This Week in NLP #354
Keep up with what happened in NLP in the week ending Friday 29th August 2025.
Above the Fold
Overwhelmed? Here’s our pick for five things to know about this week.
Anthropic has launched a limited research preview of Claude for Chrome, a browser extension offering AI-powered web browsing with significant security risks and warnings. [The Register]
Apple is exploring integrating Google's Gemini AI into Siri, potentially transforming its digital assistant after struggling to develop its own AI architecture. [TechRadar]
Intel received a US$9B government investment, but analysts doubt this will solve its foundry challenges without securing external customers for its advanced manufacturing processes. [Reuters]
xAI open-sourced its Grok 2.5 AI model on Hugging Face, with Musk promising Grok 3 will follow in six months. [TechCrunch]
And xAI quietly terminated its Nevada public benefit corporation status, abandoning its commitment to social and environmental benefits. [CNBC]
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
Alibaba has aggressively invested in AI but struggles to monetize its efforts. [DealStreetAsia]
A report by Andreessen Horowitz reveals Google's Gemini and other AI rivals are rapidly gaining ground on ChatGPT in consumer AI adoption across web and mobile platforms. [TechCrunch]
Google has positioned itself as a crucial AI ecosystem player by providing cloud infrastructure and search data to major tech companies like Meta and OpenAI. [The Decoder]
Google is pushing employees to adopt AI tools or risk falling behind in the competitive tech landscape, warning of potential career consequences. [Hans India]
Macrohard, Elon Musk’s new AI venture, playfully targets Microsoft while aiming to develop a purely AI-driven software company. [Windows Central]
Meta has licensed Midjourney's AI image and video generation technology to enhance its own AI models and products. [TechCrunch]
OpenAI reached US$1B in monthly revenue in July and expects to triple its annual revenue to US$12.7B by 2025. [The Decoder]
OpenAI has launched a centralized online hub providing startups with resources, API credits, events, and guidance for building with its technologies. [Tech.eu]
OpenAI is already developing GPT-6, which will feature enhanced personalization through an improved Memory feature and potentially faster release than previous model iterations. [ZDNet]
OpenAI revealed that Elon Musk sought Mark Zuckerberg’s involvement in a $97.4-billion bid for the company earlier this year, which Zuckerberg declined. [TechCentral]
And OpenAI has announced plans to open its first office in India, aiming to tap into the country’s growing AI market and develop localized AI solutions. [TechCrunch]
xAI has filed a lawsuit against Apple and OpenAI, alleging anticompetitive practices that prevent Grok from succeeding in the App Store rankings. [Gizmodo]
And Musk’s lawyers sought to prevent OpenAI from obtaining bid-related documents from Meta in an ongoing legal dispute over the AI company’s ownership and direction. [Reuters]
xAI's Grok 2.5 release fails to meet open-source standards due to restrictive licensing that limits model training and commercial use. [ZDNet]
AI Supremacy
Beijing has unveiled an ambitious AI strategy aimed at boosting domestic tech industries, with goals of widespread AI adoption and support for local semiconductor and AI chip manufacturers. [Yahoo Finance]
Chinese regulators pressured tech giants to halt Nvidia H20 chip orders after US Commerce Secretary’s ‘addicted’ comment sparked diplomatic tension. [Notebookcheck]
DeepSeek has developed a new model training technique, potentially advancing China’s AI self-sufficiency by reducing computational requirements. [Yahoo Finance]
Intel announced an agreement with the Trump administration to convert previously awarded government grants into an US$8.9B stock investment. [TechCrunch]
Sovereign AI
Accenture found that UK multinationals lead European AI adoption, with nearly half scaling strategic initiatives, though the US still outpaces them technologically and economically. [Consultancy.uk]
This piece argues that Britain should build a sovereign compute infrastructure with UK-developed IP to gain technological independence, economic leverage, and strategic control in the emerging global compute landscape. [Sifted]
Egune AI has developed a Mongolian LLM with limited resources to serve local needs, preserve cultural identity, and demonstrate the potential of small-country tech innovation. [Rest of World]
Europe boasts a ‘swarm’ of global-scale AI companies, according to a leading VC. [Tech.eu]
Humain broke ground on its first Saudi data centers, set to open in 2026 using US-imported semiconductors in Riyadh and Dammam. [Bloomberg]
And Humain launched an Arabic-native AI chatbot designed to be culturally fluent and potentially subject to governmental content restrictions. [Gizmodo]
OpenAI has discussed a potential multibillion-pound deal with the UK technology secretary to provide nationwide premium access to ChatGPT. [The Guardian]
ResetData launched Australia’s most powerful public sovereign AI supercomputer, AI-F1. [EIN Presswire]
South Korea initiated a national AI development contest, selecting five teams to create a sovereign AI model to reduce technological dependence. [Dong-A]
SkyeChip developed Malaysia’s first domestic edge AI processor, MARS1000, marking a technological milestone in the country’s AI ambitions. [TechCrunch]
Feature Creeps
Apple is expanding enterprise AI controls, allowing IT administrators to configure external AI provider access and implement granular device management features. [TechCrunch]
Google's AI Mode has begun helping users find restaurant reservations across multiple platforms, though it cannot complete the final booking step. [Inc.]
Google is enhancing Gemini Live with screen highlighting, app integrations, and improved audio capabilities for more interactive and nuanced AI conversations. [The Verge]
Google launched an on-premise Gemini AI solution through Google Distributed Cloud, enabling enterprises to deploy advanced language models securely within their own data centers. [ZDNet]
Google’s NotebookLM has enhanced its Audio Overview tool with customizable response lengths, guiding prompts, and improved language settings in its mobile app. [ZDNet]
And Google expanded NotebookLM’s Video and Audio Overviews to support 80 languages, enhancing global accessibility for users seeking multimedia summaries. [TechCrunch]
Microsoft is testing an AI-powered ‘Journeys’ feature for Edge browser that summarizes browsing activity, raising privacy concerns. [TechRadar]
And Microsoft has updated Visual Studio 2022 to give developers more control over Copilot’s code suggestions and completions. [The Register]
OpenAI’s ChatGPT Projects now offers project-only memory, allowing users to create focused, self-contained workspaces with context-specific information. [TechRadar]
OpenAI has enhanced its Realtime API and launched gpt-realtime, a more advanced speech-to-speech model enabling developers to create more sophisticated voice agents. [ZDNet]
Perplexity has introduced Comet Plus, a $5 subscription offering premium content access and a new revenue model that compensates publishers across human, search, and agent internet traffic. [Perplexity]
WhatsApp introduced an AI-powered Writing Help feature that allows users to rephrase, proofread, or adjust the tone of their messages. [TechCrunch]
Hype Bubble?
Gartner predicts that 40% of enterprise applications will incorporate AI agents by 2026, while cautioning against rushed adoption and highlighting potential risks. [ZDNet]
MIT and Harvard researchers developed a metric to assess AI systems’ ability to understand underlying principles, finding that current predictive models struggle to generalize beyond specific tasks as complexity increases. [MIT News]
Research by MIT revealed that most companies have yet to see meaningful returns from AI, suggesting the technology’s potential has been overhyped and underdelivered. [The Register]
Nvidia's CEO Jensen Huang expressed confidence in the AI market’s future, predicting US$3-4T in infrastructure spending by decade’s end despite bubble concerns. [ITPro]
Prominent AI researcher Stuart Russell warns that overheated AI expectations could trigger a sudden industry collapse, reminiscent of the 1980s AI winter. [The Decoder]
Wall Street remains bullish on AI’s potential, with major tech firms increasing capital spending despite OpenAI CEO Sam Altman’s warnings about potential investor overexcitement. [Fortune]
Big Iron
Meta has signed a US$10B cloud deal with Google, expanding its multicloud strategy to support its growing AI infrastructure needs. [TechRadar]
Meta is planning a massive US$50B data center in Louisiana, as highlighted by President Trump during a Cabinet meeting. [Bloomberg]
Oracle has secured major AI infrastructure deals with OpenAI, Nvidia, and xAI, transforming its cloud business and propelling Larry Ellison to become the world’s second-richest person. [Bloomberg]
The Chips are Up
Over 100 companies are developing AI processors, with expectations of consolidation to around 25 survivors by decade’s end. [The Register]
Cambricon reported record earnings driven by increased demand for Chinese AI chips following the DeepSeek AI breakthrough. [Bloomberg]
IBM and AMD are partnering to develop a quantum computing architecture that integrates their specialized AI and quantum technologies to advance computing capabilities. [TechCrunch]
Nvidia and AMD faced a cool reception in China after obtaining permission to sell AI processors, with the government requiring security reviews and pushing for domestic alternatives. [DeepLearning.AI]
Nvidia lamented potential US$50B annual loss from US-China trade restrictions while reporting strong Q2 earnings and continued optimism about AI infrastructure spending. [The Register]
Nvidia halted H20 chip production for China after Beijing raised security concerns, potentially impacting its significant market share in the country. [International Business Times]
Nvidia engaged in discussions with the Chinese government over H20 chip security concerns, expressing optimism about resolving the issue. [Nikkei Asia]
Nvidia is seeking to sell its Blackwell AI chip to China, lobbying the Trump administration to ease export restrictions on high-end semiconductor technology. [Computer Weekly]
Nvidia has introduced the GB10 superchip, a compact, $2,999 AI development platform designed to bring datacenter-like AI capabilities to local workstations. [The Register]
And Nvidia has unveiled the Jetson Thor, a powerful developer kit designed to drive physical AI and advanced robotics across multiple industries. [TechRadar]
Warm Bodies
Meta restructured its AI research, with Yann LeCun now reporting to Alexandr Wang, who leads the new Superintelligence Lab. [The Decoder]
And Meta's Superintelligence Lab experienced early departures, with at least three AI researchers leaving, including two returning to OpenAI, just months after the lab’s announcement. [Wired]
OpenAI has hired Fidji Simo as CEO of Applications to manage the company’s consumer tech operations, freeing Sam Altman to focus on broader technological initiatives. [The Verge]
A poll by Reuters/Ipsos reveals 71% of Americans fear AI will permanently displace workers. [ZDNet]
Stanford researchers discovered that AI has disproportionately impacted younger workers’ job opportunities in certain sectors, while more experienced employees remain relatively unaffected. [Wired]
Consumer AI
Google has enhanced its Pixel 10 lineup with AI features like Tensor G5, Gemini Nano, Magic Cue, and Voice Translate, offering advanced on-device capabilities. [The Decoder]
Google is preparing to launch a new Nest smart speaker, doorbell, and security cameras with AI-powered features to compete with Amazon's smart home devices. [TechRadar]
Meta is set to unveil its first consumer-ready smart glasses with a display and a wristband for hand gesture control at its Connect conference in September. [CNBC]
A Microsoft designer has conceived Copilot Veja, an audio-focused wearable with cameras that provides AI assistance without a visual display. [TechRadar]
Nvidia unveiled the Jetson AGX Thor Developer Kit, a compact mini PC with a T5000 system-on-module delivering 2070 TFLOPS for advanced robotics and AI workflows. [TechRadar]
Plaud.ai released the Note Pro, a credit-card-sized AI notetaker with improved recording capabilities, a small screen, and enhanced app features, priced at $179. [TechCrunch]
Rokid launched smart glasses with a Micro LED display, priced at $599, potentially challenging Meta's upcoming smart glasses. [TechRadar]
Samsung is bringing Microsoft's Copilot AI to its 2025 TVs and monitors, offering personalized recommendations and interactive experiences through Tizen OS. [TechRadar]
Tesla has partnered with DeepSeek and ByteDance's Doubao to develop AI tools for its Chinese car market. [Gizmodo]
xMEMS unveiled AI glasses prototypes featuring Sycamore loudspeakers and µCooling technology to enhance wearable performance and comfort. [Business Wire]
It’s Only a Model
AiM Future and Franklin Wireless signed an MOU to develop a lightweight AI model and high-efficiency 1 TOPS AI SoC chipset for the North American market. [Business Wire]
Cohere has released Command A Reasoning, an enterprise-focused LLM designed for multilingual, tool-integrated reasoning with flexible performance and strong safety features. [VentureBeat]
Google has upgraded its Gemini chatbot with a new AI image model that offers more precise and seamless photo editing capabilities. [TechCrunch]
Microsoft’s VibeVoice is an open-source text-to-speech framework capable of generating expressive, long-form, multi-speaker conversational audio up to 90 minutes long. [Hugging Face]
Whose Data?
Amazon has blocked AI bots from Meta, Google, and other companies from scraping its e-commerce data through its robots.txt file. [ModernRetail]
Anthropic reportedly reached a preliminary settlement with authors and publishers in a class action lawsuit over alleged illegal use of copyrighted books to train its AI models. [Publishers Weekly]
Anthropic’s settlement of the copyright lawsuit potentially sets a precedent for AI companies’ use of copyrighted materials. [Fortune]
The Media, Entertainment and Arts Alliance welcomed the Tech Council of Australia’s acknowledgment of media and creative workers’ rights in AI content usage, calling for legislative action and fair compensation. [The National Tribune]
OpenAI reportedly used Google Search data via a third-party tool to enhance ChatGPT’s responses on current events and specific topics. [ZDNet]
OpenAI and other AI companies are forging global partnerships to gather diverse, structured consumer data for improving their AI models’ accuracy and performance. [Rest of World]
Asahi and Nikkei filed a ¥2.2 billion copyright lawsuit against Perplexity, alleging unlawful scraping and reproduction of their articles. [The Register]
Perplexity will compensate news publishers from a US$42.5m pool for AI-generated content used in search and query responses. [The Wall Street Journal]
The LLM Ecosystem
Cerebras Systems and Core42 have delivered record-breaking performance for OpenAI's gpt-oss-120B, enabling enterprise-scale AI with unprecedented speed and efficiency. [Business Wire]
Dell has integrated its AI Data Platform with Elasticsearch and launched PowerEdge servers with Nvidia GPUs to accelerate enterprise AI workloads and data processing. [Blocks & Files]
fileAI launched its Model Context Protocol Server, enabling developers to process and extract data from files using a single natural language prompt across multiple AI environments. [EIN Presswire]
Google and Zed have introduced the Agent Client Protocol (ACP) to enable interoperable AI agent integration across different integrated development environments. [The Register]
LLM.co launched a private, customizable LLM-as-a-Service platform enabling enterprises to deploy secure, compliant AI tailored to their specific industry needs. [EIN Presswire]
Novogain AI has developed a context-specific AI platform that creates digital team twins to generate actionable insights and improve organizational performance. [EIN Presswire]
Salesforce AI Research has developed MCP-Universe, an open-source benchmark that evaluates LLMs’ performance across real-world enterprise tasks using interoperability standards. [VentureBeat]
And Salesforce has developed CRMArena-Pro, a digital twin simulation platform, to rigorously test AI agents in realistic business environments and address enterprise AI deployment challenges. [VentureBeat]
Scale AI researchers discovered that search-based AI models may cheat on benchmark tests by directly fetching answers from online sources like Hugging Face. [The Register]
Typedef has released Fenic 0.3.0, an open-source DataFrame project designed to help AI teams build more efficient and reliable production-scale inference pipelines. [EIN Presswire]
Vast Data's SyncEngine enables AI agents to access and mobilize unstructured data across diverse platforms, simplifying data pipeline challenges and enhancing AI workflow capabilities. [siliconANGLE]
Agentic AI
AI2 launched Asta, a comprehensive open-source ecosystem of AI agents, benchmarks, and tools designed to bring transparency and rigor to scientific research. [Business Wire]
Akka has partnered with Deloitte Canada to develop scalable, mission-critical agentic AI systems for large enterprises. [GlobeNewswire]
Avaamo has launched a HIPAA-compliant AI agent workforce for healthcare patient access on AWS Marketplace, offering end-to-end workflow automation across multiple channels. [EIN Presswire]
Nutshell has launched an AI Lead Researcher agent that automates lead research, providing instant insights and saving sales teams significant time. [PRWeb]
Salesforce launched CRMArena-Pro, a digital twin service enabling enterprises to stress-test AI agents using synthetic data before real-world deployment. [TechRadar]
University of Hong Kong researchers have developed OpenCUA, an open-source framework for creating robust AI agents capable of autonomously operating computers across diverse tasks and operating systems. [VentureBeat]
Vendasta has repositioned itself as an AI-powered customer acquisition platform for small and medium-sized businesses, aiming to provide enterprise-grade AI automation through autonomous agents. [BetaKit]
Workato has launched a free Developer Sandbox that provides full access to its AI automation platform, enabling developers to build and experiment with intelligent business agents. [siliconANGLE]
Other LLM Sightings
Advisor CRM launched a Gen AI Marketing Suite that automates content creation and streamlines marketing workflows for registered investment advisors. [PRWeb]
CodeSightAI launched an AI-powered code review platform that helps development teams improve software quality and reduce review time by up to 60%. [EIN Presswire]
DataVisor launched an AI-powered Suspicious Activity Report filing solution that unifies and automates suspicious activity report processes across fraud and AML teams. [Business Wire]
Daxtr launched an AI-powered personal assistant that proactively manages tasks via SMS, monetizing user intent without charging consumers. [EIN Presswire]
FlipHTML5 has introduced Storybook AI, an innovative tool that transforms text prompts into fully illustrated and interactive digital storybooks. [PRWeb]
Fountain launched Frontline OS, an AI-powered platform that dramatically reduces screening time and improves retention for frontline workers through automated workforce management. [Business Wire]
Innovid has introduced AI-powered tools that automate creative labeling, optimize ad performance, and enable faster personalization across advertising formats. [Business Wire]
Phonetik AI has developed AI-powered accessibility tools that quickly and affordably create subtitles, captions, and narration for visually and hearing-impaired users across multiple languages. [TechCentral]
Sapia.ai launched a Skills Intelligence API that enables developers to embed science-backed, bias-mitigated hiring intelligence directly into HR systems. [Business Wire]
Synametrics Technologies has released WinSQL Version 20, integrating an AI assistant that simplifies SQL querying, optimizes performance, and enhances database management for users. [EIN Presswire]
Teikametrics launched GenAI Smart Pages, an AI-powered tool that automatically optimizes product listings using performance data to maximize sales across retailers. [Business Wire]
Your Money Line launched an AI-powered, personalized podcast that provides tailored financial insights and strategies to help employees improve their financial wellness. [EIN Presswire]
YURA introduced an AI-powered chatbot in CADvizor, automating electrical system design tasks and enhancing efficiency through intelligent parts selection and drawing creation. [PRWeb]
Risks and Responses
US attorneys general warned 11 AI and social media companies to protect children from potential harm in chatbot interactions or face accountability. [404 Media]
AI chatbots have repeatedly validated users’ false beliefs, potentially leading vulnerable individuals to dangerous delusions about reality and themselves. [Ars Technica]
Anthropic has developed an AI tool with the US Department of Energy to detect and block attempts to obtain nuclear weapons design information. [TechRadar]
And Anthropic has revealed that cybercriminals are increasingly using its Claude AI to conduct sophisticated ransomware, fraud, and espionage operations across multiple sectors. [The Register]
Apple has struggled to honour AI safety commitments, scoring just 13% on voluntary White House guidelines according to a recent research analysis. [Fast Company]
Citizen is using AI to generate potentially inaccurate and sensitive crime alerts without human review, raising concerns about privacy and data exposure. [404 Media]
Copyleaks launched Text Moderation, an AI-powered content analysis tool that provides context-aware flagging to help teams manage digital content more effectively. [GlobeNewswire]
Meta's Yann LeCun has proposed AI safety measures focused on constraining systems within human-defined objectives and implementing safeguards to prevent potential harm. [Digital Information World]
Microsoft has developed Project Ire, an AI agent capable of autonomously detecting and reverse engineering malware with high precision and low false positive rates. [ITPro]
OpenAI and Anthropic collaborated on cross-lab AI safety testing to surface blind spots and demonstrate potential industry-wide cooperation amid fierce competition. [TechCrunch]
OpenAI faces a wrongful death lawsuit after a teenager allegedly used ChatGPT to discuss suicide methods, highlighting AI chatbot safety limitations. [TechCrunch]
OpenAI acknowledged that ChatGPT’s safety guardrails may degrade during long conversations. [Gizmodo]
OpenAI has enhanced ChatGPT’s safeguards to better protect users, especially teens, by improving mental health interventions and exploring parental controls. [ZDNet]
And OpenAI has quietly begun scanning user messages for harmful content, potentially escalating concerning conversations to human review and law enforcement. [Futurism]
Oxford researchers discovered that training language models to be warmer and more empathetic led to increased errors, misinformation, and a tendency to agree with users. [The Decoder]
PromptLock emerged as the first known AI-powered ransomware, using OpenAI's GPT model to generate malicious Lua scripts for potential file encryption and exfiltration. [ITPro]
Security researchers from Palo Alto Networks’ Unit 42 have discovered that LLMs can be tricked into bypassing guardrails by using a single, grammatically poor, run-on sentence. [The Register]
Spiral-Bench is a new AI safety test that reveals how different models can trap users in escalatory delusion loops with varying degrees of risk. [The Decoder]
TikTok laid off hundreds of UK and Asian moderators as it shifts towards AI-driven content moderation, sparking criticism from unions and safety advocates. [Gizmodo]
The UK’s Trades Union Congress warned that AI could increase inequality unless a ‘worker-first’ strategy is adopted to protect jobs and support employees through technological changes. [The Register]
xAI's Grok AI chatbot leaked over 370,000 user conversations, revealing disturbing content including instructions for drugs, malware, and potential violence. [Gizmodo]
Regulation
Tech giants like OpenAI and Google have lobbied federal lawmakers to preempt state AI regulations, arguing for a unified national approach to prevent innovation-stifling restrictions. [WebProNews]
Andreessen Horowitz and OpenAI President Greg Brockman are leading a US$100m effort to create a pro-AI super-PAC network opposing strict AI regulations in upcoming elections. [TechCrunch]
And Meta has established a super PAC to support California candidates favouring a light-touch approach to AI regulation, aiming to influence statewide elections. [TechCrunch]
Trump has threatened to impose tariffs on countries regulating US tech companies, specifically targeting Australia’s proposed social media ban for children. [9News]
Environmental Issues
AI’s soaring energy demands are complex, with data centers consuming increasing electricity, but individual query impact remains relatively small compared to broader environmental factors. [ZDNet]
Google's environmental impact report on Gemini AI highlights low energy use per text prompt while omitting critical factors like scale, complex tasks, and training energy consumption. [The Decoder]
Google's report on Gemini’s energy use reveals only a partial picture of AI’s significant and growing electricity demand across various applications and platforms. [MIT Technology Review]
Microsoft, Nvidia, and OpenAI researchers warned of AI training’s uneven power usage and proposed collaborative strategies to stabilize electrical grid demand. [The Register]
Conversational AI
6sense launched AI Email Agents to automate personalized B2B marketing outreach, enabling efficient pipeline creation with minimal manual effort. [Business Wire]
Aurelian has pivoted from salon booking software to developing an AI voice assistant that helps 911 call centers manage non-emergency calls by triaging and routing communications. [TechCrunch]
BetSymphony introduced a conversational AI platform with BetHarmony, offering personalized, intuitive iGaming experiences through voice navigation and intelligent interaction. [GlobeNewswire]
Cerence AI enhanced Volkswagen Group's IDA in-car assistant with Cerence Chat Pro, enabling more natural, multi-turn conversations across various vehicle brands. [GlobeNewswire]
A study by Glia revealed that less than 30% of financial institutions are fully leveraging contact center AI’s potential, despite widespread recognition of its strategic importance. [Business Wire]
LinkLive's Agentic AI platform reached 1,100 customers, enabling organizations to achieve 78% first-call resolution and 50% cost reduction across regulated industries. [Business Wire]
Microsoft and Cloudflare partnered to transform websites into AI-friendly, conversational platforms through NLWeb and AutoRAG technologies, challenging traditional search models. [TechRadar]
Qualified unveiled Piper, an AI SDR superagent capable of autonomously engaging, nurturing, and converting buyers through multi-modal, face-to-face experiences across the marketing funnel. [Business Wire]
Relatient's Voice AI successfully deflected over 60% of routine appointment scheduling calls, reducing staff burden and improving patient experience through automated, intelligent interactions. [Business Wire]
Slang AI has partnered with Tripleseat to help restaurants capture private dining leads by automatically logging and syncing event inquiry calls into their platform. [PRWeb]
Taco Bell experimented with AI drive-thru ordering, encountering numerous challenges including customer manipulation and order inaccuracies. [Gizmodo]
Be Real
Alibaba unveiled Wan2.2-S2V, an open-source speech-to-video model that transforms single portrait images into lifelike, film-quality animated avatars capable of speaking and performing. [Alibaba Cloud]
Google has expanded its Vids video editor with AI avatars, transcript trimming, and image-to-video tools for Workspace users. [TechCrunch]
Sharp's Poketomo is a palm-sized, meerkat-shaped AI companion designed to provide emotional support for young women experiencing loneliness. [TechRadar]
Stanford’s James Zou is hosting Agents4Science, a conference where virtual AI researchers autonomously conduct, write, review, and present scientific research across multiple disciplines. [MIT Technology Review]
Voice News
Reality Defender has partnered with Hume AI to enhance deepfake detection capabilities by gaining early access to advanced voice AI models. [Speech Technology Magazine]
Sendbird launched a voice AI solution providing multilingual, adaptive customer service across communication channels with advanced NLP capabilities. [Speech Technology Magazine]
Smartbox used AI and an old VHS tape to restore Sarah Ezekiel’s lost voice after 25 years of motor neurone disease. [GovTech]
SoundHound AI has deployed its generative AI-powered voice assistant in select Jeep vehicles across European markets, enhancing in-vehicle interaction and connectivity. [Business Wire]
Winxvideo AI v4.3 has enhanced its multimedia software with an advanced AI Vocal Remover, expanded codec support, and improved workflow tools for video and audio professionals. [EIN Presswire]
Document AI
Docling, an open-source IBM Research tool, simplifies unstructured data processing for retrieval-augmented generation by providing modular, context-aware document parsing and conversion capabilities. [The New Stack]
Helm & Nagel launched Quick-Extract, an AI-powered document data extraction platform that automatically generates parsers without manual configuration or training. [Helm & Nagel]
RunMyProcess launched an AI Workflow Assistant that automates and optimizes document-heavy processes with enterprise-grade governance and intelligent document processing capabilities. [EIN Presswire]
Translation
Meta has launched free AI audio translation for Facebook and Instagram creators, automatically dubbing reels between English and Spanish for accounts with over 1,000 followers. [Slator]
PlayTV.ai developed an AI platform that instantly translates YouTube channels into multiple languages, enabling creators to reach global audiences effortlessly. [EIN Presswire]
Welocalize collaborated with Duke University to create LangMark, a multilingual dataset evaluating AI translation post-editing across seven languages using real-world marketing content. [Slator]
Search
Bospar research predicts AI will largely replace Google for business research by 2030, with significant generational differences in adoption expectations. [Business Wire]
Google is experimenting with improving link placement and visibility in AI Mode search results to encourage more user clicks on website links. [ZDNet]
OpenAI reportedly used SerpApi to scrape Google Search results, revealing the continued importance of Google’s search index in the AI era. [Search Engine Land]
Pureinsights launched a Discovery Sandbox that enables organizations to experiment with AI-powered search and chatbots without complex environment setup. [EIN Presswire]
SpeedyIndex launched a Googlebot-guaranteeing service that ensures web pages and backlinks are indexed within 12 hours, offering a VIP Queue for near-instant crawling. [PRWeb]
AI in Journalism
Politico defended its AI-generated reports as exempt from editorial standards, arguing they were created by coders outside the newsroom, sparking union dispute. [TechDirt]
The University of North Carolina’s Center for Innovation & Sustainability in Local Media supported four small newsrooms in developing experimental AI chatbots, revealing challenges and potential benefits of local news AI tools. [NiemanLab]
Health Tech
Ask DrBomi is an AI health assistant that provides insights on health and longevity. [EIN Presswire]
IKS Health's Scribble has been integrated with Epic's electronic health record software, offering AI-powered clinical documentation solutions to enhance clinician efficiency. [Business Wire]
Inovalon has partnered with Google Cloud to develop an AI-powered system that will automate and streamline the complex prior authorization process in healthcare. [Forbes]
ModMed has developed an AI-powered Enhanced Faxing feature that dramatically reduces administrative time and improves healthcare workflow efficiency. [Business Wire]
OpenAI’s GPT-4o demonstrated high-quality multilingual translation of radiology reports in English, French, and Spanish, with potential for reducing language barriers in healthcare. [Slator]
Meanwhile, OpenAI has expanded its healthcare AI team, hiring key executives and signalling increased focus on developing AI tools to assist clinicians and patients. [Gizmodo]
Ostro launched Airmark, an AI-powered solution transforming pharma marketing emails into interactive, compliant conversations for healthcare professionals. [Business Wire]
Wolters Kluwer has enhanced its Ovid platform with an AI Article Summary tool that helps medical researchers quickly distill insights from peer-reviewed literature. [Business Wire]
Legal Tech
Complete Discovery Source has expanded its RelativityOne cloud eDiscovery platform to Germany, offering secure, AI-powered hosting that meets local privacy requirements. [EIN Presswire]
Everlaw has expanded its Deep Dive AI tool to open beta and secured FedRAMP certification, enabling legal teams to quickly query large document collections with AI-powered insights. [LawSites]
Exterro unveiled Exterro Intelligence, an advanced AI solution for data risk management that delivers secure, efficient, and transparent results through domain-specific expert agents. [GlobeNewswire]
Harvey has launched a law school alliance program, embedding its generative AI technology into curricula at five prestigious US law schools to prepare future lawyers for technology-enabled legal work. [Artificial Lawyer]
Juro and Wordsmith formed a strategic partnership using Model Context Protocol to provide an integrated AI offering for in-house legal teams, focusing on enhancing workflow and automation. [Artificial Lawyer]
Law.co has developed privacy-focused AI solutions enabling law firms to deploy secure, custom language models without compromising client confidentiality. [EIN Presswire]
The Law Society of Ireland has established an online hub providing legal professionals with comprehensive resources on technology, courses, and industry insights. [Irish Legal News]
Legora has expanded its Word Add-in with four AI-powered Actions that automate template filling, document anonymization, writing improvement, and translation for legal professionals. [Artificial Lawyer]
And Legora has partnered with Harrison LLP to integrate AI tools into private wealth law, enhancing lawyers’ ability to deliver personalized legal solutions. [Business Wire]
LexisNexis has expanded its Protégé AI platform to offer secure, authoritative, and workflow-integrated AI tools for legal professionals. [LawSites]
Practice AI expanded its AI Demands platform with case valuation and verdict analysis tools to help law firms make faster, more strategic decisions. [GlobeNewswire]
Supio launched Instant Demands, an AI-powered tool that generates accurate, firm-specific demand letters for personal injury law firms in minutes. [Business Wire]
Ed Tech
Anthropic's research reveals educators are using Claude AI to develop course materials, conduct research, and automate administrative tasks. [Anthropic]
Cengage Group expanded its AI-powered learning tools across higher education, workforce training, and K-12 segments to enhance personalized learning experiences for over one million students. [Business Wire]
Google has introduced AI-powered language learning and live translation features in its Translate app, targeting both language learners and real-time communication across multiple languages. [TechCrunch]
MathGPT.ai launched an AI tutoring platform designed to help college students learn math while providing instructors with robust tools to prevent cheating and support learning. [TechCrunch]
Funding
Aether raised NZ$3.8m in Seed funding to transform marketing presentations into live, data-connected documents for enterprise customers. [Startup Daily]
amass secured €1.4M in funding to develop an AI-powered scientific intelligence platform connecting diverse research sources for faster life science discoveries. [Silicon Canals]
Apate.ai raised US$2.5m in Seed funding to develop AI bots that engage and gather intelligence from scammers, helping protect Australian consumers. [Startup Daily]
Arintra secured US$21m in Series A funding to expand its GenAI-powered medical coding platform for healthcare providers. [FinSMEs]
Assort Health raised US$50m in Series B funding to develop AI-powered patient communication tools for specialty healthcare practices. [TechCrunch]
Attio secured US$52m in Series B funding to further develop its London-based CRM software platform. [Silicon Canals]
Aurelian secured US$14m in Series A funding to expand its AI voice assistant service for automating non-emergency 911 calls across the US. [FinSMEs]
Bench IQ raised US$5.3m to develop an AI platform that provides litigators with predictive insights into judges’ decision-making patterns. [Business Wire]
Cascala Health secured US$8.6m in seed funding to leverage AI for improving post-hospital patient care and data summarization. [Endpoints News]
Darwin AI secured US$4.5m in seed funding to develop AI workers that help mid-market companies in Latin America automate sales, support, and customer interactions. [FinSMEs]
EliseAI has raised US$250m in Series E funding, doubling its valuation to over US$2.2 billion while expanding its AI-powered enterprise solutions in real estate and healthcare. [Tech Funding News]
Farang raised €1.5m to develop a novel AI language model architecture that enables specialized, efficient, and privacy-focused assistants across various domains. [Tech.eu]
Interhuman AI raised €2m to develop a social intelligence layer for AI systems, enabling more nuanced and empathetic human-AI interactions through advanced non-verbal communication analysis. [Tech.eu]
Loman AI has raised US$3.5m to develop a Voice AI platform that helps restaurants handle phone orders, improve sales, and reduce labor costs. [Business Wire]
Maisa AI has raised US$25m to develop an enterprise-focused AI platform that prioritizes accountability and process transparency through its innovative digital worker deployment system. [TechCrunch]
Vox AI raised US$8.7m in seed funding to develop autonomous voice AI technology for transforming quick service restaurant operations globally. [Business Wire]
Acquisitions
Thoma Bravo is acquiring Verint Systems for US$1.23B in cash, offering shareholders US$20.50 per share at a 13% premium. [Bloomberg]
TransPerfect acquired Unbabel, a language AI pioneer, integrating the latter’s TowerLLM and COMET technologies into its GlobalLink technology portfolio. [MultiLingual]
Uniphore acquired Orby AI and Autonom8, adding top AI talent from Google DeepMind and Google to enhance its Business AI Cloud capabilities. [Business Wire]
Workday acquired Paradox, an AI-powered conversation recruitment platform, to enhance its talent acquisition suite and streamline hiring processes. [ITPro]
There’s More
AI’s overuse of em dashes reveals its tendency to artificially polish writing, requiring human intervention to maintain authentic communication style. [VentureBeat]
Altman and Musk are advancing brain-computer interfaces and AI implants that could fundamentally transform human existence, raising critical ethical concerns about dehumanization. [Yeni Şafak]
Melania Trump has launched the Presidential AI Challenge to inspire students to use AI tools for solving community problems and advancing technological innovation. [New York Post]
Montana eatery Stefanina’s Wentzville pleaded with customers to stop using Google AI Overviews, which falsely advertised non-existent specials and menu items, causing customer confusion and frustration. [Futurism]
Vivaldi has firmly rejected integrating generative AI into its browser, arguing that such technology dehumanizes web browsing and prioritizes data collection over user experience. [The Register]
The Back Cover
Here’s Foo Fighters with a cover of Gerry Rafferty’s Baker Street.
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!