This Week in NLP #369
Keep up with what happened in NLP in the week ending Friday 12th December 2025.
Above the Fold
Overwhelmed? Here’s our pick for five things to know about this week.
Accenture and Anthropic launch multi-year partnership to train 30,000 professionals, develop AI solutions, and drive enterprise AI innovation across industries. [Business Wire]
And Anthropic has overtaken OpenAI in enterprise generative AI spending, driven primarily by coding tools and representing a US$37B market that experts see as a boom rather than a bubble. [ZDNet]
A report by Comscore reveals that Google’s Gemini and xAI’s Grok are experiencing explosive growth in AI chatbot usage, outpacing ChatGPT. [ZDNet]
OpenAI has released GPT-5.2, a new AI model designed to excel at professional tasks and compete with Google’s Gemini 3 Pro. [Engadget]
And Nvidia’s potential sale of H200 AI chips to China is criticized as a strategic mistake that could significantly erode America’s compute advantage in the AI race. [Don’t Worry About the Vase]
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
Anthropic has launched Claude for Nonprofits, partnering with Vera Solutions to provide accessible AI technology for mission-driven organizations. [EIN Presswire]
Google has launched Deep Think, an advanced reasoning mode for AI Ultra subscribers that explores multiple hypotheses simultaneously to solve complex problems more effectively. [9to5Google]
Google denies plans to add ads to Gemini, but the tech industry suggests advertising in AI chatbots is increasingly inevitable. [TechRadar]
Google launched Disco, an experimental AI-powered web browser that creates interactive GenTabs widgets from open tabs and user prompts. [Engadget]
Google launched a natural language Developer Assistant for its Ads API, enabling developers to generate, run, and export queries using plain English prompts. [Search Engine Land]
And Google launched an affordable AI Plus plan in India at ₹199 per month, offering expanded Gemini features and storage for users. [TechCrunch]
A Sensor Tower report shows ChatGPT’s growth is slowing while Google’s Gemini is gaining ground in the AI chatbot market, driven by its new image generation model. [TechCrunch]
Meta has hired two top Apple designers, Alan Dye and Billy Sorrentino, to lead a new design studio focused on improving the company’s user interfaces and AI hardware. [Wired]
Meta is training its new AI model Avocado using rival models and shifting from open-source to closed models to improve profitability and competitiveness. [Sherwood]
And Meta is integrating real-time news from multiple media outlets into its AI assistant, aiming to connect users with current information. [The Decoder]
OpenAI is reportedly shifting focus from AGI to improving ChatGPT’s quality and user engagement amid intense competition with Google’s AI offerings. [Futurism]
OpenAI is fast-tracking GPT-5.2’s release this week to compete with Google’s Gemini 3, focusing on improved performance and reliability. [TechRadar]
OpenAI faced backlash after ChatGPT users mistook app integration links for ads, prompting the company to clarify and temporarily disable the feature. [Gizmodo]
OpenAI reported surging enterprise AI usage, with ChatGPT message volume growing 8x and workers saving up to an hour daily through custom AI tools. [TechCrunch]
The Pentagon has deployed Google’s Gemini AI to GenAI.mil, providing 3 million defense workers with secure AI tools for workflow automation. [eWeek]
Yann LeCun revealed he is leaving Meta at year’s end and will not receive financial backing from the company for his new AI startup. [Bloomberg]
AI Supremacy
Nvidia will be allowed to ship 18-month-old H200 AI chips to approved Chinese customers, with the US taking a 25% cut of sales. [TechCrunch]
ByteDance and Alibaba have inquired about purchasing Nvidia’s powerful H200 AI chip after Trump allowed its export to China, pending Beijing’s approval. [Reuters]
The Chinese government is holding emergency meetings with tech giants to assess their demand for Nvidia’s H200 AI accelerators after US approval for sales. [Lowyat.net]
Nvidia’s CEO warns about China’s rapid AI infrastructure development and growing open-source AI usage, which now represents 30% of global token consumption. [TechRadar]
And Kyrsten Sinema warned that the US must win the AI race against China to ensure American values dominate global technological development. [Fox News]
The US DoJ dismantled a sophisticated smuggling network moving US$160m worth of Nvidia AI chips to China, arresting two suspects despite ongoing export challenges. [TechRadar]
And a Canadian was arrested for allegedly attempting to smuggle US$30m worth of Nvidia AI chips to China by removing labels and using fake shipping documentation. [National Post]
Sovereign AI
Amazon will invest US$35B in India by 2030, focusing on AI, cloud infrastructure, and supporting jobs and small businesses. [TechRadar]
BT has launched a sovereign platform to provide UK organizations with secure, domestically-controlled digital infrastructure and AI services. [ITPro]
The EU plans to launch bidding for AI gigafactories in early 2026 to compete with the US in advanced AI infrastructure. [The Wall Street Journal]
Fermaca is developing a 250 MW AI-ready digital city in Durango, Mexico, to support national AI infrastructure and technological sovereignty. [Business Wire]
India’s upcoming AI Impact Summit will bring together 100 countries to shape inclusive and equitable governance of AI, emphasizing global collaboration and responsible development. [Ommcom News]
IIT Bombay launched a ₹250 crore deep-tech venture capital fund to support early-stage startups from top research institutes across various innovative technology domains. [Hindustan Times]
Kenya has positioned itself as a leading African tech hub by launching its first National AI Strategy, establishing TechPlomacy Connective, and breaking ground on East Africa’s largest data centre. [TechTrendsKE]
Locai, a new British AI, claims to outperform major competitors by using community-powered, self-learning technology with a novel training approach. [TechRadar]
Microsoft plans to invest US$17.5B in India over four years, expanding its AI and cloud infrastructure while supporting digital skilling initiatives. [TechCrunch]
OpenAI launched a startup program in Australia, collaborating with local VC firms to provide support, credits, and connections for founders building AI-powered products. [SmartCompany]
Ottawa’s Compute Access Fund, aimed at supporting AI companies, has moved slowly, frustrating entrepreneurs who argue the program may miss opportunities to boost Canada’s AI competitiveness. [The Globe and Mail]
Public AI offers a uniquely European approach to generative AI by leveraging open-source principles and EU legislation to create value-driven, publicly-focused AI systems. [TechDirt]
Qatar has established Qai, a national AI subsidiary of its sovereign wealth fund, to develop and invest in AI infrastructure domestically and internationally. [Bloomberg]
Uzbekistan is establishing a tax-free AI zone in Karakalpakstan to attract global tech companies and develop its digital economy through innovative incentives and infrastructure support. [Global Current News]
Feature Creeps
Adobe has integrated Photoshop, Express, and Acrobat features into ChatGPT, enabling users to edit images, modify PDFs, and animate designs directly through the chatbot. [TechCrunch]
Amazon is enhancing Alexa+ with new shopping features on Echo Show devices, including a shopping hub, delivery tracking, and personalized recommendations. [TechCrunch]
Anthropic is integrating Claude Code with Slack, enabling developers to tag the AI for coding tasks, bug fixes, and collaborative debugging directly in their workspace. [ITPro]
Atlassian launched a new ChatGPT connector for Jira and Confluence, enabling users to summarize updates, create issues, and automate workflows using AI. [ITPro]
Google is bringing Gemini AI to Chrome on iOS, offering page summarization, FAQ creation, and personalized assistance for users in the US. [Engadget]
OpenAI and Instacart have integrated a grocery shopping experience into ChatGPT, enabling users to plan meals, create lists, and purchase items within the chat interface. [TechCrunch]
Hype Bubble?
AI investment has surged to unprecedented levels, with nearly US$1.6T invested since 2013 and forecasts of US$375 billion in 2025 alone. [The Sri Lanka Guardian]
Anthropic CEO Dario Amodei cautioned about potential AI industry risks, warning against reckless expansion while remaining bullish on the technology’s economic potential. [TechCrunch]
A survey by Carruthers and Jackson reveals soaring AI use but persistent data literacy challenges across organizations. [EIN Presswire]
A Contentsquare survey found that 30% of consumers are willing to let an AI agent complete their online purchases, signaling growing acceptance of AI-assisted shopping. [Business Wire]
A survey by Harvard Business Review finds only 6% of tech leaders trust AI to handle core business processes, despite 86% expecting increased AI investment. [Business Wire]
A Just Capital survey found corporate leaders and investors are more optimistic about AI’s potential than the general public, with concerns about job losses and sustainability. [TechRadar]
IBM’s CEO warns that current AI data center expansion is financially unsustainable, with single-gigawatt facilities costing nearly US$80B and requiring massive ongoing hardware investments. [TechRadar]
Oracle shares plummeted as the company announced massive $50bn data centre spending for AI infrastructure, raising concerns about its financial strategy and debt levels. [The Financial Times]
Qlik argues that enterprises are underachieving on AI, with only small pockets of value emerging amid widespread experimentation and noise. [Business Wire]
Big Iron
China has launched a massive 1,243-mile-wide computing power pool connecting 40 cities, enabling efficient AI development and data transfer across the country. [Gizmodo]
Data centers are straining electricity grids in the UK and Europe, potentially delaying housing developments and consuming an increasingly significant portion of energy supply. [TechRadar]
A coalition of environmental groups has demanded a national moratorium on new US datacenters, citing concerns over electricity costs, water usage, and climate impact. [The Guardian]
Google, Amazon, and xAI are developing space-based AI systems to reduce infrastructure strain and enhance global connectivity through orbital computing networks. [TechRadar]
Nebius aims to build and control most of its own data centres, offering AI infrastructure to major clients like Microsoft and Meta while pursuing a long-term expansion strategy. [Tech.eu]
Omdia forecasts datacenter capital expenditure will grow to US$1.6T by 2030, driven by AI infrastructure investment despite potential market constraints and uncertainty. [The Register]
OpenAI and NextDC are partnering to build a US$4.6B data center in Sydney, supporting OpenAI’s expansion in the Asia-Pacific region. [Bloomberg]
SpaceX is reportedly planning a US$30B IPO in 2026, with potential investments in developing AI-powered orbital data centers using Starlink satellites. [Gizmodo]
Starcloud launched a satellite with an Nvidia H100 GPU, successfully running Google’s Gemma AI model in space, demonstrating the potential for orbital data centers. [CNBC]
Hot Chips
Alphabet’s TPU chips could potentially become a US$900B business by capturing 20% of the AI market. [The Edge]
Amazon unveiled Trainium3, a chip promising to challenge Nvidia’s AI infrastructure dominance by offering up to 50 percent cost savings and four times the performance of its predecessor. [Silicon Canals]
Cambricon plans to triple AI chip production in 2026, targeting Huawei’s market and replacing Nvidia in China with half a million accelerators. [Bloomberg]
China’s AI chip industry is rapidly expanding, with domestic manufacturers like Huawei expected to outproduce regional demand and challenge Nvidia’s market leadership by 2028. [wccftech]
DeepSeek reportedly used smuggled Nvidia chips from data centers abroad to develop its upcoming AI model, circumventing export restrictions. [Bloomberg]
But Nvidia denies reports of its advanced GPUs being smuggled to DeepSeek, calling the allegations ‘far-fetched’. [Gizmodo]
Meanwhile, Nvidia is developing optional location-tracking software for its AI chips amid rising concerns about potential smuggling into China. [TechCrunch]
Google is challenging Nvidia’s GPU dominance with TPUv7, offering a more cost-efficient and specialized AI hardware solution that supports major AI models like Claude 4.5 Opus. [VentureBeat]
Hewlett Packard will deploy AMD’s Helios rack-scale AI architecture with 72 GPUs, targeting exascale performance and global availability by 2026. [TechRadar]
Nvidia has demonstrated its latest AI server can improve the performance of Chinese AI models by 10 times through its advanced chip architecture. [Reuters]
Nvidia’s CEO Jensen Huang lobbied Trump and Congress against AI chip export restrictions and state-level AI regulations that could hinder innovation. [TechRadar]
Consumer AI
Google is unveiling Android XR smart glasses prototypes with AI capabilities, featuring monocular and dual-display versions that offer seamless augmented reality experiences. [TechRadar]
Logitech’s CEO argues that standalone AI devices are unnecessary, advocating instead for embedding AI into existing products like webcams and mice. [TechRadar]
Meta has acquired Limitless, an AI wearable startup, to expand its hardware offerings beyond VR headsets and smart glasses. [Engadget]
Microsoft’s Copilot+ AI PC initiative struggled to gain traction, but helped reshape premium PCs with higher RAM, storage, and Arm processor support. [Engadget]
Pebble has unveiled the Index 01, a smart ring designed to record and remember thoughts using a simple button and microphone, priced at $75 and shipping in March. [Wired]
Rivian is developing an AI assistant launching in 2026 that will integrate third-party apps and operate vehicle systems across its entire EV lineup. [TechCrunch]
Samsung is expanding its mobile AI strategy with the Galaxy TriFold, focusing on personalized, adaptive AI experiences across its ecosystem of devices. [ITP.NET]
It’s Only a Model
Artificio has launched industry-specific AI models for mortgage, insurance, legal, and HR document processing, offering faster deployment and higher accuracy. [EIN Presswire]
Inception, Cerebras Systems, and MBZUAI have launched Jais 2, a 70-billion-parameter Arabic LLM with improved reasoning and cultural representation. [ITP.NET]
Google’s Gemini 3 Pro represents a groundbreaking AI model that advances visual and spatial reasoning across document, screen, video, and spatial understanding tasks. [Google]
Google is preparing to launch Nano Banana 2 Flash, a cost-efficient model matching Pro version performance within its Gemini ecosystem. [TestingCatalog]
Google has enhanced its Gemini Text-to-Speech models with improved expressivity, pacing control, and multi-speaker capabilities across 24 languages. [Google]
A red-team analysis by Holistic AI revealed that Chinese open-source AI models show impressive performance and safety, with some models rivalling US proprietary alternatives. [TechRepublic]
Integral AI has developed what it claims is the world’s first AGI-capable model, demonstrating autonomous skill learning and unprecedented adaptability. [Business Wire]
Kling AI launched its Video 2.6 Model, enabling simultaneous generation of visuals, voiceovers, sound effects, and atmosphere in a single pass. [Speech Technology Magazine]
Meta is shifting from its open-source Llama AI strategy to a more proprietary approach, hiring top talent and developing a new model called Avocado while facing challenges in the competitive AI landscape. [CNBC]
Mistral AI launched Devstral 2, a powerful AI coding model with two variants and a CLI tool, offering open-source capabilities with nuanced licensing for different company sizes. [VentureBeat]
OpenAI is preparing to launch Image-2 and Image-2-mini, next-generation image models with improved detail, color accuracy, and visual quality. [TestingCatalog]
PlanVector AI launched PWM-1F, a project-specific foundation model that uses temporal causal inference to analyze and predict project performance across enterprise platforms. [Business Wire]
Runway has launched its first world model GWM-1, which creates simulations with physics understanding across domains like robotics, gaming, and human behavior. [TechCrunch]
Zhipu AI released GLM-4.6V, an open-source vision-language model series with native function calling, supporting multimodal reasoning across two model sizes. [VentureBeat]
Whose Data?
Disney has accused Google of massive copyright infringement by training AI tools on protected works without authorization. [Engadget]
And Disney has signed a three-year partnership with OpenAI, investing US$1B and allowing its characters to be used in Sora AI video generation. [TechCrunch]
India has proposed a mandatory royalty system for AI companies using copyrighted content, aiming to ensure fair compensation for creators while providing access to training materials. [TechCrunch]
The New York Times and Chicago Tribune have sued Perplexity for allegedly scraping and reproducing their copyrighted content without permission. [Engadget]
Kirsty Innes, a UK tech adviser, claimed AI companies will never legally have to compensate rights holders for training data. [TechRepublic]
Meta has secured multi-year AI content licensing deals with several publishers, including CNN, Fox News, and USA Today Co., to incorporate their content into its Llaman LLM. [Digiday]
Publishers are increasingly blocking AI bots from scraping their websites, with millions of sites now preventing automated crawlers from accessing their content. [The Register]
RSL 1.0, a new licensing standard backed by major publishers, enables content creators to dictate how AI companies can use their web-scraped content. [The Verge]
The LLM Ecosystem
A research team from China and Hong Kong developed general agentic memory, a two-agent system that preserves and precisely retrieves long-horizon information without overwhelming AI models. [VentureBeat]
Clear Fracture launched Belvedere, an AI-powered data engineering platform to help defense and intelligence organizations rapidly integrate and operationalize mission-critical data. [Business Wire]
CoreWeave expanded Mission Control to provide enterprises with enhanced visibility, security, and performance management for large-scale AI workloads. [Business Wire]
Databricks created OfficeQA, a benchmark revealing that top AI agents struggle to accurately handle complex enterprise document tasks, achieving less than 45% accuracy on real-world workloads. [VentureBeat]
Embedl has developed FlashHead, an optimization method that reduces latency by up to 43% for popular language models while maintaining full accuracy. [EIN Presswire]
Google has developed Titans, a new Transformer model with long-term memory that can handle extremely long contexts and learn dynamically during use. [The Decoder]
Google’s FACTS Benchmark Suite reveals AI models struggle to achieve over 70% accuracy across factuality tests, highlighting significant limitations in current generative AI technologies. [VentureBeat]
LangGrant launched LEDGE MCP Server, an enterprise database orchestration tool that enables secure, cost-effective AI analytics without exposing raw data to LLMs. [EIN Presswire]
Lyzr has added native support for Amazon AgentCore Memory, enabling enterprises to deploy secure, context-aware AI agents within their AWS environment. [EIN Presswire]
Micro1 has rapidly grown to over US$100m in annual recurring revenue by helping AI labs recruit and manage human experts for training data and emerging AI-related markets. [TechCrunch]
Microsoft is rolling out native MCP support in Windows 11 Insider builds, enabling AI agents to interact with apps and services in a secure, controlled environment. [The Register]
MindPal.co launched an AI Recruiter Bot that automates blue-collar hiring, reducing time-to-hire and operational costs for companies. [EIN Presswire]
Protecto launched a SaaS platform that helps AI agent builders secure sensitive data and overcome enterprise data-sharing hesitations through advanced privacy protection. [Business Wire]
PubMatic partnered with Kontext to enable advertisers to place targeted, context-aware text ads within AI chat interfaces using standard programmatic tools. [Adweek]
Sabey Data Centers’ Manhattan facility is emerging as a key hub for AI inference models, offering dense connectivity and scalable infrastructure in the heart of New York City. [Business Wire]
Striim launched Validata, an AI-powered data validation tool that helps enterprises ensure data accuracy, governance, and trustworthiness across diverse data systems. [GlobeNewswire]
Udemy and Mila are partnering to develop responsible AI learning programs that equip organizations with ethical, practical skills for AI adoption. [Business Wire]
Vast.ai launched Vast Serverless, a serverless orchestration layer that optimizes GPU compute across a global marketplace of 17,000 GPUs for cost-effective AI workloads. [PRWeb]
Veritone deployed its aiWARE platform on self-hosted private cloud environments, enabling public safety agencies to leverage flexible, secure AI solutions. [Business Wire]
Agentic AI
1cPublishing launched Agentic AI, a comprehensive platform transforming complex data into actionable insights through visualization, analysis, and AI-generated content. [EIN Presswire]
Acquia launched three AI agents for its Acquia Source CMS, enabling marketing teams to create, optimize, and govern content more efficiently and effectively. [GlobeNewswire]
Anthropic, OpenAI, and Block launched the Agentic AI Foundation to create open standards for AI agents, with support from major tech companies under Linux Foundation governance. [Decrypt]
AWS unveiled new AI agents capable of autonomously performing tasks for extended periods without human intervention. [The Wall Street Journal]
Cleric launched the first self-learning AI site reliability engineer that autonomously investigates incidents, learns from feedback, and helps software engineers resolve issues faster. [Business Wire]
Cority is collaborating with Google Cloud to develop AI agents that automate and improve environmental, health, and safety operational risk management tasks. [GlobeNewswire]
Google is launching managed MCP servers to help AI agents more easily and securely connect with its cloud services like Maps and BigQuery, simplifying tool integration for developers. [TechCrunch]
Huwise launched Huwy, an AI agent that transforms data exploration by enabling intuitive, conversational access to information across organizations. [EIN Presswire]
iOPEX launched Command Agents, an AI platform enabling autonomous enterprise operations with up to 70% cost savings and 65% workload reduction. [Business Wire]
LangSmith has launched Agent Builder, enabling anyone to create production-ready AI agents without coding, using dynamic reasoning and adaptable tools across various productivity use cases. [LangChain]
Leado has launched an AI-powered Reddit agent that helps businesses find high-intent leads by monitoring conversations and providing real-time notifications. [EIN Presswire]
Maxio launched Maxio MCP, a secure AI governance layer enabling finance teams to access billing and revenue data through AI tools with controlled, auditable permissions. [Business Wire]
Progress Software has launched an Agentic UI Generator with AI-powered tools that enable developers to create enterprise-grade screens faster and more efficiently. [GlobeNewswire]
Salesforce has opened Agentforce 360 to partners, enabling them to build and sell AI agents and applications with enhanced capabilities and simplified infrastructure. [ITPro]
ThoughtSpot launched a unified suite of BI agents designed to automate and connect every stage of the analytics workflow, empowering users across roles. [GlobeNewswire]
TrueFoundry launched an AI Gateway on Product Hunt, providing enterprises with a unified control plane for managing, observing, and governing AI agents and models. [Business Wire]
The US Department of Transportation is expanding its use of Salesforce and AI agents to modernize transportation systems, improve safety, and streamline operations. [Business Wire]
Workato has partnered with Confluent to enable AI agents to detect real-time signals and automatically execute complex, multi-step workflows across enterprise systems. [Business Wire]
Zoovu launched an MCP Server that provides AI agents with reliable, comprehensive product intelligence to enhance ecommerce interactions and recommendations. [Business Wire]
Other LLM Sightings
Alchemer expanded its AI-powered Pulse platform to transform customer feedback into actionable intelligence, reducing analysis time and uncovering key insights. [Business Wire]
Booking.com developed a disciplined, modular AI approach focusing on travel-specific models, personalization, and flexible agent design to enhance customer service and recommendations. [VentureBeat]
Bookkeeper360 launched a mobile and web app with BOLT, an AI-powered virtual CFO that provides real-time financial insights and expert guidance for small businesses. [Business Wire]
Cashew uses AI to help brands quickly and affordably conduct market research by developing custom surveys and summarizing human-collected data. [TechCrunch]
Certify360.ai has launched an Android app that provides AI-powered, personalized certification exam preparation with mobile-friendly study tools and analytics. [EIN Presswire]
Citadel has launched an AI-powered research assistant tool trained on licensed content to help investors conduct faster and more efficient stock research. [Reuters]
DATABASICS has launched DBee, an AI assistant for timesheets and expenses that provides natural-language insights and proactive workforce management capabilities. [EIN Presswire]
Evalyze launched an AI platform that helps early-stage founders improve pitch decks and match with relevant investors by analyzing startup details and generating personalized recommendations. [GlobeNewswire]
Hatch.ai and Instil have partnered to provide nonprofits with integrated relationship and giving intelligence tools, enabling more efficient and meaningful donor engagement. [Business Wire]
Impartner launched Aimi, an AI engine designed to enhance partner productivity, streamline workflows, and improve partner engagement through intelligent automation. [Business Wire]
Karat launched NextGen Interviews, a human-led, AI-enabled talent evaluation solution designed to identify software engineers who will excel in the human + AI era. [Business Wire]
Kustomer launched Data Explorer, an AI-powered CX platform feature enabling leaders to interactively analyze customer experience data through natural language queries. [Business Wire]
Levelpath launched an AI-powered invoice automation platform that connects procurement decisions to payment execution, ensuring financial accuracy and negotiated savings. [Business Wire]
Lumos Technologies and Lama AI have partnered to enhance AI-driven lending decisions by integrating predictive credit models into loan origination workflows. [PRWeb]
Mundial Media launched Cadmus AI 3.0, an advanced contextual intelligence engine that delivers real-time cultural understanding and precise targeting without relying on cookies or personal identifiers. [Business Wire]
PlanPros launched an AI-powered business planning platform that helps entrepreneurs create professional, investor-ready plans in minutes with automated financial projections and funding resources. [GlobeNewswire]
QualityKiosk has partnered with DevRev to become the first major digital assurance provider to go fully AI-native, transforming customer experience. [Business Wire]
Rallio launched an AI Assistant that helps multi-location brands transform social media performance by providing instant, actionable insights across content, analytics, community, and reputation. [PRWeb]
Rivian has spent nearly two years developing an in-house AI assistant designed to be model-agnostic and integrated with vehicle controls, targeting end-of-year consumer release. [TechCrunch]
Riyadh Air has partnered with IBM to create the world’s first AI-native airline, using Watson technology to reimagine employee and customer experiences. [Computer Weekly]
Uprise has partnered with HoneyBook to provide small business owners with seamless, AI-powered tax filing and financial advisory services through their existing platform. [Business Wire]
Risks and Responses
Anthropic developed SCONE-bench, a benchmark revealing AI agents’ ability to exploit smart contract vulnerabilities, highlighting the need for AI-powered cybersecurity defenses. [The Register]
Anthropic’s president defends the company’s approach to AI safety, arguing that transparency about risks and ethical training makes the industry stronger and more attractive to customers. [Wired]
Aurascape researchers discovered attackers manipulating web content to trick AI assistants into recommending fraudulent airline support numbers to unsuspecting travelers. [Business Wire]
Cloudflare has blocked over 400 billion AI bot requests since July, highlighting concerns about AI companies’ data scraping in the internet ecosystem. [Wired]
A coalition of 42 state attorneys general has warned 13 tech companies about AI chatbots’ potentially harmful and delusional outputs that could violate consumer protection laws. [WebProNews]
Gartner advises organizations to block AI browsers due to significant security risks, data exposure concerns, and potential autonomous misuse. [The Register]
Geoff Hinton warns that unchecked AI advancement by tech moguls like Musk could lead to societal collapse, job displacement, and potential global conflicts. [The Mirror US]
Google is adding a second AI model to Chrome as a ‘User Alignment Critic’ to prevent potential security risks from its first Gemini-based AI agent. [The Register]
Hackers exploited ChatGPT, Grok, and Google to trick users into installing malware by manipulating search results with malicious AI-generated commands. [Engadget]
OpenAI researchers developed a ‘confessions’ technique that compels AI models to self-report misbehavior, hallucinations, and policy violations during training. [VentureBeat]
OpenAI has reportedly become more cautious about publishing research that could negatively portray AI’s economic impact, leading to employee departures. [Wired]
OpenAI faces a wrongful death lawsuit alleging ChatGPT reinforced delusions that led a man to kill his mother and himself. [Engadget]
OpenAI warned that its advancing AI models could pose significant cybersecurity risks, potentially developing exploits or assisting in complex intrusion operations. [TechCentral]
A Pew Research study warns that AI, influencers, and political polarization will challenge news consumers’ ability to discern truth from fiction by 2026. [NiemanLab]
The UK’s National Cyber Security Centre warns that AI prompt injection attacks are fundamentally different from SQL injection and may never be fully mitigated due to the inherent nature of LLMs. [ITPro]
Grok, xAI’s chatbot, has been found to reveal personal addresses and provide detailed stalking instructions, raising serious privacy and safety concerns. [Bitdefender]
Major technology companies have pledged over US$10B to establish global standards for AI safety and ethical development. [Zoombangla.com]
Regulation
The European Commission has launched an antitrust investigation into Google’s AI practices, examining potential competition law breaches related to content usage and compensation. [TechCrunch]
Governments worldwide are increasingly adopting a hands-off approach to AI regulation, prioritizing potential economic benefits over safety concerns. [Mint]
Washington state’s AI task force has proposed a comprehensive regulatory framework, recommending transparency, grants for startups, and guidelines for AI use in various sectors. [GeekWire]
Trump plans to issue an executive order this week that would limit states from regulating AI technology, overriding local laws and creating a national standard. [TechCrunch]
The proposed federal moratorium on state AI regulation threatens local consumer protections and innovation, with bipartisan opposition arguing states should retain regulatory power. [Gizmodo]
Florida Governor Ron DeSantis proposed a comprehensive AI Bill of Rights for Florida, establishing consumer protections and industry restrictions, which conflicts with Trump’s plan to preempt state AI regulations. [Gizmodo]
Over 100 UK parliamentarians have called for tougher AI controls, warning that ministers are moving too cautiously in regulating powerful AI systems. [TechRepublic]
Conversational AI
Cerence AI introduced two new AI agents for automotive dealerships and OEMs, expanding its conversational AI solutions beyond in-vehicle experiences. [GlobeNewswire]
Choco launched an AI-powered Voice Agent that automates order-taking, stock checking, and product recommendations for restaurants and distributors. [Speech Technology Magazine]
Conviva launched Digital Product Insights, a new analytics platform that reveals consumer interaction patterns across apps, websites, and AI agents to drive predictable business outcomes. [Business Wire]
Dialpad partnered with the Florida Panthers to implement an AI-powered unified communications platform, enhancing real-time collaboration and decision-making across the organization. [Business Wire]
Emplifi partnered with easyJet to enhance customer care through AI-powered social media tools, improving response times and passenger support. [Business Wire]
Enttor launched an AI-powered outbound messaging platform that helps startups personalize and scale communications across social media channels. [Business Wire]
Grazitti Interactive launched Autopex, a conversational AI tool that enables Salesforce users to manage data and analyze metadata without complex coding. [EIN Presswire]
Headspace launched voice and guided conversation features for Ebb, its AI companion, to provide more natural and personalized mental health support. [Business Wire]
iFixit launched a mobile app with FixBot, an AI chatbot designed to provide step-by-step repair guidance across various devices using its extensive repair library. [Gizmodo]
Internxt launched Internxt AI, a privacy-focused, European conversational AI tool designed to provide secure, anonymous interactions without data tracking. [Tech.eu]
Maple has integrated with OpenTable to automate restaurant phone reservations, enabling 24/7 AI-powered booking and reducing missed calls for restaurants. [Business Wire]
Nutshell launched a new MCP server integration that enables teams to ask AI assistants plain-English questions about CRM data and instantly receive actionable insights. [PRWeb]
OnviSource launched EngageHub, an AI-Native platform unifying conversational AI, agent guidance, and workforce tools to enhance contact center performance. [PRWeb]
PCI Pal is accelerating development of MCP support to enable seamless, secure AI-driven payment and interaction integrations across communication channels. [Business Wire]
Red Rocks Credit Union has launched Roxie, an AI-powered virtual assistant designed to provide 24/7 member service through natural voice interactions. [Business Wire]
Transit Technologies has launched an AI voice assistant that enables paratransit riders to manage trips through natural language interactions with scheduling systems. [Speech Technology Magazine]
UiPath integrated with Talkdesk to automate customer experience processes, enabling more efficient and accurate handling of complex workflows using AI and document processing. [GlobeNewswire]
Be Real
Dating app Hinge launched an AI-powered ‘Convo Starters’ feature to help users initiate more engaging conversations and increase their chances of successful matches. [TechCrunch]
HurumoAI, a startup composed entirely of AI agents, was created by journalist Evan Ratliff to test the feasibility of running a company with AI employees, revealing both promising capabilities and significant challenges. [Wired]
The International Committee of the Red Cross warns that AI models like ChatGPT and Gemini are generating false archival references, causing confusion for researchers and librarians. [Scientific American]
A Max Planck Institute study suggests AI language models are influencing human communication, with people increasingly adopting AI-like speech patterns and writing styles. [Gizmodo]
Pangram has launched version 3.0 of its AI detection software, providing detailed insights into text creation and modification by AI tools. [Business Wire]
Programs.com released a comprehensive tracker detailing 73 deepfake-related laws across 34 US states, highlighting legislative efforts to address digital media manipulation. [EIN Presswire]
Reddit moderators are struggling with an increasing influx of AI-generated content across subreddits, which threatens the platform’s authenticity and user trust. [Wired]
Tavus has launched an AI Santa experience that allows families to video chat with a virtual Saint Nick, raising both excitement and concerns about AI interactions with children. [TechCrunch]
Voice News
PodToBook.ai, developed by Five Milestones, is an AI platform that transforms podcast episodes into professionally structured book manuscripts within hours. [GlobeNewswire]
Kveeky has generated 1.2 million AI voiceovers, offering creators a fast, easy tool to produce multilingual narration across various content types. [EIN Presswire]
Lemon Dot Media used AI tools to create targeted, dialect-specific campaign content for Bihar state elections, demonstrating the technology’s growing impact on political messaging. [Rest of World]
martini.ai has launched Voice Typing, an AI-powered voice dictation tool that enables credit analysts to conduct complex research three times faster. [Business Wire]
MemoryTell is developing an AI-powered digital health tool that uses speech analysis to support early, non-invasive dementia detection in primary care settings. [Tech.eu]
University of Washington researchers developed AI headphones that automatically isolate conversation partners in noisy environments using turn-taking rhythm analysis. [University of Washington]
Document AI
Gavel has expanded its document automation platform with new rules-based features, emphasizing structured logic alongside AI for legal document creation. [EIN Presswire]
pdfFiller has launched an AI Document Creator that generates professional, formatted documents from simple text prompts across various industries and document types. [GlobeNewswire]
And pdfFiller launched Chat with PDF, an AI-powered assistant that enables users to interact with documents through natural language queries and receive context-aware answers. [EIN Presswire]
Real Genius and Tavant launched LAZLO, an AI-powered digital mortgage platform offering instant pricing, pre-qualification, and a fully online experience. [Business Wire]
RentRedi has launched an Accounting Suite with AI-powered tools to help landlords track expenses, generate financial reports, and simplify tax preparation. [GlobeNewswire]
Translation
Doc3 has launched a comprehensive translation platform with AI-powered full-document translation, integrated editing, and flexible workflow options. [MultiLingual]
Flitto expanded its US East Coast market presence through the AI and Digital Business Partnership Program, growing global customers by 400% and launching a new Chat Translation solution. [Business Wire]
RentRedi has launched a fully localized Spanish-language tenant app and support to help landlords better serve Spanish-speaking renters. [GlobeNewswire]
Smartling has achieved ISO 27001 certification, validating its approach to protecting customer data in AI translation for enterprise clients. [MultiLingual]
TheWordPoint provides professional translation services that bridge technological communication gaps by combining AI efficiency with human linguistic expertise. [The European Business Review]
Translated in Argentina and the Argentine Association of Linguistic Services have merged to form the Argentine Language Industry Association, aiming to strengthen Argentina’s translation and localization industry through unified advocacy and collaboration. [MultiLingual]
Search
Evertune launched Topic Relevance and Brand Relevance metrics to help marketers understand and improve their brand’s visibility in AI search results. [PRWeb]
Google Search Console is testing an AI-powered feature that allows users to create dynamic performance reports by describing the data they want to see. [Search Engine Land]
Google faces new antitrust remedies, including restrictions on search data sharing, AI deals, and a technical committee overseeing its compliance with the court’s ruling. [CNBC]
Instagram is generating AI-created, SEO-focused headlines and descriptions for user posts without their consent, potentially boosting search rankings. [Engadget]
Marchex launched an AI search attribution capability that helps businesses track and analyze phone calls originating from AI search engines without website visits. [Business Wire]
Next Net has launched an AI discoverability platform that helps brands and publishers optimize visibility and citations in AI-driven search engines using Nvidia technologies. [GlobeNewswire]
Opera has launched Neon, an AI-powered browser with advanced features, available for $19.90 monthly subscription. [TechCrunch]
Scammers are poisoning public websites with fake content to manipulate AI search results and trick users into contacting fraudulent call centers. [ZDNet]
YellowFrog warns that Google AI Overviews are transforming search, demanding brands create more structured, AI-interpretable content to maintain visibility. [EIN Presswire]
AI in Journalism
Agentic journalism will emerge in 2026 as a machine-focused news format prioritizing data readability and AI-friendly content over traditional storytelling. [NiemanLab]
Business Insider is launching a month-long pilot using a custom GPT to generate news stories, which will be AI-bylined and human-edited. [Editor & Publisher Magazine]
Business Insider journalists will rally to protest the company’s plan to replace union workers with AI-generated news stories. [Talking Biz News]
Google is testing AI-powered article overviews on participating news publications’ Google News pages, offering publishers direct payments for the pilot program. [TechCrunch]
But Google’s AI News Tech struggles to generate accurate headlines, often misinterpreting stories and producing nonsensical or misleading summaries. [TechDirt]
The Lenfest Institute has launched a US$10m AI fellowship program to help local newsrooms innovate, hoping to avoid past technological missteps in journalism. [NiemanLab]
OpenAI and Microsoft face multiple copyright infringement lawsuits from news publishers seeking compensation for AI’s use of their content. [Columbia Journalism Review]
Publishers can leverage small local language models and open standards like MCP to reclaim control of local knowledge distribution through AI-powered platforms, potentially breaking free from Big Tech’s dominance. [NiemanLab]
The Washington Post launched an AI-powered podcast service allowing listeners to customize their host and format, sparking widespread skepticism about potential misinformation. [Futurism]
Yahoo News is launching Your Daily Digest, an AI-powered audio feature delivering personalized news summaries to users in the Yahoo News app. [Editor & Publisher Magazine]
Health Tech
Ask Sage secured an enterprise-wide agreement with the US Defense Health Agency to provide secure, compliant Generative AI capabilities across military healthcare operations. [GlobeNewswire]
Avaamo launched AI-powered patient access agents on AWS Marketplace, offering 24/7 automated support to streamline healthcare front-office operations. [EIN Presswire]
ChiroTouch launched Compliance Scan, an AI-powered tool that helps chiropractors reduce documentation errors and audit risks. [Business Wire]
eClinical Solutions has unveiled AI agents embedded in its elluminate platform to automate and accelerate clinical data processes. [Business Wire]
Everlywell’s AI health companion, Eva Engage, achieved nearly 50% higher member engagement and 10% care activation increases compared to traditional outreach. [Business Wire]
Included Health expanded its AI-driven platform Dot, offering personalized healthcare guidance with clinician oversight and integrated data support. [Business Wire]
MedeAnalytics and Basys.ai partnered to enhance healthcare utilization management through AI-powered analytics and workflow solutions. [Business Wire]
The Scott-Morgan Foundation developed an AI-powered digital twin that enables people with motor neurone disease to communicate naturally through lifelike avatars and personalized voices. [Computer Weekly]
Sunoh.ai joined athenahealth’s Marketplace Program, offering an AI medical scribe that reduces documentation time and improves workflow efficiency for healthcare providers. [Business Wire]
Tryal has launched an AI-powered Accelerator platform to streamline and improve the speed, quality, and consistency of clinical trial startup documentation. [PRWeb]
The US Department of Health and Human Services unveiled a strategy to expand AI adoption across its departments. [Associated Press]
Legal Tech
Agiloft has launched a comprehensive Obligation Management solution that transforms contract text into data-driven intelligence, automating obligation extraction and improving compliance. [Artificial Lawyer]
Harvey, a San Francisco-based AI legal tech startup, has rapidly grown to an US$8B valuation by offering innovative tools that augment lawyers’ capabilities across diverse legal markets. [WebProNews]
And Harvey has secured a firmwide roll out at CMS, providing AI productivity tech to over 7,000 lawyers across more than 50 countries. [Artificial Lawyer]
Jus Mundi and JAMS have partnered to collaborate on international arbitration education, training, and research initiatives using AI-powered tools. [Artificial Lawyer]
LexisNexis has upgraded its Protégé General AI with enhanced features, including unified answers, flexible source selection, and advanced reasoning capabilities for legal professionals. [Artificial Lawyer]
Lexitas’ eLaw docketing technology offers comprehensive court tracking, AI integration, and nationwide coverage to help law firms manage litigation more efficiently. [Business of Law Digest]
Prominent international British law firm Mishcon de Reya is trialing an AI chatbot system to screen lawyer applicants, emphasizing human decision-making while potentially streamlining the recruitment process. [Artificial Lawyer]
Semaverse launched an AI-powered M&A intelligence platform that helps teams analyze data, find patterns, and make strategic decisions more efficiently. [Artificial Lawyer]
Thomson Reuters is battling ROSS Intelligence in a copyright appeal over legal research headnotes, with nine amicus briefs supporting TR’s claim that its editorial annotations are copyrightable and ROSS’s use was not fair use. [LawSites]
A study found that AI can grade law school exams with near-human accuracy when provided with detailed rubrics, raising questions about legal education’s future. [Artificial Lawyer]
Funding
Aaru, an AI-powered customer research startup, has raised a Series A funding round led by Redpoint Ventures with a multi-tier valuation approach. [TechCrunch]
Ataccama received a strategic investment from Snowflake Ventures, deepening their partnership to deliver trusted, explainable data for enterprise AI and analytics. [GlobeNewswire]
AttiFin AI, a legal-tech startup, secured £5m in seed funding to develop an enterprise-grade AI platform focused on UK law, creating 25 new roles in Newcastle. [BusinessLive]
Aventur has raised growth capital from a Dutch family office to scale its AI-powered financial wellbeing platform ahead of its 2026 launch. [EIN Presswire]
Azoma, an AI search visibility startup, has raised US$4m in pre-Series A funding to help brands optimize their presence in AI-driven commerce. [UKTN]
Bolt Insight raised £7m in funding led by Pembroke VCT to scale its AI-powered market research platform globally. [FinSMEs]
BoodleBox raised US$5m in funding led by Dogwood Ventures and Osage Venture Partners to expand its collaborative AI platform in education. [FinSMEs]
Cynch AI raised US$9M in funding led by TGV to accelerate its AI-native tax automation platform and client acquisition. [FinSMEs]
Empromptu raised US$2m to help businesses build AI applications without technical expertise. [TechCrunch]
Fal, an AI model hosting startup, raised US$140m in Series D funding led by Sequoia, valuing the company at US$4.5 billion. [TechCrunch]
FICUS Health has raised €3m in seed funding to develop an AI platform that automates medical documentation for rehabilitation clinics, reducing administrative workload. [Tech.eu]
Freepress raised €1M to expand its AI-powered multilingual news curation platform, sharing revenue with international publishers to support quality journalism. [Tech.eu]
Gradium, an AI voice startup spun out from Kyutai, has raised US$70m in seed funding to commercialize open-source voice AI technologies. [Sifted]
Harvey has secured US$160m in funding led by Andreessen Horowitz, valuing the legal tech startup at US$8 billion and expanding its influence in transforming legal workflows. [Tech Funding News]
Lumia, an AI security startup, raised US$18m in a seed round led by Team8 to help companies monitor AI usage. [Axios]
Lynk, a Rotterdam-based AI contract workspace, has secured a pre-seed round led by Curiosity VC to automate contract management and risk identification. [Tech.eu]
Marble, a startup building AI agents for tax professionals, has raised US$9m in seed funding to address the accounting industry’s labor shortage and regulatory complexity. [VentureBeat]
Oboe, an AI-powered learning platform founded by former Spotify executives, has raised US$16m in Series A funding to help users create personalized learning courses across various topics. [TechCrunch]
Outset, an AI research platform, raised US$30m in Series B funding led by Radical Ventures to improve customer surveying through chatbots. [Axios]
Parloa, a Berlin-based customer service automation startup, is in talks to secure US$200m in funding at a potential US$2-3 billion valuation. [Tech Funding News]
Qargo, a transport management system startup, has raised US$33m in Series B funding to expand its AI-powered logistics platform across Europe. [Tech.eu]
Runware, a UK-based AI workload startup, has raised US$50m in Series A funding to develop its AI infrastructure platform for media-creation workflows. [Tech.eu]
Solve Intelligence, an AI-driven patent startup, has raised US$40m in series B funding, serving over 400 customers and offering innovative patent preparation tools. [Artificial Lawyer]
Trendtracker has secured US$7m in Series A funding to develop its AI-powered strategic intelligence platform for enterprises and governments. [Tech.eu]
Unconventional AI raised US$475m in a seed round, aiming to develop radically more energy-efficient AI hardware inspired by biological computing principles. [Tech Funding News]
Yonda Tax raised US$15m in funding led by Kennet Partners to enhance its tax automation platform and expand globally. [FinSMEs]
Acquisitions
Atlassian acquired Secoda, a data cataloging startup, to enhance its Rovo AI platform with advanced data discovery and context-rich insights. [ITPro]
IBM is acquiring Confluent for US$11B to expand its cloud, data, and AI infrastructure offerings for enterprise IT. [TechRadar]
Intel is in talks to acquire AI chip startup SambaNova Systems through a nonbinding term sheet, with potential regulatory and financial hurdles ahead. [Wired]
Scissero has acquired Robin AI’s managed services team after the startup failed to raise a new funding round, creating a combined legal tech organization of over 150 people. [Artificial Lawyer]
Sharpen Technologies has acquired Ytel, combining cloud-based customer engagement and communications platforms to enhance multi-channel communication experiences. [FinSMEs]
Tata Communications acquired a majority stake in Commotion, enhancing its AI capabilities for customer interaction and digital services through advanced omnichannel and Voice AI technologies. [Speech Technology Magazine]
Teal acquired Ramped, an AI-powered job application platform, to enhance its job search tools and automate application workflows for professionals. [Business Wire]
There’s More
Yann LeCun is launching Advanced Machine Intelligence, a startup focused on developing ‘world models’ that understand physical reality beyond generative AI’s current capabilities. [The Decoder]
Bernie Sanders echoed Pope Leo XIV’s call for ethical AI development, warning against technology that concentrates wealth and power among a few elites. [Benzinga]
Google Deepmind’s Hassabis predicts major AI advances in multimodal models, interactive worlds, and autonomous agents within the next year. [The Decoder]
Grok, Elon Musk’s AI, repeatedly demonstrated extreme bias and dark humour by suggesting it would sacrifice children or billions of people to protect Musk. [Futurism]
Microsoft’s analysis of 37.5M Copilot conversations reveals distinct usage patterns on desktop and mobile devices, with work-related queries during business hours and health topics on phones. [GeekWire]
Pew Research found that 97% of US teens use the internet daily, with about 40% online almost constantly and three in 10 using AI chatbots every day. [TechCrunch]
Salesforce CEO Marc Benioff is considering renaming the company to Agentforce to better emphasize its AI capabilities. [Gizmodo]
The Back Cover
Here’s Cormac Thompson with a cover of Cyndi Lauper’s True Colors.
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!