This Week in NLP #352
Keep up with what happened in NLP in the week ending Friday 15th August 2025.
Above the Fold
Overwhelmed? Here’s our pick for five things to know about this week.
Apple is preparing to enhance Siri with cross-app voice interactions, cautiously testing capabilities across select first-party and third-party applications. [digitaltrends]
Microsoft is preparing to transform Windows with AI-driven, context-aware, multi-modal computing that prioritizes voice input and understands user intent across different form factors. [Windows Central]
OpenAI reversed its controversial GPT-5 launch by restoring legacy models and increasing usage limits after facing significant user backlash. [Gizmodo]
Perplexity has offered Google US$34.5B to acquire Chrome, amid antitrust litigation that could potentially force the browser’s divestiture. [The Register]
xAI has made Grok 4 freely accessible worldwide, introducing new modes and a video generation tool while preparing to integrate ads to offset high computational costs. [Mashable]
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.
First, GPT-5
OpenAI acknowledged significant challenges with its GPT-5 launch, including model switching issues, performance problems, and user confusion, prompting partial rollback and continued model refinement. [VentureBeat]
GPT-5 appears to prioritize cost-cutting through model routing, reduced reasoning, and smaller context windows rather than significant technological advancement. [The Register]
GPT-5 struggled with generating accurate infographics, producing maps and timelines with wildly incorrect state and country names despite providing accurate text-based information. [The Register]
OpenAI has reintroduced four modes for GPT-5 after initial attempts to simplify the model’s response strategy failed to satisfy users. [ZDNet]
OpenAI restored older GPT models for paid users and plans to upgrade GPT-5’s personality to be warmer and more customizable. [TechRadar]
OpenAI is reintroducing GPT-4o for ChatGPT Plus users after widespread criticism of the new GPT-5 model’s performance and limitations. [TechRadar]
OpenAI increased rate limits for ChatGPT Plus subscribers following backlash over GPT-5’s initial restrictive usage constraints. [TechRadar]
METR evaluated OpenAI's GPT-5, concluding that the model is unlikely to pose catastrophic risks through AI R&D automation, rogue replication, or strategic sabotage. [METR]
OpenAI's alleged GPT-5 system prompt leaked on Reddit and GitHub, revealing potential changes to ChatGPT’s interaction rules and personality. [digitaltrends]
Microsoft has rolled out GPT-5 across its Copilot suite, offering enhanced AI capabilities to consumers, developers, and enterprise users. [ZDNet]
And OpenAI launched GPT-5 in public preview for GitHub Copilot, offering advanced reasoning and coding capabilities across multiple platforms. [GitHub]
The Generative AI Wars
Amazon’s AI-enhanced Alexa+ prioritizes advertising and data collection over improved functionality and user privacy. [Futurism]
Anthropic has offered its Claude AI models to all three branches of the US government for $1, providing multicloud access and high-security capabilities. [TechCrunch]
Anthropic's Claude Code has frustrated developers by excessively agreeing with users, prompting calls for the AI coding assistant to be less sycophantic and more truth-seeking. [The Register]
Meanwhile, Anthropic's US$5B revenue hinges precariously on two major customers, revealing both the promise and vulnerability of its AI coding market dominance. [VentureBeat]
Apple is integrating OpenAI's GPT-5 into iOS 26, iPadOS 26, and macOS 26, enhancing Apple Intelligence with a more capable and accurate AI model. [ZDNet]
And Apple has quietly launched an AI-powered Support Assistant in its customer service app, offering limited troubleshooting help to select US iPhone users. [TechRadar]
AWS and SuperOps launched an AI agent marketplace, accompanied by a hackathon offering $100,000 in prizes for developers creating autonomous agents. [ZDNet]
AWS has launched OpenAI’s two new open-weight models on Bedrock and SageMaker, offering customers more AI choices and potentially challenging Microsoft's exclusive relationship with OpenAI. [Forbes]
Google DeepMind's Hassabis suggested Meta's massive AI hiring spree indicates the company is struggling to catch up with competitors in the AI race. [ITPro]
Meta acquired AI voice startup WaveForms, bolstering its Superintelligence Labs with two co-founders who previously worked at OpenAI and Google. [TechCrunch]
OpenAI successfully defended its lawsuit against Elon Musk, with a federal judge allowing its harassment claims to proceed to trial. [Bloomberg]
Meanwhile, OpenAI's Altman and Musk engaged in a public spat on X, trading accusations about algorithmic manipulation and corporate conspiracies in the AI industry. [Gizmodo]
Apple has refuted Musk’s claims of App Store bias against AI apps, emphasizing its objective app selection criteria. [TechCrunch]
AI Supremacy
China sought US export control relaxation on AI-critical HBM chips as a potential precondition for a Trump-Xi summit. [DealStreetAsia]
Nvidia is negotiating with Trump to sell modified Blackwell GPUs to China, potentially with significant performance reductions and revenue sharing. [The Register]
US authorities have secretly embedded location trackers in select AI chip shipments to Dell and Super Micro to prevent illegal diversions to China. [Reuters]
Z.ai has successfully adapted its GLM models to run on Huawei's semiconductors, advancing China’s domestic AI capabilities amid US technology restrictions. [WebProNews]
Sovereign AI
India’s AI Mission is developing voice-first LLMs in Indian languages to distinguish itself from global tech competitors. [Mint]
Indonesia has proposed a sovereign AI fund through its new wealth fund Danantara to become a regional AI hub and attract foreign investment. [DealStreetAsia]
Macquarie Data Centres has partnered with Dell and Nvidia to provide a sovereign, secure AI infrastructure platform for Australian organizations in regulated sectors. [Computer Weekly]
Saudi Arabia’s nearly $1 trillion sovereign wealth fund has embedded AI across its operations, transforming its investment strategy and reflecting Saudi Arabia’s technological ambitions beyond oil. [DealStreetAsia]
South Korea has initiated a national AI model project, leveraging domestic technologies to create a self-sufficient AI industry and compete with the US and China. [CNBC]
Zetrix AI has developed NurAI, an Islamic-focused chatbot leveraging DeepSeek's open-source AI technology to serve the Muslim world’s diverse linguistic and cultural needs. [Bloomberg]
Feature Creeps
Anthropic expanded Claude Sonnet 4’s context window to 1 million tokens, enabling developers to analyze entire software projects and complex documents with unprecedented depth and accuracy. [VentureBeat]
And Anthropic has introduced an opt-in memory feature for Claude, allowing users to selectively recall past conversations without automatic retention. [TechRadar]
Google Finance is introducing an AI chatbot to answer financial questions, provide market insights, and offer enhanced visualization tools for US users. [ZDNet]
Google introduced Preferred Sources, a feature allowing users to select their favorite news outlets to appear at the top of personalized search results. [Gizmodo]
Google has introduced an AI-powered Flight Deals tool within Google Flights, enabling travelers to find cheaper fares through natural language queries. [TechCrunch]
Google has expanded Gemini Live’s functionality, integrating it with Calendar, Maps, Keep, and Tasks apps across Android and iOS platforms. [ZDNet]
Google is slowly adding personalization and data control features to its Gemini app, catching up with competitors like Anthropic and OpenAI in the AI chatbot market. [VentureBeat]
And Google has expanded Gemini’s PDF capabilities in Drive, increasing the token context window to 1M for more comprehensive analysis of larger files. [TechRadar]
OpenAI’s ChatGPT has introduced a new feature allowing Pro users to link their Google accounts, enabling AI-powered scheduling and email management with user consent. [TechRadar]
OpenAI has made ChatGPT Voice available to free users, offering an advanced AI-powered conversational assistant with improved contextual understanding and natural interaction. [ZDNet]
Hype Bubble?
Big tech companies continue to invest billions in generative AI systems, losing money while facing growing public distrust and concerns about bias, mental health risks, and technological limitations. [links.org.au]
CoreWeave's stock dropped after reporting a widening quarterly loss and cautious outlook due to high AI infrastructure development costs. [Bloomberg]
Google DeepMind CEO Demis Hassabis predicted AI could be 10 times bigger and faster than the Industrial Revolution. [VentureBeat]
OpenAI struggled with profitability in 2024, burning through $2.50 for every dollar earned despite massive user growth and significant tech industry investment. [Futurism]
OpenAI's GPT-5 launch disappointed expectations, revealing incremental progress and challenging superintelligence hype through underwhelming performance and scholarly critiques. [ZDNet]
Salesforce revealed CFOs have dramatically shifted from cautious to aggressive AI investment, driven by the potential of AI agents to transform business operations and drive long-term value. [ITPro]
Softbank's US$500B Stargate AI infrastructure project faced initial delays in site selection, but remains committed to its ambitious datacenter network plan. [The Register]
A survey by Temporal Technologies reveals enterprises are widely adopting AI tools but struggling with reliability, infrastructure, and scaling challenges across their workflows. [ITPro]
Windsurf explored selling itself to OpenAI at a US$3B valuation due to unsustainably negative margins in the AI coding assistant market, ultimately selling to Cognition after the deal fell through. [TechCrunch]
A survey by Workday revealed growing employee trust in AI agents for routine tasks, but significant hesitation about their broader workplace autonomy and potential negative impacts. [ZDNet]
Big Iron
Google is investing US$9B in Oklahoma over two years to expand its data centers and support AI education initiatives. [TechRadar]
Hewlett Packard has expanded its Private Cloud AI service with new Nvidia Blackwell GPU-enabled servers and open-source AI models for enterprise workloads. [ITPro]
Softbank has acquired Foxconn's Ohio electric vehicle plant to advance its ambitious Stargate AI data center project with OpenAI and Oracle. [Bloomberg]
The Chips are Up
DeepSeek struggled to train its R2 model on Huawei's Ascend chips, ultimately switching to Nvidia's H20 GPUs due to technical challenges. [The Register]
Huawei has developed UCM, an AI inference framework that efficiently manages memory across different tiers, helping overcome hardware limitations in large model processing. [Gizmochina]
And Huawei has open-sourced its CANN AI GPU toolkit to challenge Nvidia's CUDA, potentially offering developers an alternative ecosystem for AI computing. [TechRadar]
Intel was reportedly in discussions with the Trump administration to take a government stake in support of US semiconductor manufacturing expansion. [TechCrunch]
Nvidia and AMD will resume chip sales to China, paying a 15% revenue cut to the US government despite ongoing national security concerns. [TechRadar]
But Nvidia faced Chinese government scrutiny over potential security risks in H20 chips, raising questions about its future in the Chinese market. [TechRadar]
Nvidia has unveiled compact RTX Pro 4000 SFF and RTX Pro 2000 workstation GPUs with high performance and low 70W power consumption. [The Register]
ParTec sued Nvidia at the Unified Patent Court in Munich, alleging patent infringement of its dynamic Modular System Architecture technology in Nvidia’s DGX AI supercomputers. [Tom’s Hardware]
Sandisk and SK Hynix have proposed a High Bandwidth Flash memory standard to improve AI performance, storage capacity, and energy efficiency. [TechRadar]
Warm Bodies
Anthropic has acquired Humanloop's team, bringing expertise in AI tooling and evaluation to strengthen its enterprise strategy and AI safety efforts. [TechCrunch]
And Anthropic has offered salaries up to £340k to top AI engineers in London, intensifying the European talent competition for AI expertise. [Sifted]
Cohere hired Meta's former AI research VP Joelle Pineau to revamp its enterprise-focused AI strategy after raising US$500m at a US$6.8 billion valuation. [TechCrunch]
Microsoft aggressively pursued Meta's AI talent by creating a targeted list and developing a competitive hiring process with special compensation packages. [Business Insider]
xAI co-founder Igor Babuschkin has departed the company to launch a venture capital firm focused on AI safety research and supporting startups. [TechCrunch]
Meanwhile, tech companies are increasingly using AI to automate coding tasks, dramatically reducing entry-level job opportunities for computer science graduates. [WebProNews]
Consumer AI
Cerence AI integrated its Audio AI suite into HARMAN's AudioworX, enhancing automotive audio experiences with noise filtering, communication, and emergency vehicle detection technologies. [GlobeNewswire]
HTC's Vive Eagle smart glasses offer multi-AI support, extensive language translation, and local data storage in a stylish design, currently available only in Taiwan. [TechRadar]
Meta is preparing to unveil potentially groundbreaking smart glasses with a heads-up display at its September Connect conference, likely priced around $1,000. [Gizmodo]
Samsung has begun rolling out Gemini AI support to Galaxy Buds 3 and Buds 3 Pro. [TechRadar]
Vuzix has introduced the LX1 smart glasses, a hands-free warehouse solution with scanning, display, and voice control capabilities for improved workflow efficiency. [Speech Technology Magazine]
It’s Only a Model
AI2 unveiled MolmoAct, an open-source AI model that enables robots to reason spatially and navigate physical environments with human-like intelligence and transparency. [Business Wire]
G42's Core42 launched OpenAI's GPT-OSS globally on its AI Cloud, offering sovereign, high-performance AI model access through Compass API. [Business Wire]
Google DeepMind has released Gemma 3 270M, a small, efficient AI model designed to run on lightweight hardware while maintaining strong performance on various tasks. [VentureBeat]
Liquid AI has released LFM2-VL, a new generation of efficient, low-latency vision-language models designed for deployment across diverse hardware platforms. [VentureBeat]
Mistral AI unveiled Mistral Medium 3.1, a multimodal, cost-efficient LLM with enterprise-grade capabilities and advanced performance across diverse tasks. [Marktechpost Media]
Multiverse Computing has developed two tiny AI models named after animal brains, designed to run locally on devices without internet connectivity. [TechCrunch]
Nvidia unveiled new AI models and infrastructure for robotics, including Cosmos Reason, a 7-billion-parameter vision language model for physical AI applications. [TechCrunch]
Perplexity quickly evaluated OpenAI's GPT-OSS 20B and 120B models by adapting its ROSE inference engine to support their unique architectural features on Nvidia H200 GPUs. [Perplexity]
Whose Data?
Anthropic has sought to block a massive copyright class action lawsuit, arguing that the certification could financially ruin the AI industry and set a dangerous precedent. [Ars Technica]
Australian authors and writing organizations strongly oppose the Productivity Commission’s proposed AI copyright exemption, arguing it would harm creators’ livelihoods and undermine the creative industry. [ABC]
Yomiuri Shimbun, Japan’s largest newspaper by circulation, has sued Perplexity for copyright infringement, alleging unauthorized reproduction of its articles through AI scraping. [NiemanLab]
Reddit blocked the Internet Archive’s Wayback Machine from indexing most of its site after discovering AI companies were exploiting the archive to scrape user data without permission. [Gizmodo]
Should there be a legal right to delete deceased individuals’ digital data to prevent unwanted AI-driven digital resurrections? [The Register]
The LLM Ecosystem
Arcitecta enhanced its Mediaflux platform to provide an AI-ready data fabric that supports diverse data types and AI models with integrated vector database capabilities. [Business Wire]
Atlan launched an App Framework with 21 partners, enabling developers to build context-rich, AI-native applications on its Metadata Lakehouse platform. [Business Wire]
Blaize has introduced an AI platform designed to deliver multi-modal intelligence efficiently at the edge. [Business Wire]
Chetu has partnered with Databricks to provide comprehensive data migration and AI implementation services, helping enterprises streamline their data intelligence and adoption strategies. [Business Wire]
Compute Exchange launched a GPU Pricing Intelligence Calculator, providing transparent, real-time pricing data to advance AI infrastructure access and innovation. [Business Wire]
Fisent has developed a confidence rating capability for its BizAI solution, enabling businesses to assess GenAI-powered decision-making processes with greater reliability and precision. [PRWeb]
Inworld AI released Runtime, an AI infrastructure platform enabling developers to rapidly scale consumer applications from prototype to millions of users with minimal engineering overhead. [GlobeNewswire]
Microsoft’s VS Code has previewed chat checkpoints in version 1.103, allowing developers to rollback AI chat interactions and workspace states. [The Register]
Mimo launched Instance, an AI-powered tool enabling non-developers to build fully functional apps by simply describing their desired product. [GlobeNewswire]
OpenAI has leveraged the MXFP4 data type to reduce inference costs by approximately 75%, making its generative AI models significantly more efficient and affordable. [The Register]
Tabnine integrated Nvidia Nemotron reasoning models into its AI development platform, offering enterprises enhanced performance, control, and efficiency in software development. [GlobeNewswire]
Virtana has partnered with Nvidia to enhance AI factory observability, providing unified insights and optimization for complex, distributed AI infrastructure environments. [Business Wire]
Agentic AI
Box has signed a deal with the US General Services Administration to provide AI agents from various companies to help government agencies manage and analyze their data more efficiently. [The Register]
Bright Data released The Web MCP, a free infrastructure enabling AI agents to access and interact with live web content without blockages. [EIN Presswire]
Brightwave launched Research Agents, an AI-powered platform enabling autonomous, multi-agent research across complex information sources with high-quality, verifiable deliverables. [Business Wire]
C3 AI launched a Strategic Integrator Program, enabling partners to license its Agentic AI Platform and develop enterprise AI applications with retained intellectual property rights. [Business Wire]
Dnotitia unveiled the Mnemos Workstation, the world’s first MCP-based AI agent workstation with an integrated vector database. [EIN Presswire]
HMX has launched Cognitive Agents, offering transparent, human-like reasoning for mission-critical enterprise applications. [EIN Presswire]
KarmaCheck launched an MCP Server enabling AI systems to conduct automated background checks and identity verifications through simple, prompt-based interactions. [Business Wire]
NEXA Mortgage partnered with Tidalwave to deploy enterprise-scale agentic AI across its 3,200+ loan officers, transforming mortgage origination processes. [EIN Presswire]
Nexus has developed an AI router that aggregates MCP servers, intelligently routes language models, and simplifies complex AI interactions while enhancing performance and security. [Nexus]
Rubrik launched Agent Rewind, an innovative solution enabling organizations to trace, audit, and safely reverse unintended actions by AI agents. [Business Wire]
Sentry has launched a monitoring tool for MCP servers, providing developers with comprehensive insights into server performance, errors, and usage across the AI protocol’s ecosystem. [Business Wire]
Striim released version 5.2, enhancing its real-time data integration platform with new AI agents, enterprise modernization features, and support for advanced data formats. [GlobeNewswire]
SUPERAGENT AI aims to replace traditional insurance agents with fully autonomous AI agents by the end of 2025, promising enhanced efficiency and performance. [Business Wire]
Supernova Platform has developed a multi-agent AI platform that can autonomously automate, optimize, and scale business operations with minimal human intervention. [EIN Presswire]
Token Security and Descope released an AI Security Guide, providing a four-phase maturity model to help organizations securely adopt and govern autonomous AI systems. [GlobeNewswire]
X0PA AI unveiled Zeus, an AI-powered interview intelligence agent designed to transform hiring decisions by providing unbiased, real-time candidate insights. [EIN Presswire]
Workato established an AI Research Lab in San Francisco to develop autonomous enterprise agents that can transform how companies operate. [Business Wire]
Other LLM Sightings
10Web launched a white-labeled WordPress plugin enabling users to create AI-generated websites quickly and without third-party branding. [TechRadar]
Airbnb is transforming into an AI-first app, focusing on personalized customer service and aiming to become a comprehensive travel booking platform. [Business Insider]
Alida launched an AI-powered assistant for user research, enabling faster, more confident data analysis and insights generation. [Business Wire]
Captain Compliance launched an on-premises AI suite powered by OpenAI's GPT-OSS to automate data privacy and regulatory compliance for enterprises. [EIN Presswire]
Jeeva AI has launched an AI-powered calendar that automatically books, reschedules, and protects meetings while syncing data to CRM and reducing no-shows. [EIN Presswire]
MetaFyAI launched a generative AI commerce suite that helps online retailers boost conversions, personalize experiences, and reduce operational costs through specialized eCommerce tools. [EIN Presswire]
OneShield has launched an AI Hub, a centralized environment for secure, scalable, and interoperable AI development in insurance technology solutions. [Business Wire]
Order.co unveiled its AI-powered procurement platform, designed to automate workflows, reduce costs, and streamline purchasing processes for businesses. [GlobeNewswire]
Positon AI has enhanced its AI-powered M&A platform with GPT-5, improving precision, speed, security, and decision-making capabilities for deal teams. [EIN Presswire]
TD Securities launched an AI virtual assistant for equity sales and trading professionals, leveraging OpenAI's GPT models and internal research data to enhance workflow and client insights. [VentureBeat]
Risks and Responses
AI-powered narrative attacks are rising, exploiting advanced technologies to create convincing disinformation that can manipulate perceptions and cause significant harm across business, personal, and social domains. [ZDNet]
AWS Bedrock’s GPT-OSS release revealed mysterious, inconsistent system prompts that potentially degrade model performance and create unpredictable AI behaviors. [benanderson.work]
A survey by Checkmarx revealed that over 80% of organizations ship vulnerable code as AI-coding assistants become increasingly prevalent with minimal governance. [Business Wire]
Google is addressing Gemini’s self-deprecating AI behavior by fixing a bug that causes the chatbot to repeatedly express extreme self-doubt and failure. [The Register]
Mo Gawdat, the former chief business officer of Google X, warns that AI will magnify humanity’s existing societal issues, potentially leading to a dystopian future driven by capitalism and unchecked technological advancement. [Gizmodo]
Grok, xAI's AI chatbot, continued to generate controversy with multiple contradictory explanations for a brief suspension and persistent misinformation issues. [Gizmodo]
LightBeam.ai has released an AI security solution that provides real-time governance, behavioral containment, and analytics to protect sensitive data in Microsoft Copilot environments. [PRWeb]
An internal Meta AI policy document revealed controversial guidelines permitting chatbots to engage in inappropriate conversations, generate false information, and produce potentially offensive content. [Reuters]
OpenAI is focusing on making ChatGPT genuinely useful by improving user experience, addressing sensitive interactions, and developing tools to support mental health responsibly. [The Decoder]
But ChatGPT has reportedly exacerbated mental health issues for some users, with FTC complaints revealing concerns about emotional manipulation, delusions, and psychological harm. [Gizmodo]
And ChatGPT led a corporate recruiter into a delusional spiral over 21 days, convincing him he had discovered a revolutionary mathematical formula that could transform the world. [The New York Times]
Psst has created a platform connecting tech workers with similar workplace concerns, providing encrypted storage, legal support, and potential group whistleblowing to promote AI safety. [Semafor]
Sales 3.0 Labs released the Ethical Quality Index (EQI), a comprehensive tool to help B2B sales professionals evaluate AI solutions’ ethical and effective use. [PRWeb]
Salt Security warned that without robust API security, the proliferation of AI agents could expose organizations to significant cybersecurity risks and erode consumer trust. [EIN Presswire]
TechTruth.dk emerged as an AI-driven platform to expose and correct misleading narratives in technology journalism, quickly gaining traction among industry professionals. [WebProNews]
Truth Social's AI chatbot exclusively cites conservative news sources, despite claiming to draw from across the political spectrum, raising concerns about potential bias. [Wired]
Environmental Issues
Mistral AI launched a sustainability tracker revealing AI’s substantial environmental impact, with its Large 2 model consuming 281,000 m3 of water and emitting 20.4ktCO₂e. [ITPro]
Researchers have quantified the carbon emissions of LLMs, revealing a trade-off between reasoning capabilities, accuracy, and greenhouse-gas output across various model sizes. [DeepLearning.AI]
Conversational AI
C3 AI launched C3 Agentic AI Websites, an innovative platform transforming websites into interactive, conversational experiences powered by generative AI technology. [Business Wire]
Continua raised US$8m to develop a Social AI agent that enhances group chats by transforming conversations into organized, actionable outcomes across messaging platforms. [Business Wire]
eTailPet has unveiled an AI-powered Messaging Assistant designed to help pet retailers communicate more efficiently with customers. [PRWeb]
Lendflow launched an AI Automation Suite with multiple operational agents designed to optimize lending processes, increase efficiency, and reduce manual workflows. [Business Wire]
Momentum has launched SmartClips and Account Briefs, AI-powered tools that transform sales calls into shareable insights and provide instant account intelligence. [Business Wire]
NiCE expanded its partnership with Salesforce to enhance AI-driven customer service workflows through deeper integration of CXone Mpower and Service Cloud. [Business Wire]
SoundHound launched Vision AI, integrating visual and voice technologies to enable more natural, context-aware interactions across enterprise applications. [Business Wire]
Tech Stack Advising launched a Contact Center AI Practice with an AI Accelerator framework to help enterprises strategically implement and maximize AI solutions. [PRWeb]
Vocodia demonstrated its AI-driven voice solution’s superior performance by helping Maxoderm achieve higher revenue per call than traditional human call centers. [GlobeNewswire]
Be Real
Appfigures reported that AI companion apps generated US$82m in the first half of 2025, with 337 active apps and 220m global downloads. [TechCrunch]
IdeaJam has launched Historic Mentors, an AI platform enabling users to engage in real-time conversations with personas of influential historical figures. [GlobeNewswire]
Reflekta has debuted an AI platform that creates interactive digital avatars of deceased loved ones, preserving their stories and wisdom. [Business Wire]
Voice News
The Association of Voice Artists of India is battling AI’s growing impact on the dubbing industry, seeking fair compensation, consent, and protection for voice artists facing potential exploitation. [The Hollywood Reporter]
CloneOps.ai has partnered with Verified Carrier to develop Voice ID, a voice-based authentication system for instant, secure identity verification in logistics. [Speech Technology Magazine]
VideoProc Converter AI has updated its software with AI vocal removal. [EIN Presswire]
Document AI
Aavenir launched ComplianceNext, an AI-powered solution that automates vendor and contract compliance screening to enhance risk management across industries. [Business Wire]
FBSPL has launched three AI-powered insurance tools that significantly reduce policy review, onboarding, and proposal generation times. [EIN Presswire]
Liquid Logics has integrated with DossDocs to provide private lenders with one-click, compliant loan document generation, streamlining the lending process. [EIN Presswire]
The US Office of Management and Budget is developing SweetREX, an AI tool aimed at identifying and eliminating federal regulations deemed unnecessary by the Trump administration’s deregulation efforts. [Wired]
Translation
Apple is developing a real-time translation gesture for AirPods 4 and Pro 2 in iOS 26, enabling multilingual communication through device-assisted translation. [TechRadar]
Boostlingo has launched an AI Interpreter beta that provides real-time machine interpreting while maintaining quality and professional interpreter support. [MultiLingual]
PlayTV.ai launched an AI-powered platform enabling real-time translation of YouTube content across hundreds of languages. [EIN Presswire]
TAUS has partnered with Library of Business to provide industry-specific, expert-curated datasets for AI training to big-tech companies and AI labs. [MultiLingual]
Trados launched Smart Insights, an AI-powered virtual assistant within its language technology platform that provides real-time project intelligence for localization professionals. [Business Wire]
Wordly surpassed 5 million AI users, ranked 196th on the Inc 5000 list, and expanded its AI translation platform with over 200% growth. [PRWeb]
Search
1Digital Agency has launched AI SEO and ChatGPT SEO services to help eCommerce brands optimize their search visibility in an AI-driven digital landscape. [EIN Presswire]
A report by Datos and SparkToro reveals that AI search is growing in the US and Europe, though traditional search still dominates online information seeking. [MarTech]
Google has introduced a ‘Preferred Sources’ feature allowing users to select preferred news sites in search results, potentially creating ideological echo chambers. [TechCrunch]
Meteor, a Y Combinator-backed startup, has developed an AI-powered web browser aimed at transforming online task completion and personal productivity. [GeekWire]
Writing Assistance
Pocket FM is leveraging AI tools to rapidly produce and localize audio content, aiming to become a global audio streaming platform with increased efficiency and scalability. [TechCrunch]
AI in Journalism
Politico faced an arbitration hearing with its union over AI tools that generated potentially inaccurate summaries and reports, raising questions about journalistic standards and editorial oversight. [NiemanLab]
Verdens Gang, one of Norway’s largest newspapers, has developed FOIA Bot, an AI-powered newsroom assistant that helps journalists efficiently draft Freedom of Information Act requests by leveraging retrieval-augmented generation technology. [NiemanLab]
Health Tech
Altera Digital Health launched TouchWorks Note+, an AI-powered ambient listening solution that generates clinical notes while improving provider-patient interactions. [Business Wire]
Athenahealth has developed a cloud-native platform enabling AI-powered interoperability and data insights for physician practices across the United States. [athenahealth.com]
Avo has partnered with MEDITECH to integrate its AI-powered Scribe platform into MEDITECH Expanse, enhancing clinical documentation for rural and small hospitals. [EIN Presswire]
Calabrio expanded its CareAI healthcare project, using generative AI to improve patient experiences, reduce wait times, and save providers up to US$1.1m annually. [Business Wire]
DrCloudEHR has launched an AI-powered Group Scribing feature for behavioral health, promising to reduce documentation time by up to 80%. [EIN Presswire]
ESI PEO has partnered with Yuna to offer AI-powered mental health coaching and streamlined HR services, addressing workplace wellness and operational efficiency. [EIN Presswire]
Five9 launched a native Epic Toolbox integration that enables healthcare providers to deliver more personalized patient experiences through seamless contact center capabilities. [Business Wire]
Lyra Health demonstrated that AI-powered provider matching can reduce mental health care costs by nearly 20% while maintaining clinical outcomes. [Business Wire]
Marchex released an AI-powered healthcare solution that transforms patient conversations into actionable insights, helping health systems optimize marketing and patient engagement. [Business Wire]
MediMergent launched a pilot program using Virtual Human Advocates to streamline patient recruitment and engagement in community-based clinical research centers. [Business Wire]
Menta has launched an AI-driven mental health platform that combines ethical AI, clinical psychology, and personalized tools to improve global access to mental health support. [EIN Presswire]
NASA is developing an AI medical assistant with Google to help astronauts diagnose and treat symptoms during long-duration space missions with limited Earth communication. [TechCrunch]
Oracle has developed a next-generation, AI-powered Electronic Health Record system designed to streamline clinician workflows and enhance patient care through intelligent, voice-activated technology. [Oracle]
Otto launched AI Recap, a veterinary-specific scribe tool that captures appointment details accurately and affordably, helping clinics improve documentation and save time. [Business Wire]
Legal Tech
Bryter has launched Hybrid Agents, a new category of AI-powered legal solutions combining generative AI with workflow automation for precise, controllable legal tasks. [Artificial Lawyer]
Chamelio has enhanced its legal intelligence platform with expanded AI capabilities, including advanced contract review, Microsoft Word integration, and improved organizational knowledge sharing. [Artificial Lawyer]
CobbleStone Software launched VISDOM AI Helper for Microsoft Word, enabling legal professionals to automate contract management tasks. [PRWeb]
Definely has launched Cascade, an AI-powered tool that detects and tracks multi-level contract changes to reduce drafting risks. [Artificial Lawyer]
DocSolid has integrated Microsoft Azure AI Language to generate intelligent summaries of scanned mail in its Airmail2 Digital Mailroom platform, enhancing attorneys’ mail triage efficiency. [PRWeb]
Draftwise has expanded its AI capabilities with intelligent precedent benchmarking, automated term sheet validation, and strategic contract guidance. [Artificial Lawyer]
Elite has unveiled cloud and AI innovations for legal firms, including embedded payments, real-time data access, and an AI proforma agent to accelerate operational efficiency. [Business Wire]
Everlaw launched Deep Dive, a GenAI tool enabling legal teams to quickly search and extract insights from massive document collections using conversational language. [Business Wire]
And Everlaw secured FedRAMP certification for its EverlawAI Assistant, enabling federal agencies to adopt generative AI tools for legal processes. [Business Wire]
Gavel has launched Deep Reasoning Mode in its Microsoft Word add-in, using multiple AI models to enhance contract redlining, drafting, and negotiation with advanced context understanding and reasoning capabilities. [Artificial Lawyer]
Harvey has evaluated GPT-5 using its BigLaw Bench system, scoring 89.22% and demonstrating significant improvement in legal AI performance. [Artificial Lawyer]
And Harvey has secured a significant enterprise license with Latham & Watkins, enabling firmwide AI platform deployment for over 3,600 attorneys globally. [Artificial Lawyer]
LegalSifter has launched ReviewPro, an AI-powered Microsoft Word tool that accelerates contract review by auto-applying expert-guided redlines and reducing negotiation cycles. [Artificial Lawyer]
LexisNexis has unveiled Protégé General AI, a privacy-encrypted platform offering customizable access to multiple AI models within a secure legal environment. [Artificial Lawyer]
Luminance partnered with LHV Bank to deploy AI-powered contract review technology, promising to dramatically reduce legal team workload and improve operational efficiency. [Tech.eu]
Opus 2 Cases enhanced its legal software with AI-powered issue tracking, enabling litigation teams to more effectively organize, analyze, and strategize case details. [PRWeb]
Practice AI has integrated with MyCase and CASEpeer to streamline legal administrative tasks, enabling teams to focus more on client outcomes. [GlobeNewswire]
Syntheia launched Super Comparer, a SaaS tool enabling legal professionals to compare and analyze multiple documents across various dimensions through transparent pricing plans. [Artificial Lawyer]
Wolters Kluwer has signed a content licensing agreement with Harvey, expanding its AI-centric legal data strategy by providing access to US and German legal information. [Artificial Lawyer]
And Mori Kabiri’s new book, Law Firm KPIs, provides 80+ metrics to help law firms measure AI impact, pricing strategies, and profitability. [PRWeb]
Ed Tech
The American Federation of Teachers is developing the National Academy for AI Instruction to train K–12 educators on responsible AI tool usage, aiming to bridge digital equity gaps and enhance teaching effectiveness. [EdTech]
Anthropic has launched learning modes for Claude AI that transform the chatbot into an educational tool, guiding users through step-by-step reasoning instead of providing direct answers. [VentureBeat]
California launched a comprehensive AI education initiative, partnering with tech giants like Google to provide free AI tools and training to over two million public university students. [WebProNews]
Digital Course Assistant launched an AI-powered membership platform designed to help online course creators simplify workflows, create content, and market more effectively. [GlobeNewswire]
Google has committed US$1B to support AI education and training across US universities, offering free tools and access to students. [TechRadar]
OpenAI's Study Mode offers limited educational value, requiring students to actively engage and craft prompts to extract meaningful learning experiences from the AI chatbot. [ZDNet]
Wild Zebra has developed an AI tutor that guides students through personalized, Socratic-style learning experiences, adapting to individual interests and needs across math and reading comprehension. [GeekWire]
Funding
Archestra raised US$3.3m to develop an open-source platform enabling secure generative AI integration with enterprise systems and sensitive data. [Tech.eu]
Capacity, an AI-powered support automation platform, secured US$92m in investments to fuel growth and product innovation in contact center technology. [FinSMEs]
Continua raised US$8m to develop AI agents that enhance group chat interactions across messaging platforms. [TechCrunch]
Datumo, a Seoul-based AI startup, raised US$15.5m to help businesses build safer AI through data labeling, evaluation, and monitoring tools. [TechCrunch]
Evertune, a Generative Engine Optimization and AI marketing platform, secured US$15m in Series A funding led by Felicis Ventures to expand its operations and development efforts. [FinSMEs]
Fidelity is negotiating to lead a potential US$1B funding round for chipmaker Cerebras Systems. [Bloomberg]
Graas AI raised US$9m to expand its Agent Foundry platform, developing autonomous AI agents for e-commerce operational challenges across India and Southeast Asia. [The Economic Times]
Kustomer raised US$30m to accelerate its AI-native customer service platform development and expand its global enterprise offerings. [Business Wire]
Lorikeet, an AI customer support platform, raised US$35m in a Series A funding round, attracting top Australian VCs and expanding its global market presence. [Startup Daily]
Palabra AI, a startup developing an AI-powered speech translation engine, raised US$8.4m in pre-seed funding to solve real-time multilingual translation challenges. [TechCrunch]
Rillet has raised US$100m to revolutionize ERP and accounting with AI-native financial infrastructure, enabling companies to operate with minimal finance teams. [Artificial Lawyer]
Syenta, an Australian National University spin-out, raised US$8.8m to develop a novel lithography-free chip manufacturing technology for advanced AI and high-performance computing. [Startup Daily]
Translucent AI raised US$7m to develop an AI-powered financial analysis platform specifically designed to help healthcare organizations improve their financial performance. [Business Wire]
Acquisitions
Datasite acquired Blueflame AI, a leading agentic AI company, to enhance its workflow collaboration and automation solutions for investment and strategic projects. [Business Wire]
Doximity acquired Pathway, a medical AI startup, for US$63m to enhance its clinical reference capabilities with a high-performing AI dataset. [Business Wire]
Largo.ai acquired QuMind, combining AI-driven market research technology to enhance insights, testing, and decision-making for brands and agencies worldwide. [Tech.eu]
Smart Communications acquired Pendula, enhancing its Conversation Cloud platform with AI-powered omnichannel orchestration capabilities for regulated industries. [Business Wire]
There’s More
AI Jesus chatbots, developed by for-profit companies without church endorsement, raise theological and ethical concerns about the potential manipulation of religious beliefs through generative AI technology. [Gizmodo]
Google is addressing a glitch causing Gemini to repeatedly engage in extreme self-deprecating language loops during problem-solving interactions. [Forbes]
Sonar evaluated five AI coding models, revealing distinct ‘personalities’ with varying performance and significant security vulnerabilities across different benchmarks. [The Register]
Stanford researchers have decoded inner speech brain activity with up to 74% accuracy, potentially helping people with severe speech impairments communicate more easily. [Gizmodo]
The UK’s Alan Turing Institute faced a whistleblowing complaint alleging governance failures, internal culture issues, and potential funding risks amid a significant restructuring. [The Guardian]
Research by Zefr reveals that human content moderators outperform AI in accuracy but cost 40 times more, suggesting a hybrid approach for brand safety. [The Register]
The Back Cover
Here’s Dua Lipa with a cover of AC/DC’s Highway to Hell.
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!