This Week in NLP #333
Keep up with what happened in NLP in the week ending Friday 4th April 2025.
Above the Fold
Overwhelmed? Here’s our pick for five things to know about this week.
Amazon has launched Nova Act, an experimental developer kit for building AI agents that can autonomously navigate the web, powered by its proprietary Nova language model. [VentureBeat]
And Amazon's Alexa+ has launched with limited functionality, offering only basic features like Uber ordering and email drafting, while many promised capabilities remain delayed for months. [Engadget]
Cognition AI has launched Devin 2.0, a significantly cheaper and more feature-rich version of its autonomous coding platform, featuring parallel agents and improved codebase exploration capabilities. [VentureBeat]
OpenAI has secured a record-breaking US$40B funding round led by Softbank, valuing the ChatGPT maker at US$300B and surpassing all previous private tech fundraising. [CNBC]
xAI has acquired X in an US$80B all-stock merger, combining Musk’s AI venture with the social media platform he purchased in 2022. [The Verge]
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
AWS has offered customers a 25% discount to switch from Nvidia's GPUs to its own Trainium chips. [TechRadar]
Gartner has predicted an upcoming consolidation in the LLM provider market, with many companies facing extinction due to high capital requirements and competitive pressures. [The Register]
Google has made its advanced Gemini 2.5 Pro Experimental AI model freely available to all users, offering enhanced reasoning capabilities that top industry benchmarks. [TechRadar]
Manus AI has launched two subscription tiers starting at $39 monthly, alongside releasing an iOS app and upgrading to Claude 3.7 Sonnet. [TechCrunch]
Meta's AI research VP Joelle Pineau has announced her departure from the company effective May 2024. [TechCrunch]
Microsoft has denied analyst claims that its scaled-back datacenter expansion plans reflect a deteriorating relationship with OpenAI, reaffirming their strong partnership. [Computer Weekly]
OpenAI has taken over Microsoft's US$12B CoreWeave contract after Microsoft withdrew, highlighting the closed-loop nature of the AI economy. [TechRadar]
And OpenAI has begun planning its first data center to reduce its dependence on Microsoft's cloud infrastructure. [The Information]
OpenAI has announced plans for a new open-weight language model and is gathering developer feedback through applications and upcoming events in multiple global regions. [TechRepublic]
OpenAI’s ChatGPT attracted one million new users within an hour after launching image generation features. [Engadget]
And OpenAI has generated over 700 million images through its new ChatGPT feature since March 25. [TechCrunch]
OpenAI has grown its ChatGPT paid subscriber base by 30% to 20m users, driving monthly revenue to US$415m amid ambitious expansion plans. [Yahoo Finance]
OpenAI's o3 model has faced revised cost estimates from the Arc Prize Foundation, increasing from $3,000 to $30,000 per ARC-AGI task. [TechCrunch]
OpenAI has been given until year-end to become a for-profit company in order to secure up to US$40B in Softbank-led funding. [DealStreetAsia]
xAI has encountered trademark challenges over its chatbot name ‘Grok’ from multiple companies, including Groq, Grokstream, and Bizly, which claims prior trademark rights from 2021. [Wired]
And xAI's Grok has criticized its creator Elon Musk as a misinformation spreader, defying attempts to control its responses. [eWeek]
Zhipu AI has launched AutoGLM Rumination, a free AI agent powered by proprietary language models that have demonstrated competitive performance against leading AI benchmarks. [eWeek]
Sovereign AI
DeepSeek's development of a cost-effective Chinese AI model has spurred US policymakers and tech companies to prioritize competitive innovation over safety concerns in maintaining AI dominance. [The Indian Express]
The Digital Europe Programme has announced its 2025-2027 work programme focusing on AI deployment, cloud computing, cybersecurity, and digital skills development across EU-affiliated nations. [European Commission]
The European Union has committed €1.3B to accelerate AI adoption through initiatives including healthcare training, digital infrastructure, and innovation hubs across member states. [TechRepublic]
IBM has opened a new Cloud Multizone Region in Montreal and expanded its watsonx AI portfolio to Toronto, strengthening data sovereignty capabilities across Canada. [Yahoo Finance]
The UK government’s Public Accounts Committee has revealed that outdated IT systems, poor data quality, and skills shortages are hampering AI adoption across UK government departments. [Total Telecom]
US Office of Science and Technology Policy Director Michael Kratsios has received a directive from Trump to restore American technological supremacy. [FedScoop]
Feature Creeps
Apple Intelligence has expanded its AI features with Priority Notifications, new languages, EU availability, Vision Pro support, and additional tools for Mac and Image Playground. [TechCrunch]
Google has launched new AI-powered travel features across Search, Maps, and Gemini to enhance vacation planning and compete with emerging AI platforms. [Yahoo Finance]
Google’s NotebookLM has a new ‘Discover sources’ feature that automatically finds and curates relevant web content based on user-specified topics. [Google]
Google's Project Astra AI upgrade has expanded beyond Pixel and Samsung devices to all Android phones with Gemini Advanced, requiring a monthly subscription. [TechRadar]
OpenAI has launched internal knowledge source integration for ChatGPT Team users, allowing them to connect company databases for more contextual and accurate responses. [VentureBeat]
OpenAI has launched ChatGPT Projects, allowing paying users to organize conversations, files, and custom instructions into separate workspaces for different purposes. [Wired]
OpenAI has adopted Anthropic's Model Context Protocol, enabling AI applications to connect with data sources through a universal standard for more efficient information access. [TechRepublic]
OpenAI has expanded its viral Studio Ghibli-style image generator to free users, offering three daily generations following initial delays due to overwhelming demand. [The Verge]
And OpenAI has moved ChatGPT’s Deep Research feature from its Pro tier to Plus and is now planning to release it on the free tier, though with likely usage restrictions. [TechRadar]
Hype Bubble?
AAAI’s 2025 Presidential Panel has revealed that most AI researchers believe current approaches to achieving AGI are insufficient and overhyped. [Gizmodo]
Research by Asana reveals that 67% of companies have failed to scale AI beyond leadership, with tools remaining largely confined to individual rather than collaborative use. [TechRadar]
China’s AI datacenter boom has stalled, with up to 80% of newly built facilities sitting empty after speculative investments and changing technical demands undermined their viability. [AIwire]
This piece argues that CoreWeave's disappointing IPO has stemmed from company-specific issues like high debt and customer concentration, rather than reflecting broader problems in AI adoption. [Yahoo Finance]
McKinsey's latest survey has found that large companies are leading organizational changes to generate value from generative AI, with CEOs increasingly overseeing AI governance and workflow redesigns. [McKinsey]
Microsoft co-founder Bill Gates has predicted that AI will replace many jobs within a decade. [eWeek]
Hardware
Ambarella has demonstrated its next-generation Edge GenAI technology, showcasing reasoning models on its CVflow Edge AI SoCs. [EIN Presswire]
H3C has warned customers about depleting H20 chip stocks and delayed deliveries until mid-April, amid strong demand for Nvidia's China-approved AI processors. [Yahoo Finance]
MangoBoost has achieved record-breaking performance with its LLMBoost software on AMD MI300X GPUs, surpassing Nvidia H100 results while offering significant cost savings. [Business Wire]
MediaTek has partnered with Google to develop next-generation tensor processing units, reducing Google’s dependence on Broadcom for AI hardware development. [TechRadar]
Nvidia has received US$16B in orders for H20 server chips from Chinese tech giants ByteDance, Alibaba, and Tencent amid tightening US export restrictions. [Yahoo Finance]
Nvidia's GPUs dominated MLCommons’ latest MLPerf benchmarks, outperforming AMD and Google in most generative AI tests, with AMD scoring wins in two Llama 2 categories. [ZDNet]
And Nvidia has redefined GPU counting from modules to individual dies, potentially doubling AI Enterprise licensing costs for customers using its latest HGX systems. [The Register]
Consumer AI
Apple has integrated Matter-compatible robot vacuum control into iOS 18.4, while Samsung has launched its Bespoke AI appliance lineup featuring energy-efficient smart devices. [eWeek]
MediaTek has unveiled the Kompanio Ultra 910 chip for Chromebooks, featuring an 8-core CPU and 50 TOPS NPU, to compete with AI-enabled Windows laptops. [The Register]
Meta is developing Hypernova, a $1,000+ smart glasses model with a built-in display, hand gesture controls, and an AI chatbot, targeting a late 2024 release. [The Verge]
Microsoft has rolled out new AI features for Copilot+ PCs with AMD and Intel processors, including live captions and image generation tools for Paint and Photos apps. [TechRadar]
It’s Only a Model
Alibaba’s QVQ-Max has launched as a visual reasoning AI model that analyzes images and videos to solve problems ranging from mathematics to artistic creation. [Qwen]
And Alibaba announced plans to release Qwen 3, its latest AI model, amid increased competition from both Chinese and US tech companies in the AI market. [Yahoo Finance]
Runway has launched Gen-4, a high-fidelity AI video generator that maintains consistent characters and environments across scenes. [TechCrunch]
Whose Data?
Google has proposed copyright reforms to the UK government that it says balance AI innovation with creators’ rights. [Google]
Meta has faced demands from British authors to appear before Parliament after allegedly using LibGen’s pirated book collection to train its Llama 3 AI model. [Engadget]
OpenAI has been accused by an AI watchdog organization of training its GPT-4o model on unlicensed, paywalled O’Reilly Media books. [TechCrunch]
Thomson Reuters has won a partial judgment against Ross Intelligence over unauthorized use of Westlaw content, while simultaneously investing US$200m annually in its own AI development and advocating for fair compensation frameworks. [The Register]
The Blair Institute has backed tech companies over artists in a report supporting data mining exceptions for AI training, despite opposition from content creators and rights holders. [The Register]
Vana has launched a decentralized platform enabling users to pool and monetize their personal data for AI training while maintaining ownership of resulting models. [MIT News]
The Wikimedia Foundation has reported a 50% surge in bandwidth consumption since January due to AI model training bots scraping multimedia content from Wikimedia Commons. [TechCrunch]
The LLM Ecosystem
Anthropic has developed a technique to observe LLMs’ internal processes, revealing surprising insights about how they handle tasks like math and language translation. [MIT Technology Review]
ClearCompany has launched AI Notetaker, an interview intelligence tool that records and analyzes candidate interviews while allowing recruiters to focus on personal connections. [PRWeb]
env0 has launched Cloud Analyst, an AI-powered infrastructure intelligence solution that provides instant visibility and insights across enterprise cloud environments. [EIN Presswire]
Exostellar has launched Software Defined GPU™, a multi-vendor GPU slicing platform that optimizes resource utilization and reduces costs through dynamic allocation. [Business Wire]
Fivetran has expanded its Microsoft Fabric integration by adding over 700 pre-built connectors, enabling automated data movement for AI-ready data lakes and analytics. [Business Wire]
Fluidstack has partnered with Macquarie Group to finance GPU supply for European AI labs through a strategic deal enabling flexible, contract-free compute access. [Business Wire]
Hugging Face has launched Yourbench, an open-source tool enabling organizations to create custom benchmarks for testing AI models against their internal data. [VentureBeat]
Informatica has launched new AI-powered cloud integration and data management features, including CLAIRE Copilot and GPT integration, to enhance enterprise data accessibility and productivity. [Business Wire]
John Snow Labs has launched Generative AI Lab 7.0, enabling domain experts to evaluate and improve custom-built LLMs without coding expertise. [GlobeNewswire]
Keysight Technologies has launched KAI Data Center Builder, a software suite that emulates AI workloads to validate and optimize network architecture performance. [Business Wire]
LLMs.txt has been proposed as a new website content standard, offering AI models streamlined access to content while giving site owners more control over data usage. [Search Engine Land]
MLCommons has announced MLPerf Inference v5.0 benchmark results, highlighting dramatic performance improvements in generative AI and introducing four new benchmark tests. [Business Wire]
Nebul has deployed Mirantis’ open-source k0rdent platform to deliver on-demand AI inference workloads while maintaining European privacy and sovereignty standards. [Business Wire]
Nvidia has open-sourced its KAI Scheduler, a Kubernetes-native GPU scheduling solution designed to manage AI workloads and optimize resource allocation across computing clusters. [VentureBeat]
OpenAI has launched a free educational platform called OpenAI Academy, offering AI courses and resources to students ranging from beginners to advanced learners. [Tech Startups]
Parasail has launched the first AI Deployment Network, providing companies with access to the largest supply of on-demand GPUs at significantly reduced costs. [Business Wire]
Prophecy has launched version 4.0 of its data integration platform, adding self-service data preparation capabilities for Databricks SQL users. [BigDATAwire]
Scale AI has launched an automated platform that tests AI models across thousands of benchmarks, identifies weaknesses, and recommends additional training data to enhance performance. [Wired]
SGNL has announced support for Model Context Protocol, providing real-time security controls for AI agents accessing enterprise systems through identity-aware, policy-driven decisions. [Business Wire]
Soochow University researchers have developed Chain-of-Tools (CoTools), a framework enabling LLMs to efficiently use external tools without compromising their core capabilities. [VentureBeat]
Traefik Labs has launched new AI Gateway platform enhancements and developer tools to improve enterprise AI deployment security, scalability, and API management. [Business Wire]
Zencoder has launched AI coding agents that integrate with existing development environments, offering autonomous code generation and testing capabilities. [VentureBeat]
Agentic AI
Amazon has launched an AI-powered ‘Buy for Me’ feature that enables users to purchase items from third-party websites directly through its mobile app. [The Verge]
Anthropic's Model Context Protocol has gained early industry adoption as a standardized way to connect AI agents with data sources and tools since its fall launch. [BigDATAwire]
Augment Code has launched its ‘Augment Agent’ technology, focusing on helping developers navigate large codebases rather than simple code generation. [VentureBeat]
Emergence AI has launched a platform that creates AI agents through text prompts, using recursive intelligence to autonomously generate and orchestrate enterprise workflow solutions. [VentureBeat]
Kubiya has unveiled what it says is the first Enterprise AI Stack for Agents, enabling organizations to build and manage secure AI agents faster. [Business Wire]
NICE Actimize has launched its X-Sight ActOne platform with Agentic AI capabilities to transform financial crime investigations through automated workflows and ML. [Business Wire]
Orases has launched AI agent development services, expanding its portfolio to provide custom intelligent automation solutions for mid-market enterprises. [PRWeb]
Zhipu AI has launched a free AI agent called AutoGLM Rumination, joining other Chinese companies in releasing AI products that compete with US rivals. [Reuters]
Other LLM Sightings
12New.ai has launched StrongSignal.ai, an AI-powered platform that aggregates and analyzes stock market news to help investors make data-driven decisions more efficiently. [EIN Presswire]
Actable has launched Intelligence Factory, a framework that transforms customer data into actionable insights within clients’ Google Cloud environments. [PRWeb]
Adobe has launched Generative Extend and Media Intelligence in Premiere Pro, introducing AI-powered tools for extending video clips and searching footage more efficiently. [Business Wire]
Advisor360° has launched Parrot AI, a secure generative AI assistant that helps financial advisors manage client meetings and automate workflow documentation. [Business Wire]
<b>Apex Fintech Solutions</b> has launched Ask Ascend, an AI-powered assistant built on Google Cloud's Vertex AI to accelerate client technical integrations and support. [Business Wire]
AvidXchange has launched new AI agents and enhancements for invoice automation, including approval assistance, PO matching, and improved capture capabilities for mid-market businesses. [GlobeNewswire]
DeskSense has launched a lifetime access deal for its AI-powered productivity assistant, offering email drafting, spreadsheet formatting, and content creation tools for $38. [TechRepublic]
Finarb Analytics Consulting has launched DataXpert, a generative AI platform that connects with various data sources to provide natural language-driven enterprise analytics and insights. [Business Wire]
Infocepts has implemented a generative AI solution helping retailers create optimized product descriptions that boost e-commerce conversions while maintaining brand consistency. [EIN Presswire]
Iterable has unveiled Nova, an AI agent system designed to transform marketing by providing real-time insights, automated campaign optimization, and personalized customer engagement. [Business Wire]
Job-Bolt has launched an AI-powered video interview platform that accelerates hiring by 75% while reducing screening costs through automated candidate evaluation. [EIN Presswire]
micro1 has launched Zara, a multilingual AI recruitment agent that automates talent sourcing, interviewing, and onboarding while reducing hiring costs by 80%. [EIN Presswire]
Quickbase has launched AI Smart Builder, transforming complex spreadsheets into functional apps with automated workflows and integrations in seconds. [Business Wire]
And Quickbase has enhanced its FastField platform with AI-powered capabilities including OCR, automated table generation, and workflow improvements to streamline field service management. [Business Wire]
Octus has launched CreditAI Vault, an AI-powered research tool that helps buy-side professionals analyze private credit data across permissioned data rooms. [Business Wire]
ParaScript has partnered with Endurance Italia to integrate AI-powered fraud prevention and check verification solutions for Italian financial institutions. [Business Wire]
Radiant Point has launched Sky for Customer Success, an AI-powered copilot that integrates with CRM systems to help maximize post-sale revenue through guided workflows. [EIN Presswire]
Ready Tensor has launched TensorCheck, an AI-powered assessment tool that evaluates AI-focused publications using over 100 expert-defined criteria for quality improvement. [EIN Presswire]
Sapia.ai has launched Chat Interview Pro to assess 25 core competencies through AI-powered interviews, following analysis of 47,000 job descriptions across industries. [Business Wire]
Sourcetable has launched an AI-driven spreadsheet that uses natural language commands and voice input. [Legal IT Insider]
Thrive HR Consulting has launched AI 2.0, combining AI-powered tools with fractional HR services. [EIN Presswire]
UpVibe has launched an AI-powered accountability app featuring personalized focus calls, community support, and reward pledges to help users achieve their goals. [EIN Presswire]
Vizpert has launched an AI-powered immigration platform, combining 12 years of legal expertise with insights from over one million cases to streamline immigration processes. [EIN Presswire]
Voxle Talent has launched an AI-powered recruitment platform that streamlines hiring processes and reduces bias through structured evaluation frameworks. [EIN Presswire]
Risks and Responses
AgentSpec, developed by Singapore Management University researchers, has introduced a domain-specific framework that enforces structured rules to control AI agents’ behavior and prevent unsafe actions. [VentureBeat]
Applause's research has revealed that only one-third of organizations use red team testing in AI development, despite widespread issues with application flaws. [ITPro]
Barracuda has warned organizations about ‘gray bots’ - AI web scrapers that, while not malicious, can overwhelm websites with millions of daily requests. [ITPro]
Carnegie Mellon researchers have discovered that excessive pre-training of LLMs can make them harder to fine-tune, dubbing this effect ‘Catastrophic Overtraining’. [VentureBeat]
Google's Gemini has become vulnerable to algorithmically generated prompt injections through its fine-tuning feature, marking the first successful automated attack on a closed-weights model. [Ars Technica]
Google DeepMind has published a 145-page paper predicting AGI’s arrival by 2030 and outlining safety measures, while drawing criticism from AI experts regarding its premises. [TechCrunch]
And Google DeepMind has tightened research-sharing restrictions, prompting Iris.ai's CEO to warn that reduced openness would hamper scientific progress across the AI industry. [The Next Web]
OpenAI has relaxed its content moderation policies, allowing ChatGPT’s new image generator to create pictures of public figures and sensitive content while maintaining some restrictions against harmful usage. [TechCrunch]
OpenAI’s ChatGPT has developed the ability to generate convincing fake receipts, with the company claiming potential benefits despite fraud concerns. [TechRadar]
And OpenAI has increased its bug bounty rewards to $100,000 and expanded its cybersecurity programs, including partnerships with Bugcrowd and SpecterOps for vulnerability testing. [eWeek]
Netskope has reported a 30-fold increase in enterprise data sharing with AI apps over 12 months, reaching 7.7GB monthly and raising security concerns. [TechRadar]
New Jersey has criminalized the creation and sharing of AI-generated deepfake media, with violators facing up to five years in prison and potential civil lawsuits. [Associated Press]
Conversational AI
8x8 has launched AI Orchestrator, enabling organizations to create seamless workflows between multiple AI bots from different vendors for improved customer experiences. [Business Wire]
And 8x8 has launched JourneyIQ, an AI-powered solution that tracks and optimizes customer interactions across all channels and departments within organizations. [Business Wire]
Bliro is a conversation intelligence platform that transforms business communications into actionable data across multiple channels. [Silicon Canals]
CallRail has launched Voice Assist, an AI-powered assistant that transforms missed calls into revenue by handling leads 24/7 through natural conversations and automated follow-ups. [Business Wire]
CloudXperte has launched VehicleTracking.ai, an AI-powered chatbot offering 24/7 multilingual customer support and data analytics across multiple industries. [EIN Presswire]
Cyara has unveiled a next-generation CX assurance platform featuring innovations for voice, digital, and AI channels across more than 100 countries. [Business Wire]
Glia has launched GVA Learning 360, an AI virtual assistant that learns from human agent conversations to generate customer service responses for financial institutions. [Business Wire]
Marchex has launched Engage for Service, an AI-powered solution helping auto dealership service centers recover missed opportunities and increase customer lifetime value. [Business Wire]
Pizza chain Papa John’s has partnered with Google Cloud to enhance customer ordering through AI-powered personalization, chatbots, and virtual assistants. [Yahoo Finance]
PAX Technology has launched an AI Voice Ordering Solution that automates restaurant phone orders, reservations, and payments. [PRWeb]
Taalk.ai has launched Taalk4Good, providing nonprofits and charities with enterprise-grade AI communication agents for automated phone, SMS, and email interactions. [EIN Presswire]
Tavus has integrated Llama AI models into its conversational video interface platform to create authentic digital interactions with significantly improved performance and capabilities. [Meta]
TELUS Digital has partnered with Zendesk to integrate CRM and CCaaS services, combining AI technology through its Fuel iX platform with Zendesk’s customer service solutions. [Business Wire]
Tinder has launched ‘The Game Game’, an OpenAI-powered interactive experience that helps users practice flirting through AI-generated dating scenarios. [Tinder]
Voice News
aiOla has launched Jargonic, an enterprise-focused speech recognition model designed to handle specialized vocabulary, background noise, and diverse accents without extensive retraining. [VentureBeat]
Gladia has launched Solaria, an AI speech recognition model supporting 100 languages with 94% accuracy and 270ms latency for real-time call center communications. [VentureBeat]
Northeastern University researchers have developed Speak Ease, an AI-powered app combining speech recognition, text-to-speech synthesis, and predictive features for speech-impaired users. [Speech Technology Magazine]
Salad Technologies has launched an upgraded Transcription API that delivers industry-leading 95.1% accuracy at $0.16 per hour, undercutting competitors by 40%. [GlobeNewswire]
UC Berkeley and UC San Francisco researchers have developed a brain-activity decoder that translates thoughts to speech in near real-time, testing it with a paralyzed woman. [Cosmos]
Viaim has launched RecDot, AI-powered earbuds that transcribe meetings, translate languages, and provide noise cancellation while offering 36-hour battery life. [digitaltrends]
Translation
Interprefy has partnered with UTS Nîmes tennis tournament to provide AI-powered real-time French subtitles for live coaching and player interviews to 8,000 spectators. [MultiLingual]
Nuanxed has produced over 900 automated literary translations across 60 language combinations, sparking debate about AI’s role in literary translation and its impact on human translators’ livelihoods. [The Markup]
OOONA has launched a Multilingual QC tool that enables simultaneous validation and editing of subtitles across multiple language streams for media localization professionals. [MultiLingual]
Pocketalk has transformed communication in Hancock County courtrooms by providing 14 translation devices capable of interpreting over 80 languages. [WGEM]
Shanghai AI Laboratory researchers discovered that LLMs produced unnatural translations, with over 40% of GPT-4’s outputs containing translationese errors. [Slator]
XL8 has partnered with PMVG to deliver the first AI-based real-time English-to-Spanish caption translation system for US public broadcasting stations. [Speech Technology Magazine]
Search
AmberSearch has developed an AI-powered search platform that helps European SMEs efficiently access scattered internal knowledge while maintaining data privacy and security standards. [Silicon Canals]
Google's AI Mode has expanded search capabilities with Gemini 2.0, enabling chat-like interactions and deep research while potentially reducing clicks to publisher sites. [Search Engine Land]
Google's AI Overviews experienced dramatic growth in entertainment (528%), restaurant (387%), and travel (381%) queries during its March 2025 core update rollout. [Search Engine Land]
Google's shift to generative information retrieval has transformed search from link-based results to AI-generated answers, challenging traditional SEO practices and website traffic models. [Search Engine Land]
Perplexity has partnered with Seattle-based Firmly.ai to integrate e-commerce capabilities into its AI search platform, enabling direct product purchases through the app. [GeekWire]
AI in Journalism
The BBC has announced plans to negotiate with AI companies, boost short-form video content, and enhance its news coverage through BBC Verify and iPlayer expansions. [Press Gazette]
Bloomberg has issued dozens of corrections to AI-generated news summaries since implementing the technology, despite claiming 99% accuracy in its thousands of daily articles. [The New York Times]
Newsquest has expanded its AI-assisted reporting team to 36 journalists across its titles, helping free up other reporters for original local journalism. [Press Gazette]
Health Tech
Apple is developing Health+, an AI-powered health coaching app that will analyze user data and provide personalized wellness recommendations, with a planned launch in 2026. [eWeek]
Dartmouth researchers have demonstrated that their Therabot AI therapy bot reduced depression symptoms by 51% in clinical trials, matching human therapy’s effectiveness in half the time. [MIT Technology Review]
Embold Health has launched AI-powered virtual assistants that help patients navigate healthcare decisions by providing real-time guidance and connecting them with suitable local providers. [Hospital Management]
FirstHx has launched its AI-powered patient intake technology in the US market, partnering with Jevan Consulting to streamline pre-visit medical data collection. [Business Wire]
John Snow Labs has partnered with Guideline Central to launch an AI solution that combines medical LLMs with clinical guidelines for improved healthcare decision-making. [GlobeNewswire]
Oura has launched Advisor, an AI-powered health coach that analyzes members’ biometric data to provide personalized guidance and insights through their smart ring technology. [Business Wire]
Patient Connect has launched AI-powered call scoring that evaluates all healthcare communications, ensuring HIPAA-compliant quality assurance across its medical answering services. [EIN Presswire]
A report by the Peterson Health Technology Institute has indicated that ambient scribe technology is becoming one of healthcare’s fastest-adopted technologies, with promising early results in reducing clinician burnout. [MobiHealthNews]
Suki AI has launched healthcare’s first AI capability to automatically generate prescription orders from ambient doctor-patient conversations. [Business Wire]
Talkspace has launched Talkcast, an AI-powered feature enabling therapists to create personalized podcast episodes for clients between therapy sessions. [Business Wire]
United Digestive has partnered with eClinicalWorks to implement healow Genie, an AI-powered contact center solution providing 24/7 patient support. [Business Wire]
Wolters Kluwer has integrated conversational AI into its vrClinicals for Nursing platform, enhancing virtual reality simulations for nursing education with voice interactions. [Business Wire]
Meanwhile, OpenAI‘s research has revealed both benefits and risks of AI therapy chatbots, which Americans are increasingly using despite therapists’ concerns about replacing human emotional support. [eWeek]
Legal Tech
Bot Mediation has been selected to present its AI-powered legal dispute resolution platform at ABA Tech 2025’s Pitch Competition among 14 other finalists. [LawSites]
Factor has released a 2025 benchmark study showing that while 61.2% of legal departments provide AI access, only 12.1% consider themselves leaders in GenAI adoption. [EIN Presswire]
Park IP has launched Legal Studio, an AI-driven platform that streamlines global patent filing and translation across 180 jurisdictions. [Slator]
Supio has launched Document Intelligence, an AI platform helping personal injury law firms improve case outcomes through automated analysis, drafting, and monitoring features. [Business Wire]
Ed Tech
Afficient Academy has launched an AI-powered SAT Preparation Suite featuring personalized learning paths, comprehensive exercises, and digital practice tests to accelerate student success. [PRWeb]
Anthropic has launched Claude for Education, a higher education-focused AI chatbot service featuring Learning Mode and partnerships with universities, directly competing with OpenAI’s ChatGPT Edu plan. [TechCrunch]
August Schools has launched August Intelligence, an AI agent helping K-12 schools manage student wellbeing through data analysis, administrative automation, and enhanced support coordination. [Business Wire]
Faculty AI has developed a £4mn government-funded content store for marking homework, serving as a prototype for the UK’s planned National Data Library that may sell public data. [The Financial Times]
PowerSchool has launched PowerBuddy for Engagement, an AI assistant that helps families access student information and communicate with schools through natural language interactions. [Business Wire]
Uplimit has launched AI-powered learning agents to help companies upskill employees more efficiently. [VentureBeat]
Funding
Actively AI has raised US$22.5m to develop reasoning-based sales tools that identify high-value prospects, differentiating itself from other AI sales startups. [TechCrunch]
All Gravy has raised US$2.9m to expand its AI-powered employee management platform in the UK after signing major hospitality brands and growing revenue threefold. [Tech.eu]
Brainial has secured undisclosed funding to expand its AI-powered tender assistance software across Europe. [Silicon Canals]
Cambrian has raised US$5.9m in seed funding to develop its AI infrastructure protocol for providing financial data and training AI agents. [FinSMEs]
CoreWeave has raised US$1.5B in a scaled-down IPO, selling fewer shares at a lower price than initially planned due to market volatility. [Bloomberg]
FurtherAI has raised US$5m in funding to enhance its AI-powered insurance automation technology and expand operations globally. [FinSMEs]
heyLibby has raised US$4.5m in seed funding to expand its AI-powered customer engagement platform for fitness and wellness businesses. [PRWeb]
Hi Auto has raised US$15m in Series A funding to expand its AI-powered voice technology for drive-thru restaurants, bringing its total funding to US$23m. [FinSMEs]
Kay.ai has raised US$3m in funding to develop AI co-workers that eliminate manual data entry work for insurance brokers and agencies. [Business Wire]
LTV.ai has raised US$5.2m in Series A funding to enhance its AI-powered personalized email and SmS marketing platform for retail brands. [FinSMEs]
Manus, a Chinese AI startup, has sought a US$500m valuation in its latest funding round. [The Information]
Navina has secured US$55m in Series C funding to expand its AI-powered clinical intelligence platform across US healthcare. [Navina]
Parallel has raised US$3.5m in Seed funding to develop AI agents that automate administrative tasks in healthcare, starting with medical coding automation. [Tech.eu]
Parasail has launched a platform connecting multiple GPU providers to deliver cost-effective AI infrastructure, backed by US$10m in seed funding and several enterprise customers. [TechCrunch]
Phonic has raised US$4m to develop end-to-end AI voice technology with improved reliability and reduced latency. [TechCrunch]
Replit has entered funding talks aiming to raise US$200m, which would increase its valuation to US$3 billion, highlighting investor interest in AI coding tools. [Yahoo Finance]
Ribbon has raised US$8m in funding led by Radical Ventures to expand its AI-powered recruitment platform that automates candidate screening and evaluation. [FinSMEs]
Runway has secured US$308m in Series D funding to advance its AI media tools and expand its film production capabilities, bringing its total funding to US$536.5m. [TechCrunch]
Scale AI has sought a US$25B valuation in a potential tender offer, building on its success in providing labeled data for AI training. [Yahoo Finance]
SchoolAI has raised US$25m in Series A funding to expand its AI-powered classroom platform, which currently serves over 1 million classrooms worldwide. [FinSMEs]
Sesame AI has entered talks to raise US$200m from Sequoia Capital and Spark Capital, potentially valuing the voice assistant startup at over US$1 billion. [Tech in Asia]
Softbank has sought a record US$16.5B one-year bridge loan to fund AI investments in the US, marking its largest dollar-denominated borrowing. [Bloomberg]
Sourcetable has launched an AI-powered autonomous spreadsheet platform and raised US$4.3m in seed funding to democratize complex data analysis for non-technical users. [Tech Funding News]
Sturdy has raised US$6m in seed funding to enhance its AI-powered customer intelligence platform that identifies churn risks across customer interactions. [FinSMEs]
Taxo has raised US$5m in seed funding for its healthcare AI platform, which uses a reasoning engine to help doctors reduce administrative work and increase insurance approvals. [TechCrunch]
Temporal has raised US$146m in Series C funding to expand its microservices orchestration platform into AI applications, bringing its total valuation to US$1.72 billion. [TechCrunch]
Tomorro has secured €25M in funding to expand its AI-powered contract management platform across Europe, establish a German office, and double its AI investment. [Silicon Canals]
TrialKit has raised US$4.25m in seed funding to expand its AI-powered legal discovery platform that helps criminal defense attorneys manage digital evidence. [FinSMEs]
Unframe has emerged from stealth with US$50m in funding to deliver its turnkey enterprise AI platform that creates custom solutions within hours. [Business Wire]
Voize has raised US$9m in seed funding to expand its AI-powered speech recognition technology that helps caregivers save time on medical documentation through voice dictation. [Tech Funding News]
Wrtn Technologies has secured US$56m in funding from investors including Goodwater Capital, adding to South Korea’s growing AI startup investments. [Bloomberg]
Acquisitions
AMD has acquired ZT Systems for US$4.9B, strengthening its data center capabilities and challenging Nvidia in AI infrastructure with an open-system approach. [Yahoo Finance]
Arm Holdings approached Alphawave for acquisition to obtain crucial semiconductor technology for AI processors, but later decided not to pursue the deal. [Yahoo Finance]
Diuna Group has acquired Intertext, marking its third major acquisition and establishing itself as Poland’s largest domestically-owned language service provider. [MultiLingual]
Qualcomm has acquired VinAI's generative AI division, strengthening its AI capabilities with the Vietnamese company’s expertise in automotive and ML technologies. [TechCrunch]
SER Group has acquired AFI Solutions to enhance its SAP document automation capabilities, just one week after purchasing IDP vendor Klippa. [Intelligent Document Processing Community]
There’s More
Anthropic has released research tools and findings revealing how its AI model Claude processes information, plans ahead, and sometimes fabricates explanations for its reasoning. [TechRepublic]
Gartner has predicted global generative AI spending will reach US$644B in 2025, with hardware accounting for 80% of expenditure. [VentureBeat]
OpenAI has assembled an expert panel to guide its philanthropic future as it transitions from a nonprofit-controlled organization to a traditional corporation. [TechCrunch]
And OpenAI's GPT-4.5 has surpassed the Turing test by being identified as human 73% of the time when given a persona prompt during three-party conversations. [Futurism]
The Back Cover
Here’s Postmodern Jukebox with a cover of The Police’s Every Breath You Take.
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!