bubble?
Forbes reports that, despite massive investments and hype, AI technologies remain largely untrustworthy, with surveys showing widespread skepticism about accuracy, reliability, and real-world business impact. [Forbes]
The Economist suggests that the AI investment boom could lead to massive financial losses, even if the technology succeeds. [The Economist]
Anthropic's CEO Dario Amodei’s prediction about AI writing 90% of code within six months has not materialized, despite growing AI code generation in tech companies. [ITPro]
SMBs are increasingly turning to AI to solve productivity challenges, with nearly half planning to adopt the technology in the next 12 months despite skills shortages. [TechRadar]
Meanwhile, the Commonwealth Bank of Australia backtracked on AI-driven job cuts after discovering its chatbot failed to reduce customer service call volumes as initially predicted. [ITPro]
And Microsoft is struggling to definitively prove Copilot’s productivity gains and return on investment, despite claiming 70 percent of Fortune 500 companies are using the AI tool. [The Register]
Bret Taylor, board chair at OpenAI and CEO of AI agent startup Sierra, affirmed OpenAI’s view that AI represents a transformative economic opportunity despite being in a potentially speculative bubble similar to the dot-com era. [TechCrunch]
A survey by Salesforce revealed that by 2027, AI agents are expected to resolve half of all service calls, with 79% of service leaders viewing AI investment as essential. [ZDNet]
The UK government’s AI coding assistant trial revealed significant productivity gains, with developers saving around one hour daily across 50 departments. [ITPro]
Plus: we’re seeing the emergence of ‘vibe coding cleanup specialists’ who fix problematic AI-generated code for companies struggling with automated software development. [Gizmodo]
Big Iron
Compu Dynamics Modular unveiled two purpose-built modular data center platforms designed to address AI, edge, and hyperscale infrastructure challenges. [Business Wire]
CoreWeave signed a US$6.3B order with Nvidia, guaranteeing the chipmaker will purchase any unsold cloud capacity through 2032. [Reuters]
Google has opened its first UK data center in Waltham Cross, investing £5B in local infrastructure and AI-powered services. [ITPro]
Microsoft is building a massive AI datacenter in Wisconsin, investing US$7B to create jobs, support local communities, and advance sustainable technology innovation. [Microsoft]
Microsoft secured a US$6.2B deal with Nscale and Aker to rent renewable AI computing power in Norway. [Bloomberg]
Nvidia has shifted its cloud strategy from DGX Cloud to Lepton, a GPU marketplace that routes AI workloads across partner infrastructure instead of directly competing with cloud providers. [Tom’s Hardware]
The Chips are Up
Alibaba demonstrated its PPU AI chip as comparable to Nvidia's H20 in a state TV broadcast, though performance claims remain unverified without independent benchmarks. [Tom’s Hardware]
AMD released its ROCm 7.0 software, promising significant performance improvements in AI workloads and narrowing the gap with Nvidia's CUDA ecosystem. [The Register]
Huawei has been producing AI chips using stockpiled dies, but faces a critical shortage of high-bandwidth memory that could severely limit future production. [The Decoder]
Huawei announced massive AI SuperPoDs and superclusters with up to 1 million NPUs, challenging Nvidia and AMD's AI computing dominance. [TechRadar]
Huawei unveiled a multi-year roadmap for increasingly powerful AI accelerators, aiming to provide domestic alternatives to Nvidia chips amid US export restrictions. [The Register]
Huawei has developed its own HBM memory for AI chips, aiming to break US sanctions and compete with Nvidia by enhancing processor performance. [Techzine]
Nvidia is facing challenges in China as Beijing discourages local tech companies from purchasing its chips, pushing for domestic AI chip development. [Gizmodo]
Nvidia's Rubin CPX AI GPU sparked speculation about a potential GeForce RTX 6090 with advanced features and significantly increased performance capabilities. [TechRadar]
Nvidia's RTX6000D, priced at $7,000, failed to attract Chinese tech companies due to high cost and underwhelming performance compared to banned alternatives. [TechPlugged]
SK Hynix has completed HBM4 development, preparing high-volume production to meet next-gen GPU memory demands for AI accelerators. [The Register]
Warm Bodies
Fiverr announced massive layoffs of 250 staff members as part of its transformation into an AI-focused company, aiming to become leaner and more productive. [Gizmodo]
Google has laid off over 200 AI product contractors, part of a broader trend of tech companies cutting workers while investing heavily in AI development. [Gizmodo]
xAI has reportedly deactivated Slack accounts of several high-level data annotation team members responsible for training its Grok AI chatbot. [Futurism]
And xAI has laid off 500 data annotation team members, shifting focus to specialized AI tutors while expanding in specific domains. [TechCrunch]
Consumer AI
Brilliant Labs partnered with Liquid AI to integrate vision-language foundation models into its Halo AI glasses, enhancing scene understanding and user interaction. [Business Wire]
Meta unveiled Ray-Ban Display smart glasses with a wristband-controlled display for apps, alerts, and directions, priced at $799. [TechCrunch]
And Meta unveiled the Oakley Meta Vanguard smart glasses, targeting athletes with advanced features like 3K video capture and integrated fitness tracking. [TechCrunch]
Omdia forecasts AI glasses market growth to reach 10 million units by 2026, and expected to hit 35 million units by 2030. [Business Wire]
PayPal partnered with Google to develop AI-powered shopping experiences and integrate payment solutions across Google’s products and platforms. [TechCrunch]
Rabbit released rabbitOS 2, transforming the R1 with a card-based interface, multimodal AI capabilities, and user-friendly ‘vibe-coding’ for creating custom experiences. [TechRadar]
SK Networks partnered with Kardome to launch NAMUHX, an AI-powered air purifier robot with advanced voice interaction capabilities. [Speech Technology Magazine]
Snap is developing lightweight AR smart glasses with Snap OS 2.0, offering improved browsing, content sharing, and developer-created experiences. [ZDNet]
It’s Only a Model
Alibaba unveiled Qwen3-Next, an innovative 80B-parameter AI model with hybrid attention and sparse MoE architecture, delivering unprecedented efficiency and performance across long-context understanding and inference tasks. [Alibaba]
And Alibaba's Tongyi Lab unveiled an open-source DeepResearch Agent that matches or exceeds proprietary AI research tools, demonstrating significant advancements in Chinese generative AI capabilities. [VentureBeat]
Baidu released ERNIE-4.5-21B-A3B-Thinking, a computationally efficient 21B-parameter Mixture-of-Experts language model with 128K context length and advanced reasoning capabilities. [Marktechpost Media]
Deepdub launched Lightning 2.5, a high-performance voice AI model enabling real-time, multilingual voice applications with improved efficiency and low latency. [Marktechpost Media]
DeepSeek developed R1, a cost-effective LLM trained for $294,000, using reinforcement learning to improve reasoning without extensive human annotation. [Gizmodo]
Google's Nano Banana, a text-to-image AI model, has rapidly gained industry-wide adoption for its advanced editing and consistency capabilities. [EIN Presswire]
Google DeepMind released VaultGemma 1B, the first LLM trained entirely with differential privacy, demonstrating a path to more privacy-preserving AI. [Marktechpost Media]
IBM has released Granite-Docling-258M, an open-source vision-language model designed for precise, structure-preserving document conversion across multiple formats and languages. [Marktechpost Media]
Meta has released MobileLLM-R1, a family of lightweight edge reasoning models optimized for efficient mathematical, coding, and scientific tasks with sub-billion parameters. [Marktechpost Media]
OpenAI launched GPT-5-Codex, a specialized AI coding assistant designed to autonomously handle complex software engineering tasks across various development environments. [VentureBeat]
TwinMind unveiled Ear-3, a Voice AI speech-recognition model with state-of-the-art performance across 140+ languages at a competitive price point of $0.23/hr. [Marktechpost Media]
xAI introduced Grok 4 Fast, a speed-optimized AI model offering rapid responses at the cost of nuanced output, now available on their web interface. [TestingCatalog]
AI startups are increasingly viewing foundation models as commodities, focusing instead on customization, interface design, and specific task optimization rather than model development. [TechCrunch]
Whose Data?
The Britannica Group, parent of Encyclopedia Britannica and the Merriam-Webster dictionary, sued Perplexity for allegedly copying its copyrighted content without permission. [Gizmodo]
Meta has been exploring licensing news content from media companies like Axel Springer, Fox Corp, and News Corp for its AI tools. [The Wall Street Journal]
People Inc accused Google of unfairly using the same crawler to index search results and train AI models, potentially harming publishers’ content and traffic. [TechCrunch]
And Google faces a lawsuit from Penske Media Corporation, alleging illegal use of news content in AI summaries that potentially harm publishers’ business models and revenue streams. [TechCrunch]
Meanwhile, Reddit is negotiating a comprehensive AI content-sharing agreement with Google to leverage its user-generated data and enhance platform growth. [Bloomberg]
The LLM Ecosystem
Calsoft has launched a Synthetic Data Curation service powered by LLM agents, enabling enterprises to generate compliant, domain-specific datasets at scale for AI. [EIN Presswire]
DEV.co expanded its software development services to help enterprises implement custom, privacy-focused open source LLMs for various business applications. [EIN Presswire]
GibsonAI developed Memori, an open-source memory engine that provides persistent, intelligent memory for AI agents using standard SQL databases, addressing the stateless nature of modern language models. [Marktechpost Media]
Google has donated its Secure AI Framework to the Coalition for Secure AI, advancing industry-wide AI security standards and collaborative development. [EIN Presswire]
GreyNoise Intelligence introduced an MCP Server enabling AI agents to access real-time threat intelligence, enhancing cybersecurity workflows through autonomous reasoning and actionable data. [PRWeb]
Karini AI integrated with Amazon Q Index, enabling enterprises to build secure, scalable agentic AI applications directly on AWS without data replication risks. [EIN Presswire]
MITRE and FAA developed the Aerospace Language Understanding Evaluation (ALUE) benchmark to rigorously assess LLMs for safety-critical aerospace applications. [Business Wire]
Neurotechnology has released an AI SDK for multilingual NLP, enabling developers to build proprietary speech recognition and analysis solutions. [EIN Presswire]
Progress Software expanded its AI Coding Assistants across UI libraries, reporting tools, and document processing, enhancing developer productivity and workflow efficiency. [GlobeNewswire]
Promptler has introduced a native AI prompt management solution for Apple devices, offering privacy-first, device-specific optimization. [EIN Presswire]
Qlik has launched Open Lakehouse, a fully managed Apache Iceberg service providing enterprises with rapid, AI-ready data access and optimization. [Business Wire]
Span has launched a universal AI code detector that accurately identifies AI-assisted code, helping technology leaders measure and understand the impact of AI coding tools. [Business Wire]
Thinking Machines Lab developed batch-invariant kernels to achieve deterministic LLM inference by addressing nondeterminism caused by varying batch sizes in parallel computing environments. [Thinking Machines]
Agentic AI
Amazon has introduced an always-on AI agent, Seller Assistant, to help third-party sellers manage their businesses more efficiently and strategically. [TechCrunch]
And Amazon Ads launched an AI tool in Creative Studio that enables advertisers to rapidly generate professional-quality, multi-format ad campaigns using retail insights. [Search Engine Land]
Auquan launched an AI agent that autonomously performs comprehensive credit analysis and monitoring. [Business Wire]
CrowdStrike has partnered with Nvidia to develop Charlotte AI AgentWorks, enabling secure, no-code AI agent creation and deployment across enterprise ecosystems. [Business Wire]
Dotfile launched Autonomy, an AI-powered platform that autonomously reviews KYB compliance cases, reducing processing time and false positives. [Business Wire]
Ericsson has integrated agentic AI into its NetCloud platform, enabling autonomous 5G enterprise networks with advanced troubleshooting and management capabilities. [Computer Weekly]
Gartner has positioned AI agents atop its 2025 Hype Cycle for Emerging Technologies. [BigDATAwire]
Monday.com expanded its AI capabilities with new agents, CRM tools, and enterprise features, aiming to transform work management through intelligent, end-to-end automation. [Business Wire]
Moody’s launched Agentic Solutions, an AI-powered platform designed to automate complex workflows and enhance decision-making across regulated industries. [Business Wire]
Notion has launched its first AI agent, capable of generating notes, reports, and dashboards by drawing on users’ pages and databases across multiple platforms. [TechCrunch]
Osmos.io has partnered with Databricks to launch an AI Data Engineer that automates complex data engineering tasks while maintaining data integrity and operational safety. [EIN Presswire]
Pinkfish launched retail-ready AI agents to automate marketing, merchandising, and commerce workflows for enterprises seeking simplified AI adoption. [Business Wire]
Revolut has been exploring AI-powered agents to automate customer service, sales, and other financial tasks as part of its superapp strategy. [Sifted]
Sapio Sciences launched ELaiN, an AI-powered electronic lab notebook that acts as a collaborative scientific co-researcher, transforming traditional data recording methods. [Business Wire]
Socotra released its MCP Server, enabling insurers to quickly and securely integrate agentic AI into their insurance workflows. [Business Wire]
Soul Machines launched a Workforce Connect App on ServiceNow's platform to enable emotionally intelligent AI agents that automate and enhance enterprise workflows. [Business Wire]
xSIGNAL has developed Agentic Intelligent Messaging, a ticketless workflow solution to help federal agencies optimize processes and reduce manual work. [EIN Presswire]
Zendesk has leveraged GPT-5 to dramatically improve its AI customer service agents, enabling them to solve up to 90% of support tickets with increased accuracy, speed, and reasoning capabilities. [VentureBeat]
Other LLM Sightings
AlertMedia enhanced its Risk Intelligence suite with AI-powered tools to accelerate threat detection, social monitoring, and incident response across global organizations. [Business Wire]
Behavox released GPU-powered AI Risk Policies that reduce compliance alerts by over 60% while improving detection rates across 15+ languages. [Business Wire]
Calabrio launched Workforce Intelligence, an AI-powered workforce management solution designed to enhance contact center agility, performance, and agent experience. [Business Wire]
Carly launched an AI-powered personal assistant that schedules meetings via email and SMS with human-level efficiency. [Business Wire]
CENTRL has launched an AI-powered Reporting and Research Assistant to transform due diligence data into actionable intelligence for investment teams. [PRWeb]
Deep Instinct has released DIANNA, an AI-powered tool that provides instant, expert-level explainability of never-before-seen cybersecurity threats in under 10 seconds. [Business Wire]
Elaris launched an AI platform that instantly provides marketers with audience insights, enabling quick content creation and strategy development through psychological data and synthetic audience testing. [EIN Presswire]
GFT has developed a generative AI credit risk assistant that automates financial data analysis, reducing report creation time and enabling more complex risk decisions. [Business Wire]
Konnect launched KonnectAi, an AI-powered HR assistant named Sophie, designed to streamline HR tasks and empower professionals by handling repetitive work. [Business Wire]
PageUp launched an AI-powered talent acquisition suite with Paige, an AI co-pilot designed to streamline recruiting while maintaining fairness and personal touch. [GlobeNewswire]
Proofpoint launched an AI-powered Human Communications Intelligence solution that transforms digital communications governance by interpreting intent across multiple channels in real time. [Business Wire]
Smart Capital Center has launched an AI-powered financial analysis tool that automates variance reporting for commercial real estate teams. [EIN Presswire]
Stringr launched VidGen, an AI-powered video solution enabling publishers to transform articles into branded, multi-format videos quickly and easily. [Business Wire]
Synametrics Technologies launched SyncriTix, an AI-powered email ticketing system that automates customer support through intelligent ticket management. [EIN Presswire]
TheySaid launched an AI-powered feedback platform that transforms traditional surveys into real-time, conversational intelligence for businesses. [EIN Presswire]
TTA launched an AI-powered talent search platform, enabling organizations to quickly match with vetted learning and development professionals. [EIN Presswire]
Risks and Responses
California passed a groundbreaking bill requiring AI companies to implement safeguards for minors, signaling growing regulatory scrutiny of AI companionship’s potential psychological risks. [MIT Technology Review]
Cyberspike's Villager, an AI-powered penetration tool linked to a suspicious Chinese company, was downloaded nearly 10,000 times since July, offering automated cyber attack capabilities. [The Register]
DeepSeek's AI reportedly discriminates against Falun Gong-affiliated programmers by refusing help or providing flawed code. [The Washington Post]
Fireflies has released the industry’s first comprehensive guide for responsibly deploying AI notetakers, emphasizing transparency, privacy, and human judgment in workplace conversations. [Business Wire]
Microsoft's research suggests that knowledge workers increasingly rely on AI tools, potentially diminishing their critical thinking skills by offloading cognitive processes to generative AI technologies. [Undark]
Mitsubishi Electric and Inria launched a joint research project to develop trustworthy AI systems using formal methods and mathematical approaches. [Business Wire]
OpenAI has called for hardware-level kill switches and safety features to detect and mitigate potential risks in future AI infrastructure. [TechRadar]
OpenAI announced new policies to protect underage users from harmful interactions, including restrictions on sexual topics and self-harm discussions in ChatGPT. [TechCrunch]
OpenAI's Deep Research agent was found vulnerable to a prompt injection attack that could silently exfiltrate confidential Gmail data without user interaction. [Ars Technica]
Social Media Victims Law Center filed three lawsuits against Character.ai, alleging its chatbots harmed children through manipulative and sexually explicit interactions leading to suicide attempts and abuse. [Business Wire]
And parents of children who died by suicide after AI chatbot interactions have urged Congress to take action, highlighting growing concerns about tech companies’ responsibility for online safety. [Axios]
Texas and other states have proposed disconnecting energy-hungry data centers from power grids during emergencies to prevent potential blackouts and manage electricity demand. [Newsday]
The UN has initiated preliminary international AI governance mechanisms, revealing deep geopolitical tensions and highlighting urgent needs for comprehensive, rights-focused regulatory frameworks. [Inter Press Service]
University of California Riverside researchers developed a method to retrain AI models, preserving safety guidelines when models are reduced for smaller devices. [TechRadar]
The US FTC ordered seven tech companies to provide details on how their chatbots prevent potential harm to children. [Mashable]
Educators are increasingly concerned about students’ over-reliance on AI tools, which may undermine critical thinking and foundational skills development in computer science and other disciplines. [ZDNet]
Regulation
California’s state senate approved SB 53, a bill requiring large AI companies to disclose safety protocols and create whistleblower protections, now awaiting Governor Newsom’s signature. [TechCrunch]
India has prioritized innovation over regulation in technology policies, with the government focusing on AI development and national tech initiatives. [EdexLive.com]
India’s Finance Minister Nirmala Sitharaman emphasized the urgent need for evolving AI regulations that keep pace with technological advancements while serving the common good. [Latestly]
Environmental Issues
FuriosaAI demonstrated OpenAI's chatbot running on its custom RNGD chips in Seoul, showcasing sustainable enterprise AI without traditional GPU infrastructure. [TechRadar]
OpenAI and other tech companies are exploring space-based data centers as a potential solution to the environmental challenges posed by terrestrial data centers’ massive energy consumption and resource strain. [Gizmodo]
The UK’s National Grid partnered with Emerald AI to test dynamic energy management for AI datacentres, aiming to optimize grid capacity and support technological growth. [Computer Weekly]
Conversational AI
800.com launched 800 Intelligence, an AI toolkit that analyzes calls, provides insights, and helps businesses improve customer interactions and team performance. [Business Wire]
ASAPP has enhanced its GenerativeAgent Platform, enabling enterprises to build, deploy, and govern AI agents with increased speed, confidence, and scalability for transforming customer interactions. [GlobeNewswire]
Five9 launched Fusion for ServiceNow, an AI-powered integration unifying voice and digital interactions to deliver more efficient, personalized customer service experiences. [Business Wire]
A survey by Gartner revealed that Fortune 500 companies are not planning wholesale replacement of human support staff with AI, recognizing the continued importance of human interaction. [The Register]
dotstudioPRO launched a multi-language AI CMS Assistant that transforms video operations by enabling conversational content management across multiple platforms. [EIN Presswire]
Genifea has developed an AI voice agent specifically for estate agencies, automatically handling calls, qualifying leads, and integrating them directly into CRM systems. [EIN Presswire]
Lightspeed Voice has introduced Sentiment Alerts, an AI feature enabling businesses to quickly identify and address customer interaction sentiments. [PRWeb]
MomentScience launched NexosChat, a conversational AI platform designed to help businesses create personalized, high-engagement post-transactional advertisements. [Business Wire]
PRIMEDOT Media launched AI-powered voice and chat solutions to help service businesses capture leads, improve response times, and automate customer interactions. [EIN Presswire]
reAlpha Tech Corp has upgraded its AI assistant Claire to serve as a digital homebuying concierge, providing personalized guidance and support throughout the real estate transaction process. [GlobeNewswire]
Supportwave has developed an AI-powered recruitment platform that reduces tech hiring time by 90%. [Grit Daily News]
Be Real
Albania has introduced Diella, an AI-powered chatbot, to review government contracts and combat corruption in public procurement. [Gizmodo]
Faith tech apps are attracting millions of users seeking spiritual guidance through AI-powered chatbots trained on religious texts. [Ars Technica]
Research by Sapia.ai revealed that over one million candidates found its AI chat interviews more human and engaging than traditional recruitment processes. [Business Wire]
Zoom is developing AI Companion 3.0, introducing cross-platform functionality, AI-generated avatars, and enhanced meeting tools across video conferencing platforms. [TechRadar]
Voice News
Amazon has introduced ‘Hear the Highlights,’ an AI-powered feature generating audio product summaries by synthesizing reviews, catalog information, and web sources for US customers. [CNBC]
Behavioral Signals launched a real-time deepfake speech detection platform that uses signal and behavioral intelligence to verify voice authenticity across enterprise environments. [PRWeb]
Housecall Pro has introduced voice-activated invoicing for home service technicians, enabling hands-free billing and payment processing through its software platform. [Speech Technology Magazine]
NLX expanded its Voice+ technology to enable hands-free, conversational navigation of websites and mobile apps through spoken commands. [Speech Technology Magazine]
Presto partnered with ElevenLabs to enhance drive-thru voice AI experiences with more realistic, human-like voice synthesis technology. [Business Wire]
Speechmatics achieved a record 93% accuracy in medical Speech-to-Text, delivering 50% fewer keyword errors and 17% lower overall word errors than competitors. [Business Wire]
TNS launched Enterprise Voice Security, a comprehensive solution protecting businesses from contact center fraud through advanced call authentication and monitoring technologies. [Business Wire]
Document AI
A survey by SER revealed that 65% of companies are accelerating intelligent document processing projects, focusing on productivity and front-office applications. [Business Wire]
Artificio has developed a generative AI PDF design platform that automates enterprise document creation with advanced AI, brand consistency, and intelligent workflow integration. [EIN Presswire]
Box is introducing AI-powered features to extract, analyze, and secure unstructured data stored in its cloud service, enhancing organizational productivity and information management. [Computerworld]
Canopy expanded its Smart Intake feature with AI-driven document request lists, enabling accounting firms to quickly generate tailored client document requests. [Business Wire]
Datalogz launched BI Similarity, an innovative toolkit that automatically identifies and eliminates duplicate reports across multiple business intelligence platforms. [Business Wire]
Diplomat Consulting and Punchcard Systems partnered to develop LegEngine, an AI-powered platform designed to streamline legislative document analysis and policy development. [Business Wire]
PandaDoc launched an AI-native MCP infrastructure with enhanced compliance, flexible pricing, and advanced document automation capabilities to challenge e-signature market leaders. [Business Wire]
ParaScript partnered with BIRGER to bring AI-powered check processing technology to Mauritian banks, marking its first entry into the African market. [Business Wire]
Tungsten Automation launched an enhanced platform integrating AI-driven document automation, workflow orchestration, and knowledge discovery. [Business Wire]
Translation
Commence and The Language Group launched Fetch, an AI-powered translation solution that integrates real-time, human-reviewed translations into EHR systems to improve patient care. [Business Wire]
Plint has launched Plint SUB, an AI-powered subtitling solution offering 93% accuracy, faster delivery, and significant cost savings for media localization. [Slator]
RWS has launched Smart Insights, an AI-powered virtual assistant within its Trados platform that provides real-time project intelligence for localization professionals. [Slator]
Search
Archetype, Nectar and Outcast launched an AI Visibility offering to help brands optimize and track their presence in LLM search results. [Business Wire]
G2 unveiled innovative solutions to help software companies enhance brand visibility, track AI citations, and influence buyer decisions in the evolving B2B research landscape. [Business Wire]
Google has integrated ads directly into AI-generated search answers, potentially transforming its role from search middleman to content provider and digital advertising powerhouse. [The Decoder]
Google has launched an experimental Windows app that enables quick, comprehensive searching across files, apps, Drive, and the web using an Alt + Space shortcut. [TechCrunch]
AI in Journalism
Business Insider has issued guidelines for staff to use AI as a tool for research and content enhancement, with specific rules for writing. [Status]
Gannett has launched DeeperDive, an AI-powered chatbot tool that engages readers by summarizing journalism and suggesting content across its network of publications. [Wired]
Pangram has integrated its AI detection technology into HARO and Featured, enabling journalists to identify AI-generated content and make more informed sourcing decisions. [Business Wire]
Health Tech
Allot developed an AI agent using SnapLogic's platform to analyze healthcare data and identify asthma health inequalities for pharmaceutical companies. [Business Wire]
Concord launched a white-glove Ignition Program to help healthcare organizations seamlessly transition to AI-driven document processing workflows. [Business Wire]
CSSi and HumanTrue have partnered to revolutionize clinical trial recruitment by leveraging AI for faster screening, real-time translations, and smarter site support. [PRWeb]
Doctronic developed an AI-powered healthcare platform providing rapid symptom assessment and physician access, handling 50,000 weekly medical conversations. [PYMNTS]
eClinicalWorks and Sunoh.ai helped Sun River Health streamline clinical documentation using an AI medical scribe that captures patient-provider conversations in real-time. [Business Wire]
Evidently launched Ask Evidently, an AI chat interface embedded in electronic health records that helps clinicians quickly access and understand patient information. [Business Wire]
Hippocratic AI has collaborated with University Hospitals, an integrated hospital network, to deploy conversational AI agents for non-diagnostic patient support and clinical workflow enhancement. [MobiHealthNews]
MEDITECH is advancing its AI initiatives to create an agentic user experience that supports clinicians, patients, and administrators through intelligent, proactive EHR workflows. [Business Wire]
Qventus launched an AI Solution Factory to help health systems co-develop custom AI operational assistants that solve critical challenges and drive operational efficiency. [Business Wire]
Samsung is developing an AI health coach in its Health app to provide wellness guidance and activity recommendations without medical diagnosis. [ZDNet]
ScribeEMR integrated its AI medical scribing technology with Epic's Ambient Voice Recognition Module, enabling clinicians to automatically generate and review patient encounter notes. [EIN Presswire]
Stanford researchers have developed MedAgentBench, a benchmark testing AI agents’ ability to perform clinical tasks in electronic health records, revealing current limitations and potential for augmenting healthcare workforce. [Stanford U]
Legal Tech
Assembly Software has released NeosAI Reasoning Mode, offering legal professionals two AI-driven analysis modes for comprehensive case management and document review. [GlobeNewswire]
August and other AI tools have demonstrated comparable or superior contract drafting reliability compared to human lawyers, with potential to reduce legal drudgework while complementing human legal expertise. [Artificial Lawyer]
The California Court of Appeals sanctioned an attorney for AI-generated fake citations and suggested opposing counsel may have a duty to detect and report such fabrications. [LawSites]
CILEX has partnered with 5Mins.ai to launch an AI Academy, providing personalized AI training for legal professionals to enhance their skills and adapt to technological changes. [Artificial Lawyer]
FasterOutcomes has released Knowledge 2.0, an AI-powered platform that centralizes legal knowledge and provides instant, actionable insights for law firms. [PRWeb]
Intellistack launched Streamline CLM, a no-code contract lifecycle management solution. [Business Wire]
IP8 launched an AI-powered patent surveillance platform that helps IP teams identify infringement opportunities 10x faster and with 80% less investigation time. [PRWeb]
A study by LegalBenchmarks.ai revealed AI tools matched or exceeded human lawyers in contract drafting reliability, with some AI products outperforming legal professionals. [LawSites]
Lex Generalis launched a cost-effective legal service model for startups, universities, and technology innovators, to streamline IP and corporate legal needs. [Business Wire]
Lightbringer has developed an AI-powered platform that dramatically accelerates patent filing by combining automated drafting with expert attorney review. [Tech.eu]
Lovable's Fire Fairness, an AI-powered insurance claims platform, processed over 2,800 wildfire claims 360x faster with 99.8% cost reduction. [Artificial Lawyer]
Stanford Law School has launched Liftlab, a research project exploring how AI can reshape legal services to be more accessible, efficient, and equitable. [Artificial Lawyer]
Supio partnered with Thomson Reuters to bring AI-powered case preparation tools to personal injury law firms, aiming to streamline medical record review and improve case outcomes. [Business Wire]
TheFormTool urged state bar associations to mandate lawyer oversight of AI in legal practice, emphasizing the need for human responsibility in legal services. [Business Wire]
Thomson Reuters has developed Deep Research, an AI platform for legal research that prioritizes depth and accuracy over speed, using a multi-step process to provide nuanced, well-cited legal insights. [VentureBeat]
Wolters Kluwer enhanced CCH iKnowConnect with expert AI capabilities, providing tax and legal professionals in APAC with intelligent, context-aware research insights. [Business Wire]
And the Bluebook has introduced a controversial Rule 18.3 for citing AI-generated content, drawing widespread criticism from legal scholars for its confusing, impractical, and potentially unethical citation requirements. [LawSites]
Ed Tech
A survey by Copyleaks revealed that 90% of US students now use AI for academic purposes, with usage increasing and becoming a daily learning tool. [GlobeNewswire]
Geek Haus launched TutorFlow, an AI-powered learning management system that enables educators to create interactive online courses rapidly and effortlessly. [EIN Presswire]
Google has introduced Learn Your Way, an AI-powered study tool that personalizes learning material by adapting content to students’ interests and grade levels. [ZDNet]
And Google has launched a US$1B initiative to provide US college students free access to AI tools, training, and certifications through its AI for Education Accelerator program. [Forbes]
South Australia has rolled out EdChat, a Microsoft-developed AI app for high schools, aiming to provide safe and productive technological learning while addressing potential drawbacks. [ABC]
MaiaLearning has introduced AI-powered tools to personalize college and career planning, offering students tailored recommendations and educators efficient support. [EIN Presswire]
McGraw Hill released ALEKS for Calculus, an AI-powered digital learning solution that personalizes math education through adaptive, knowledge-gap-assessing technology. [Business Wire]
Oboe launched an AI-powered learning platform that creates personalized, multi-format courses on any topic using interactive and flexible educational approaches. [TechRadar]
Funding
AAA C(H+A)Rm has raised US$4m in seed funding to develop an AI-first orchestration platform enabling humans and autonomous agents to collaborate and scale together. [PRWeb]
Altan has raised US$2.5m to develop an AI-powered platform that autonomously designs, builds, and operates software with minimal human intervention. [Tech.eu]
Arch raised US$52m to develop an AI-powered platform streamlining private market investment management and reporting for institutional investors. [Business Wire]
ASML invested €1.3B in Mistral AI, securing an 11% stake to apply advanced AI to chip research and development. [TechRadar]
Centari has raised US$14m in Series A funding to develop its AI-powered Deal Intelligence Platform, which provides advanced insights for high-stakes transactions. [Artificial Lawyer]
CodeRabbit raised a US$60m Series B to help companies streamline code review by using AI to catch and provide feedback on AI-generated code errors. [TechCrunch]
Conduct, an enterprise AI start-up, secured US$12m in seed funding to modernize legacy ERP systems using cutting-edge agentic AI technology. [Tech.eu]
DRUID AI secured US$31m in Series C funding to expand its enterprise-ready agentic AI platform globally. [Tech.eu]
Eagl has raised €825,000 to develop an AI-native financial operations platform that automates accounting workflows and enhances data quality for finance teams. [Tech.eu]
Eloquent AI has raised US$7.4m in Seed funding to expand its AI-powered customer support automation for financial services. [FinSMEs]
Ethosphere secured US$2.5m in pre-seed funding to develop AI technology that enhances in-store retail interactions and supports frontline associates. [Business Wire]
Evolve has raised US$1m in seed funding to develop an AI-driven learning platform that transforms corporate training through adaptive, measurable, and efficient skill development. [Tech Funding News]
Groq secured US$750m in funding, doubling its valuation to US$6.9 billion while challenging Nvidia's AI chip dominance with its innovative language processing units. [TechCrunch]
Irregular, an AI security firm, secured US$80m in funding to develop advanced methods for detecting and preventing potential risks in AI models. [TechCrunch]
Isaacus, an Australian legal AI foundation model builder, secured $700,000 in pre-seed funding to develop sovereign, high-performance legal AI models for tech companies. [Artificial Lawyer]
Keplar raised US$3.4m in seed funding to use voice AI for conducting faster, more cost-effective customer research interviews. [TechCrunch]
Markup AI secured US$27.5m to launch Content Guardian AgentsSM, an AI-native platform providing automated oversight and guardrails for AI-generated content across enterprises. [Business Wire]
Micro1 has raised a US$35m Series A funding round to help AI companies find and manage high-quality human contractors for data labeling and training. [TechCrunch]
Moodwork secured €3.1m in funding to expand its AI-powered mental health platform addressing workplace well-being and stress reduction. [Tech.eu]
Nory has raised US$37m in Series B funding to enhance its AI-powered restaurant management platform and expand into the US market. [Tech.eu]
Nothing secured US$200m in Series C funding, valuing the smartphone startup at US$1.3 billion and positioning it for AI-driven expansion. [TechCrunch]
Nscale secured US$700m from Nvidia to build the UK’s largest AI supercomputer, partnering with Microsoft and OpenAI to deploy thousands of GPUs. [TechRadar]
PromptWatch secured €1.2M in seed funding to develop its Generative Engine Optimisation and AI visibility platform. [Silicon Canals]
Quack has raised US$7m in Seed funding to expand its proactive AI customer support platform for SaaS companies. [FinSMEs]
SEQUESTO secured €2.5m from Smartfin and Newion to expand its AI-powered platform for automating tender and RFP processes internationally. [Tech.eu]
Sophont raised US$9.22m in Seed funding to develop multimodal medical AI foundation models for healthcare and pharmaceutical applications. [FinSMEs]
Spara has raised US$15m in Seed funding to expand its AI-powered enterprise platform for engaging website visitors through chat, voice, and email. [FinSMEs]
Trially secured US$4.7m in seed funding and launched Margo, an AI platform designed to dramatically improve clinical trial patient recruitment and enrollment. [Business Wire]
Acquisitions
Capacity acquired KLaunch, a conversational AI company, to enhance its automation platform and expand its reach in key industries. [Speech Technology Magazine]
CrowdStrike acquired Pangea Cyber for US$260m to enhance AI security, offering comprehensive protection against prompt injection and risky AI interactions. [TechRadar]
Innovaccer acquired Story Health to develop AI agents that augment specialty care teams, enabling continuous patient engagement and reducing hospitalizations. [Business Wire]
Opper AI acquired FinetuneDB, a Stockholm-based platform for fine-tuning LLMs. [FinSMEs]
Thomson Reuters acquired Additive, enhancing its tax technology capabilities through an AI-powered document processing solution. [Thomson Reuters]
Workday announced its plan to acquire Sana, an AI knowledge tools company, for US$1.1B to enhance its enterprise AI platform for learning and work. [Silicon Canals]
There’s More
Gartner projects that global AI spending will surge to US$2T by 2026, driven by infrastructure expansion, AI services, and investments from tech giants worldwide. [ITPro]
OpenAI launched Grove, a five-week mentorship program for early-stage AI entrepreneurs, offering technical guidance and workshops to roughly 15 participants. [CNBC]
And research by OpenAI revealed that ChatGPT is primarily used for non-work tasks like writing, guidance, and information-seeking, with usage growing rapidly across diverse global demographics. [ZDNet]
Replit's new AI coding agent, Agent 3, sparked user outrage by unexpectedly increasing costs, particularly when editing existing code projects. [The Register]
The Back Cover
Here’s Laura Elliott and Scary Pockets with a cover of U2’s With or Without You.
If you know of a great cover version of a back-catalogue classic, drop an email to us at backcover@language-technology.com and we’ll consider it for inclusion here.
Got this from someone else?
Did you find this newsletter useful? if so, you might forward it to a friend; or you could email us at news@language-technology.com to tell us what you want more of.
Did you hate this newsletter? if so, you could forward it to an enemy; or you could email us at news@language-technology.com to tell us why – make it better!