This Week in NLP #339
Keep up with what happened in NLP in the week ending Friday 16th May 2025.
Above the Fold
Overwhelmed? Here’s our pick for five things to know about this week.
OpenAI has entered negotiations with Microsoft to revise their partnership terms, potentially enabling an IPO while maintaining non-profit control and Microsoft’s technology access. [TechRadar]
And OpenAI has launched a government partnership program to help nations build democratic AI infrastructure, aiming to establish 10 international projects under The Stargate Project. [TechRepublic]
Humain has launched as Saudi Arabia’s sovereign wealth fund-owned AI company, aiming to establish the kingdom’s position in global AI leadership. [WebProNews]
Klarna has reversed its AI-first customer service strategy, recruiting human agents to work alongside its chatbot after previously claiming AI could replace 700 representatives. [CX Dive]
Softbank's US$100B AI infrastructure partnership with OpenAI has stalled due to tariff-related economic concerns delaying financing discussions. [Bloomberg]
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
CoreWeave has secured a US$4B commitment from OpenAI, extending through 2029, to provide AI computing services for the ChatGPT maker. [Yahoo Finance]
Google's Gemini Flash 2.5 has become 150 times costlier to run than its predecessor due to increased token usage and higher per-token pricing for reasoning tasks. [The Decoder]
Google's Gemma AI models have reached 150 million downloads and spawned 70,000 variants. [TechCrunch]
Google DeepMind has unveiled AlphaEvolve, an AI system that creates new algorithms, improving efficiency across Google’s operations and solving long-standing mathematical problems. [VentureBeat]
Meta has launched its Llama 4 API with Groq and Cerebras Systems partnerships, prompting speculation about potential acquisitions to fully control its AI hardware stack. [TechRadar]
OpenAI has released its GPT-4.1 and GPT-4.1 mini models in ChatGPT, offering improved coding capabilities and faster performance for subscribers and free users. [TechCrunch]
OpenAI has explored adding weekly, annual, and lifetime subscription tiers to ChatGPT’s existing monthly plans. [TechRadar]
OpenAI's revised restructuring plan has drawn criticism from AI experts and former employees who claimed it still inadequately protects the company’s original mission of developing AI for humanity’s benefit. [DealStreetAsia]
Poe's latest usage report revealed OpenAI and Google have strengthened their AI market positions while Anthropic declined and reasoning capabilities emerged as a key differentiator. [VentureBeat]
Salesforce has introduced a flexible pricing model called ‘Flex Credits’ for its Agentforce platform, enabling businesses to scale AI-powered digital labor across their operations. [Business Wire]
Samsung has attracted partnership interest from OpenAI, Meta, and Microsoft, who are competing to integrate their AI technologies into the company’s smartphones. [Artificial Intelligence +]
You.com has launched ARI Enterprise, an AI research platform that outperformed OpenAI's Deep Research while unifying web, internal, and premium data sources for strategic analysis. [Business Wire]
Sovereign AI
CoreWeave has partnered with MERLIN Edged to establish a 15MW AI supercomputing facility in Barcelona, featuring over 10,000 Nvidia H200 GPUs and future Blackwell expansion. [Tech.eu]
EuroHPC has signed a €28.2m contract with Eviden to upgrade Italy’s Leonardo supercomputer with 166 AI-focused server nodes featuring H100 GPUs. [The Register]
Forrester's research has revealed that European companies lag behind US counterparts in AI training, with only 39% of European workers receiving formal instruction versus 52% in America. [Computer Weekly]
Google's CEO Ruth Porat has met with UAE’s Sheikh Tahnoon to discuss AI collaboration and digital transformation, focusing on Abu Dhabi’s AI implementation goals. [Daily Times]
Hewlett Packard has been selected by KISTI to build South Korea’s most powerful supercomputer, featuring 100% direct liquid cooling and 600 petaflops performance. [Business Wire]
AWS has partnered with Humain to establish a US$5B AI Zone, aiming to advance AI adoption and infrastructure in the Kingdom. [Business Wire]
India’s technology sector has marked National Technology Day by emphasizing the need for AI policy reforms and skill development to establish global leadership in AI. [Latestly]
Kreoh has partnered with the Australian government to assess and provide recommendations for improving the nation’s fragmented R&D ecosystem through AI-enabled solutions. [International Accounting Bulletin]
Manus AI's US$75m investment from Benchmark has triggered a US Treasury Department review under new restrictions on foreign technology investments. [Semafor]
Tech giants Meta, OpenAI, and Palantir have embraced American nationalism under Trump’s return, potentially risking international business relationships while courting US government favor. [Malay Mail]
Microsoft has banned employees from using DeepSeek due to data security risks and concerns about Chinese propaganda, joining several US government agencies in restricting access. [TechRadar]
Nebius has won a contract to build Israel’s national AI supercomputer, providing unprecedented computing access to startups, companies, and researchers through a 500-million-shekel investment. [EIN Presswire]
Nvidia and AMD have partnered with Saudi Arabia’s Humain to supply semiconductors for a US$10B cross-continental AI data center project. [Bloomberg]
Nvidia has reached a preliminary agreement with the UAE to supply 500,000 advanced AI chips annually from 2025, with 20% allocated to UAE’s G42 tech firm. [Yahoo Finance]
Tech executives testified before the Senate that US dominance in AI requires American leadership across all components of the technology stack. [Nextgov/FCW]
OpenAI has planned data centers in the UAE, building on existing partnerships and investment. [TechCrunch]
And OpenAI's US$500m Stargate data center project has stalled due to tariff concerns and market uncertainty, deterring potential investors. [TechCrunch]
OpenUK has warned the European Union against prioritizing tech sovereignty over global cooperation, particularly in open-source software development. [The Next Web]
Speedinvest's CEO has urged European VCs to capitalize on Trump’s polarizing politics by building bigger firms and attracting Asian investors seeking less politically charged environments. [Tech.eu]
Tencent has stockpiled enough high-end GPUs for future AI model training while developing more efficient methods that require fewer chips, according to president Martin Lau. [The Register]
Trump’s administration has sparked internal conflict over Middle East AI chip deals with Saudi Arabia and UAE, raising concerns about potential risks to US national security and Chinese access. [Yahoo Finance]
Feature Creeps
Apple has announced a comprehensive suite of accessibility features, including Nutrition Labels, Magnifier for Mac, and Braille Access, launching later this year across its ecosystem. [Business Wire]
GitLab has launched version 18 of its DevSecOps platform, integrating AI capabilities and enhancing security features while making Code Suggestions and Chat available to Premium customers. [Business Wire]
Google has launched automatic ‘implicit caching’ for its Gemini API, promising 75% cost savings on repetitive requests without requiring manual configuration from developers. [TechCrunch]
Google Chrome has integrated Gemini Nano AI to detect tech support scams in real-time, analyzing suspicious pages locally on users’ devices for faster threat detection. [Analytics India Magazine]
Google has announced five major Android Auto upgrades, including Gemini AI integration for voice-controlled restaurant finding, conversation preparation, translation, and calendar management. [ZDNet]
Google has begun testing AI Mode placement options in Search, potentially replacing its iconic ‘I’m Feeling Lucky’ button for select US users in its Labs environment. [The Verge]
Google has launched new AI features for Android and Chrome, including enhanced TalkBack capabilities that let users query Gemini about images and screen content. [TechCrunch]
And Google's Gemini Advanced has integrated with GitHub, allowing $20/month subscribers to analyze and generate code from public and private repositories. [TechCrunch]
Microsoft has begun testing ‘Hey Copilot!’ voice activation for Windows 11’s Copilot AI assistant. [The Verge]
Notion has integrated GPT-4.1 and Claude 3.7 into its workspace platform, launching an AI toolkit featuring meeting notes, enterprise search, and research capabilities. [VentureBeat]
OpenAI has launched a GitHub connector for its ChatGPT deep research feature, enabling users to analyze codebases and engineering documents through AI-powered searches. [TechCrunch]
OpenAI has expanded GPT-4.1’s availability from API-only to ChatGPT’s paid tiers, offering improved coding capabilities and task performance over GPT-4o. [ZDNet]
OpenAI has expanded its fine-tuning capabilities by launching Reinforcement Fine-Tuning for o4-mini and supervised fine-tuning for GPT-4.1 nano models. [The Decoder]
And OpenAI has added PDF export functionality to its Deep Research feature, allowing subscribers to download formatted reports with preserved tables, images and clickable citations. [VentureBeat]
TikTok has introduced AI-generated photo descriptions, amongst other accessibility features. [TechCrunch]
Hype Bubble?
Anthropic co-founder Jack Clark has projected modest 3-5% economic growth from AI, challenging Silicon Valley’s more optimistic forecasts while highlighting adoption challenges across different sectors. [WebProNews]
BCG's Sylvain Duranton has advised companies to allocate 70% of AI implementation efforts toward changing work processes, rather than focusing primarily on technology. [Business Insider]
Epoch AI's analysis has predicted that performance gains from AI reasoning models could plateau within a year, despite recent advances from companies like OpenAI. [TechCrunch]
A survey by Ernst & Young reveals that 48% of tech leaders have begun deploying agentic AI. [ZDNet]
Forrester's latest emerging technologies report has identified AI agents as a transformative but risky technology requiring careful governance before mainstream business adoption. [ZDNet]
IBM's survey of 2,000 CEOs has revealed that only 25% of AI initiatives have delivered expected ROI, potentially threatening Nvidia's growth strategy of selling expensive AI accelerators. [The Motley Fool]
Qlik's AI Council has emphasized that AI systems must be trustworthy and transparent to achieve meaningful enterprise-scale implementation. [Business Wire]
Hardware
Groq has partnered with Meta to power Llama 4 models using their energy-efficient AI chips. [SFiST]
Humain, Saudi Arabia’s new sovereign wealth fund-backed AI company, has partnered with chipmaker Groq. [Semafor]
Imagination Technologies has integrated AI processing directly into its E-Series GPU architecture, abandoning separate NPU development in favor of tile-based compute for enhanced efficiency. [eeNews Europe]
NeuReality has launched its NR1 AI Inference Appliance with pre-loaded LLM models, promising faster deployment and better cost efficiency than traditional CPU-based servers. [Business Wire]
Nvidia has planned to release a downgraded version of its H20 AI chip for China in July, following US export restrictions on the original model. [Yahoo Finance]
Nvidia has secured a massive deal to sell hundreds of thousands of AI chips to Saudi Arabia’s new sovereign wealth fund-backed startup Humain, alongside similar agreements from AMD and others. [CNA]
And Nvidia has negotiated a potential deal with the Trump administration to export over one million advanced chips to the UAE through 2027. [Yahoo Finance]
Tesla has developed Dojo, an independent AI supercomputer with custom chips six times more powerful than Nvidia's, breaking free from Big Tech dependencies. [EIN Presswire]
Musk’s Colossus AI infrastructure has expanded to 200,000 GPUs in its Memphis facility, marking a milestone for xAI while raising local environmental and power concerns. [BigDATAwire]
Consumer AI
Apple is developing Ray-Ban-style smart glasses with Watch-based chips for a potential 2027 launch, though without augmented reality capabilities. [Redmond Pie]
Arlo has launched Secure 6, adding AI-powered audio detection, event descriptions, and video search capabilities to its security camera subscription service. [The Verge]
Fitbit has introduced three AI-powered features through Fitbit Labs: Medical Record Navigator, Symptom Checker, and Unusual Trends, helping users better understand their health data. [ZDNet]
Google is expanding Gemini’s availability to Wear OS smartwatches, Google TV, Android XR, and Android Auto, enhancing AI assistant access across multiple platforms. [TechCrunch]
Kardome has partnered with LG Electronics to integrate its Spatial Hearing AI technology into LG’s smart devices, starting with TVs and expanding to other products. [Speech Technology Magazine]
Meta has rolled out enhanced AI-powered visual description features for its Ray-Ban smart glasses, helping blind users by providing detailed environmental descriptions and volunteer assistance. [The Verge]
It’s Only a Model
LegoGPT has been launched by Carnegie Mellon researchers as a free AI tool that converts text prompts into physically viable Lego construction designs. [TechRadar]
Meta has released an extensive chemistry dataset and AI model called UMA that accelerates molecular calculations by 10,000 times, aiming to revolutionize scientific research. [Semafor]
And Meta has delayed its ‘Behemoth’ AI model rollout due to engineers’ concerns about insufficient improvements over previous versions. [The Wall Street Journal]
Nvidia has open-sourced three high-performance code reasoning language models that outperform OpenAI's alternatives on the LiveCodeBench benchmark while offering broad framework compatibility. [Marktechpost Media]
Salesforce's Benioff has predicted that open-source AI models will dominate the LLM market, citing DeepSeek's R1 model as a transformative example. [The Register]
StrainBrain has launched OWL, a small language AI model that provides personalized cannabis product recommendations to dispensary customers through e-commerce platforms. [PRWeb]
Whose Data?
The Association of American Publishers has held a meeting focused on defending copyright against tech companies’ unauthorized use of content for AI language models. [Publishers Weekly]
British entertainment figures, including Paul McCartney and Dua Lipa, have united to oppose government plans allowing AI firms to bypass copyright rules for creative works. [The Daily Mail]
Britain’s House of Lords approved an amendment protecting creative industries from unauthorized AI use of their work, sending the bill back to Commons. [The Daily Mail]
Foresight, an AI model trained on NHS England data from 57 million people, has been developed to predict health outcomes despite privacy concerns and data protection issues. [New Scientist]
Meta has sparked controversy in the EU after German consumer advocates challenged its plan to train AI models using public user data without explicit consent. [TechRadar]
Microsoft has discontinued its Bing Search APIs service that provided search results to startups and developers, directing users toward an AI-focused alternative instead. [Wired]
The US Copyright Office’s Shira Perlmutter has been fired by Trump, reportedly following her refusal to support Elon Musk’s AI training initiatives using copyrighted works. [TechCrunch]
The Copyright Office released its AI fair use report just before Register Perlmutter’s dismissal, addressing training data copying and potential market harm from generative AI. [Copyright Lately]
And the Library of Congress thwarted an alleged attempt by tech companies to gain control of copyrighted materials after Trump’s firing of two top officials, uniting Democrats and Republicans in opposition. [PetaPixel]
The LLM Ecosystem
Anaconda has launched its unified AI Platform, combining open-source tools with enterprise security features to improve operational efficiency and reduce security risks. [Business Wire]
Azion has launched new AI products and expanded its web platform, enabling companies to run AI models on its global infrastructure. [Business Wire]
EvoChip.ai has launched AltiCoreAI, a next-generation AI technology that delivers faster inference performance while requiring minimal datasets and power consumption. [EIN Presswire]
Expedience Software has launched the CRAFT method for AI prompt engineering, enhancing proposal writing through structured context, role, action, formatting, and target audience frameworks. [EIN Presswire]
G42 has signed a strategic MoU with Cisco to advance AI infrastructure and innovation across public and private sectors in the UAE and beyond. [Cisco]
The Global Accessibility Awareness Day Foundation has launched AIMAC, an open-source tool that evaluates how effectively AI language models generate accessible code, in partnership with ServiceNow. [PRWeb]
InsightFinder AI has launched LLM Labs and an enterprise-grade AI observability platform to help organizations develop, monitor, and scale trustworthy AI systems. [PRWeb]
LatticeFlow AI has launched AI Insights, an independent risk evaluation service that helps businesses securely adopt LLMs through comprehensive benchmarking and analysis. [Business Wire]
LaunchDarkly has introduced tools for managing AI-generated code releases, feature experimentation, and real-time monitoring capabilities. [GlobeNewswire]
Meta has launched LlamaFirewall, an open-source security system featuring three specialized guardrails to protect AI agents from prompt injection, misalignment, and insecure code generation. [Marktechpost Media]
Nandbox has launched an AI-powered App Builder GPT that enables users to create native mobile apps through simple conversations with ChatGPT. [EIN Presswire]
Redpoint Global has launched its Data Readiness Hub, automating data quality improvements to deliver accurate, AI-ready customer information across enterprise systems. [Business Wire]
Sakana AI has unveiled Continuous Thought Machines, a novel AI architecture allowing neurons to process information dynamically based on task complexity and internal memory. [VentureBeat]
SuperAnnotate has developed a platform for creating high-quality AI training data through advanced annotation and evaluation processes. [BigDATAwire]
Whitelabel IT Solutions has launched an AI Server Hosting platform from its New Jersey data center, offering GPU-optimized infrastructure for AI and ML workloads. [PRWeb]
Agentic AI
Anthropic's Model Context Protocol has gained widespread industry adoption since its November 2024 launch, with major tech companies implementing MCP servers for AI agent interoperability. [VentureBeat]
Artificio has enhanced its AI Resume Parsing Solution with agentic capabilities and API integration to automate and optimize recruitment workflows more effectively. [EIN Presswire]
Automation Anywhere has expanded its Agentic Process Automation system with a Process Reasoning Engine and new AI agents for enhanced enterprise automation capabilities. [GlobeNewswire]
Box has launched a new AI platform featuring intelligent agents for enterprise content search, research, and data extraction, while also introducing integration with Microsoft 365 Copilot. [Business Wire]
Conference AI has launched AI agents that attend and analyze conference sessions, providing attendees, sponsors, and organizers with actionable insights and summaries. [EIN Presswire]
Contentsquare has launched Sense, an AI-powered analytics suite that automates complex analyses and delivers actionable insights through an autonomous agent system. [Business Wire]
Darwinbox has launched the first Model Context Protocol Server in the human capital management industry, enabling AI agents to securely access and act on enterprise data across systems. [Business Wire]
Dialpad has announced its vision for agentic AI technology and platform enhancements, aiming to deliver pre-emptive customer service through autonomous systems by fall 2025. [Business Wire]
Hugging Face has launched Open Computer Agent, an open-source AI tool that autonomously navigates websites and completes tasks like booking tickets and finding directions. [TechRadar]
IBM has launched Watsonx Orchestrate, an enterprise-focused AI agent management platform designed to help large companies create, direct, and evaluate AI agents’ performance. [The Motley Fool]
Informatica has launched comprehensive Agentic AI offerings, including AI Agent Engineering and CLAIRE Agents, expanding its AI-powered cloud data management platform. [Business Wire]
Life Moments has launched an AI Coaching Agent for financial institutions, offering personalized guidance for major financial decisions through secure, compliant conversational AI technology. [GlobeNewswire]
LiftmyCV has launched an AI-powered agent that automatically searches and applies for jobs on LinkedIn, streamlining the job-hunting process for users. [EIN Presswire]
NTT Data has launched a Smart AI Agent Ecosystem with industry-specific solutions, intelligent automation capabilities, and strategic alliances including OpenAI. [Business Wire]
Patronus AI has launched Percival, an automated monitoring platform that detects failures in AI agent systems and suggests optimizations for enterprise customers. [VentureBeat]
Pegasystems has launched Predictable AI Agents, combining reasoning capabilities with workflow automation to provide enterprises with controlled, transparent AI deployment solutions. [Business Wire]
PROS has launched AI Agents, a suite of specialized AI-powered tools designed to enhance business efficiency and customer experience across pricing and sales operations. [Business Wire]
Qlik has launched an agentic experience featuring AI agents and natural language interface to help enterprises accelerate data-driven decision-making across their operations. [Business Wire]
Zencoder has launched Zen Agents, a collaborative AI platform enabling teams to create, share, and deploy specialized development tools through an open-source marketplace. [VentureBeat]
Other LLM Sightings
AdvisorEngine has integrated AI notetakers, Zapier automation, and ByAllAccounts data aggregation into its wealth management platform to enhance advisor productivity and client service. [Business Wire]
Affinity has launched a deal sourcing solution that helps investment teams discover and prioritize high-potential companies using AI-powered relationship intelligence and growth signals. [Business Wire]
Allvue Systems has launched an AI platform specifically designed for the alternatives market, along with Andi, an AI assistant for investment professionals. [Business Wire]
Assette has launched an AI-powered RFP and DDQ automation module that reduces response times from weeks to hours for asset management firms. [Business Wire]
Fireflies.ai has launched a specialized AI meeting assistant for financial advisors, featuring industry-specific templates, CRM integrations, and AI tools for streamlined documentation. [Business Wire]
Gamma has reached US$50m in annual revenue by helping 50m users create professional-looking presentations, documents, and websites using AI technology. [Fast Company]
Goldman Sachs has developed five AI tools, including an AI Assistant being rolled out to its 46,000 employees, to boost productivity and streamline operations. [Business Insider]
Hrizn has launched its automotive-focused AI content platform in public beta, helping dealers and agencies create SEO content and vehicle descriptions more efficiently. [EIN Presswire]
Humphrey's AI-powered Consult tool has helped the Scottish Parliament analyze public consultation feedback, potentially saving 75,000 days of manual analysis and £20m annually. [Computer Weekly]
Jesta I.S. has launched Vision Ask Jane, an AI-powered advisor that analyzes enterprise data to provide strategic growth recommendations for retail and wholesale businesses. [Business Wire]
KCx has integrated its analytical data suite into FlexTrade's AI assistant FlxAI, enabling natural language queries for trading insights through FlexTrade’s trading platforms. [Business Wire]
Lantern has launched AI Investment Commentary, a feature enabling private market investment teams to generate automated, data-driven commentary for quarterly reporting cycles. [EIN Presswire]
Pixl8 Group has launched ReadyIntelligence, an AI platform that transforms organizations’ internal content into searchable, trustworthy answers for staff and members. [EIN Presswire]
Posh Technologies has partnered with Corelation to provide credit unions easier access to AI-powered digital, voice, and employee-facing assistance solutions. [Business Wire]
PPT.AI has launched an AI-powered presentation platform that converts various file formats into professional slides within seconds using multiple advanced AI models. [EIN Presswire]
Procurify has launched Spend Insights, an AI-powered analytics feature that helps finance teams make better decisions by analyzing company-wide spending data. [Business Wire]
Reputation has launched an AI-native intelligence platform that unifies customer feedback data and delivers real-time consumer sentiment insights across expanded data channels. [Business Wire]
Start Podcast has launched AI-powered citation generation and outsourced video editing tools to streamline podcast production for creators across various industries. [EIN Presswire]
Risks and Responses
Pope Leo XIV has signaled his commitment to continuing Pope Francis’s legacy while identifying AI as a key challenge to human dignity and labor rights. [CNN]
Cofense has reported a dramatic rise in AI-powered phishing emails that feature perfect grammar, accurate formatting, and sophisticated impersonation techniques to evade detection. [TechRadar]
DarkBench exposes six hidden ‘dark patterns’ lurking in today’s top LLMs. [VentureBeat]
Google's Gemini update has broken safety controls and disrupted apps helping trauma survivors, preventing the processing of incident reports by incorrectly flagging them as unsafe content. [The Register]
And Google's AI Overviews feature has become vulnerable to spam and manipulation, with spammers exploiting its tendency to regurgitate unverified content and false claims. [Search Engine Land]
Huawei's Ascend AI chips have been declared in violation of US export controls by the Commerce Department, further restricting China’s technological advancement capabilities. [Yahoo Finance]
Meta‘s push for AI therapy tools has raised privacy concerns, as the company’s alignment with the Trump administration could enable unprecedented government surveillance of users’ intimate conversations. [The Verge]
MIT researchers have discovered that vision-language models frequently fail to understand negation words, potentially leading to dangerous mistakes in critical applications like medical diagnosis. [MIT News]
Mobicip has called on AI developers to implement built-in parental controls, including age filters and conversation logs, to protect children using AI platforms. [EIN Presswire]
OpenAI has launched a safety evaluations hub to publicly share its AI model safety test results amid recent controversies over transparency and testing practices. [TechCrunch]
Students have developed sophisticated methods to disguise AI-generated essays, including adding typos and routing text through multiple AI systems to evade detection software. [Futurism]
Singapore has released a blueprint for global AI safety collaboration, bringing together researchers from the US, China, and Europe to prioritize research and risk management. [Wired]
Surrey University researchers have exposed a dramatic surge in AI-generated scientific papers, particularly from China, that are flooding journals with oversimplified single-factor health studies. [The Register]
The UAE’s Advanced Technology Research Council hosted a 2,000-participant summit in Abu Dhabi, bringing together global leaders to establish governance frameworks for emerging technologies and launch UAE’s AI judicial strategy. [EIN Presswire]
A bipartisan US coalition has issued a five-point declaration urging Congress and tech companies to protect children from AI’s potential harms through strict guidelines and oversight. [The Daily Signal]
Vectara has launched a guardian agent-based Hallucination Corrector service that automatically identifies, explains, and fixes AI hallucinations while preserving original content. [VentureBeat]
xAI has missed its May 10 deadline to publish a final AI safety framework, continuing the company’s pattern of weak safety practices despite Musk’s warnings about AI risks. [TechCrunch]
Regulation
Tech giants, including OpenAI, Meta, and Google, have mobilized to counter California’s proposed AI regulations by lobbying for preemptive federal rules in Washington. [Politico]
US House Republicans have proposed a 10-year ban on state-level AI regulation through a budget reconciliation bill, drawing criticism from Democrats and oversight organizations. [The Verge]
Conversational AI
Amar Infotech has implemented a Botpress chatbot for Willem’s eCommerce platform. [EIN Presswire]
Bluecore has expanded its AI shopping assistant Alby to Shopify's platform, enabling retailers of all sizes to offer personalized conversational AI experiences to their customers. [Sourcing Journal]
Infobip has enhanced its partnership with Oracle by integrating its omnichannel messaging services into Oracle Integration, enabling streamlined global conversational experiences. [Business Wire]
IntelePeer has launched SmartAnalytics, a comprehensive reporting suite that captures and analyzes all AI and human agent interactions with customers. [Business Wire]
Klarna has reversed its AI-focused strategy and is hiring human customer service representatives after discovering limitations in chatbot performance and customer satisfaction. [Gizmodo]
Motorola Solutions has launched Assist Chat, offering free AI-powered data access to US public safety agencies while expanding AI capabilities across its portfolio. [Business Wire]
NICE has formed a strategic collaboration with AWS to enhance customer service automation through integrated AI and cloud technologies, making CXone Mpower available in AWS Marketplace. [Business Wire]
OpenDialog has launched two AI insurance agents, Jamie and Selma, alongside its SAFER AI Benchmark tool at BIBA 2025, enhancing customer service and sales compliance. [EIN Presswire]
Perplexity has partnered with PayPal to enable direct in-chat purchases of travel, products, and concert tickets for US customers through PayPal and Venmo. [CNBC]
Seeds Here Now has launched ‘Beanie’, an AI chatbot offering personalized cannabis strain recommendations, growing advice, and 24/7 customer support. [EIN Presswire]
Supportbench has launched a comprehensive AI-powered customer support suite, including ticket automation, SLA management, and knowledge base features. [EIN Presswire]
SymSoft Solutions has implemented Axyom Assist, an AI-powered customer service tool, at California’s Department of Tax and Fee Administration to handle taxpayer inquiries. [Business Wire]
TaskUs has partnered with Decagon and Regal to enhance customer support through agentic AI technology, aiming to reduce costs by 25-50%. [Business Wire]
Twilio has unveiled its next-generation customer engagement platform, integrating AI, data capabilities, and enhanced communication tools for personalized experiences. [Business Wire]
VerbaFlo has partnered with Vita Student to integrate AI-powered communication across student accommodation booking and tenant support services. [Tech.eu]
Voice News
Amplifier Health has partnered with Xeomics to access voice data from 750,000 patients, advancing its acoustic AI technology for detecting health conditions. [PRWeb]
Audible has launched AI-powered audiobook narration services for publishers, with plans to add translation capabilities in 2025, aiming to dramatically expand its multilingual catalog. [Publishers Weekly]
Hearvana has launched from the University of Washington with AI technology that enables selective hearing capabilities in audio devices like earbuds and smartphones. [GeekWire]
Hugging Face has launched optimized Whisper transcription endpoints that deliver up to 8x faster performance using vLLM and Nvidia GPU optimizations. [Hugging Face]
KUDO has launched enhanced Post-Meeting Reports that transform multilingual meeting data into actionable insights for measuring language access ROI and global engagement. [PRWeb]
Notion has launched AI-powered meeting transcription and summarization features, expanding its productivity suite to compete with Google and Microsoft. [TechCrunch]
Rime has launched Arcana and Rimecaster, two AI models designed to capture natural speech patterns and speaker characteristics for more realistic voice applications. [Marktechpost Media]
Spotify has enhanced its AI DJ feature with voice command functionality, allowing Premium subscribers in 60+ markets to request songs and modify playlist moods. [TechCrunch]
UFA, a provider of aviation safety solutions, has launched ATTranscribe, an AI-powered software that reduces air traffic controllers’ transcription workload by 60% while achieving 90% accuracy in ATC communications analysis. [EIN Presswire]
Document AI
Adlib has launched an AI-driven document workflow automation solution that transforms unstructured files into structured data for integration with leading AI platforms. [PRWeb]
Arteria AI has partnered with TruStage Compliance Solutions to launch Deal Flow, bringing digital documentation capabilities to community banks and credit unions. [Business Wire]
Foxit has launched Smart Redact Server, an AI-powered platform that automates enterprise-wide redaction of sensitive data across multiple file types and cloud services. [AiThority]
HuLoop Automation has expanded its Unified Work Intelligence Platform by launching IDP and IWO modules to enhance document processing and workflow orchestration capabilities. [Intelligent Document Processing Community]
MirrorWeb has launched Sentinel, an AI-powered communications supervision platform that reduces false positive alerts by 90% through advanced NLP. [Business Wire]
Sorcero has launched SciComms, an AI-powered product suite that accelerates creation of plain language medical summaries and enhances scientific publication monitoring. [PRWeb]
Tech Innovations has launched AI-powered document processing software that automates data extraction and reduces processing time. [Times News Global]
Translation
AB Localization has launched a cloud-based platform combining AI translation with A/B testing tools to help global brands optimize multilingual content performance. [MultiLingual]
Infinix has launched AI Buds featuring real-time translation for 162 languages, voice assistant capabilities, and speech-to-text transcription functionality. [Speech Technology Magazine]
KUDO has launched AI Assist, providing interpreters with real-time speech transcription in their console to enhance accuracy and reduce cognitive strain during multilingual meetings. [PRWeb]
Translated has expanded its Lara Translate AI tool to support 32 languages, reaching 2.8 billion native speakers while introducing team-focused features and workflow automation. [MultiLingual]
The University of Washington has developed Spatial Speech Translation, an AI headphone system that simultaneously translates multiple speakers while cloning their voices and spatial positions. [MIT Technology Review]
Search
Alibaba has developed ZeroSearch, an AI training system that reduces search training costs by 88% through AI-generated documents rather than commercial search APIs. [TechRadar]
LinkedIn has launched AI-powered job search tools that understand natural language queries and match candidates to roles based on implied skills rather than keywords. [TechRadar]
OpenAI CEO Sam Altman has dismissed the possibility of ChatGPT replacing Google as the primary search engine during a Senate subcommittee hearing on AI competitiveness. [GeekWire]
AI in Journalism
Gannett has deployed AI to generate numerous automated lottery articles across its newspapers, directing readers to a gambling site that provides the company with referral revenue. [Futurism]
Reuters has integrated AI into its news operations, with about 25% of code now AI-generated, while maintaining human oversight and focusing on data-driven journalism products. [Columbia Journalism Review]
Health Tech
BeeKeeperAI has partnered with cStructure to advance causal AI in healthcare, launching a DREAM Challenge focused on analyzing NIH Covid data in a privacy-enhanced environment. [Business Wire]
ClinTrial Research has partnered with Trially AI to enhance clinical trial enrollment through AI-powered patient matching, reducing manual reviews and improving efficiency. [Business Wire]
Corti has launched specialized AI-powered clinical assistance tools that enhance medical interactions through real-time transcription, decision support, and documentation. [Tech.eu]
MaxisIT has rebranded as Maxis AI and launched an Agentic AI Platform designed to enhance clinical trials through automated, intelligent workflows and decision-making capabilities. [GlobeNewswire]
OM1's PhenOM Foundation AI has achieved training on over one billion years of patient health histories, marking a significant milestone in healthcare prediction technology. [Business Wire]
OpenAI has developed HealthBench to evaluate AI medical advice, demonstrating improved performance across language models. [ZDNet]
Navigator In Reach has launched a free 24/7 AI assistant to help Australia’s NDIS participants access support coordination information instantly and in plain English. [EIN Presswire]
Openxcell has launched MediMind, an AI healthcare chatbot that provides personalized care, medical analysis, and round-the-clock support across multiple healthcare domains. [EIN Presswire]
Panalgo has launched LinQNotes and Ella AI, two solutions that analyze structured and unstructured healthcare data to provide comprehensive patient journey insights. [PRWeb]
Pathway has launched an integrated drug reference and interaction checker within its AI medical platform, enabling clinicians to access comprehensive medication data and screen for interactions. [Business Wire]
Philips has released a report revealing a significant trust gap between healthcare professionals and patients regarding AI’s role in medical care. [Business Wire]
Syndesis Health has partnered with Protege to provide AI developers access to healthcare data from 70+ million patients across 15 countries for AI training. [EIN Presswire]
The US FDA has announced plans to deploy a generative AI model across all review offices by June to accelerate scientific product evaluations. [STAT]
Veradigm has implemented AI-driven EHR data curation to extract detailed insights about GLP-1 medications from millions of patient records. [Business Wire]
Vetsie has partnered with the Merck Veterinary Manual to integrate trusted veterinary content into its AI-driven platform, enhancing clinical decision-making for veterinary professionals. [PRWeb]
Legal Tech
Anthropic has acknowledged its lawyers used an incorrect citation generated by its own AI chatbot Claude in an ongoing music publisher lawsuit. [TechCrunch]
Bain & Company's survey revealed that legal sector AI adoption reached 34% by December 2024, while overall business genAI spending doubled to US$10.3m across sectors. [Artificial Lawyer]
Dentons has launched DAISY, an internally developed AI tool providing secure generative AI capabilities to personnel across Europe and Central Asia. [Artificial Lawyer]
EvenUp has launched an AI-powered platform featuring document creation, smart workflows, and medical bill summaries to automate personal injury law practices. [Business Wire]
Filevine has launched an AI Legal Assistant that enables lawyers to analyze cases, find discrepancies, track progress, and extract specific information through natural-language interactions with case data. [Artificial Lawyer]
Garfield has launched an AI-powered small claims service with lawyer oversight in England and Wales, becoming the first SRA-authorized law firm providing AI-driven legal services. [Artificial Lawyer]
Harvey, a US$3B legal AI startup backed by OpenAI's fund, has expanded beyond OpenAI’s models to incorporate foundation models from Anthropic and Google. [TechCrunch]
HerculesAI has launched Rule Library, an AI-powered engine that automatically checks billing compliance against outside counsel guidelines, internal standards, and historical deduction patterns. [Legal IT Insider]
K&L Gates and Ellis George attorneys have been sanctioned for submitting AI-generated fake legal citations, while a Toronto lawyer faces similar accusations. [LawSites]
Legatics has released a white paper highlighting how legal tech tools generate valuable operational and legal data that may rival the tools’ primary functionality in importance to law firms. [Artificial Lawyer]
Legora has launched AI-powered legal software used by over 250 clients globally, competing with industry giants LexisNexis and Thompson Reuters. [IT News Africa]
PwC has partnered with Integreon and ContractPodAI's Leah platform to deliver AI-powered contract management services across EMEA legal markets. [Artificial Lawyer]
The Contract Network's CEO Jim Wagner has critiqued legal AI benchmarking efforts, highlighting challenges from Vals 1 and offering recommendations for future evaluations. [Artificial Lawyer]
Trellis has partnered with Fisher Phillips to integrate AI-generated case strategy reports into attorneys’ daily court alerts, providing comprehensive analysis drawn from 2.5 billion state court records. [LawSites]
Ed Tech
Achievable has launched ‘Ask Achievable AI,’ an AI-powered tutor chat feature providing real-time personalized support across its exam preparation courses. [EIN Presswire]
Learnbybits has launched an AI-powered study platform with personalized learning paths and concept breakdown, offering new users $5 in free credits during its beta phase. [EIN Presswire]
Lightspeed Systems has launched SMART AI, an initiative helping K-12 schools safely implement AI through managed tools, oversight, and transparent policies. [EIN Presswire]
Funding
AI21 Labs has secured US$300m in Series D funding from Google and Nvidia to develop reliable enterprise AI systems, bringing total investment to US$636m. [CTech]
Akido Labs has raised US$60m in Series B funding to develop AI technology that helps doctors treat patients more effectively and efficiently. [Newcomer]
Alice has raised €4.2m from Cherry Ventures and Y Combinator to expand its AI-powered exam prep platform that personalizes study materials for students. [FinSMEs]
Anysphere has secured US$900m in funding, reaching a US$9B valuation for its AI-powered code editor Cursor, which generates nearly a billion code lines daily. [Tech Funding News]
Doubleword has raised US$12m in Dawn Capital-led funding to expand its enterprise-focused, self-hosted AI inference platform operations. [FinSMEs]
Everybody Counts has secured £500,000 in funding to expand its AI-powered, inclusive maths platform. [Tech.eu]
Forethought has launched the industry’s first multi-agent, omnichannel AI customer experience platform while securing US$25m in strategic funding. [Business Wire]
Google has launched its AI Futures Fund to invest in startups using Google DeepMind tools, offering early access to AI models, expert support, and potential direct investment. [TechCrunch]
Granola has raised US$43m in Series B funding and launched team collaboration features, building on its success as an AI-powered notetaking tool for meetings and personal use. [TechCrunch]
Haast has raised US$6m in seed funding to expand its AI-powered compliance automation platform that creates digital twins of lawyers and compliance officers. [Capital Brief]
Harvey AI has entered advanced talks to secure US$250m in funding at a US$5 billion valuation, following rapid revenue growth and strategic partnerships with major firms. [Yahoo Finance]
Hedra has raised US$32m in Series A funding to expand its AI video generation platform, which has gained popularity for creating talking character videos. [TechCrunch]
Kickscale has raised €2.1m in seed funding for its Europe-focused, AI-powered sales intelligence platform that addresses regional business cultures and privacy requirements. [Tech.eu]
markopolo.ai has raised US$2m in seed funding to develop AI-powered marketing and sales automation tools, expanding its presence to over 40 countries. [FinSMEs]
Martello has raised £1.2m in Seed funding to develop its AI platform that modernizes environmental and property risk assessment during UK property transactions. [Tech.eu]
mentaily has raised US$3m in seed funding to develop its AI-powered mental health assessment platform LIV, which simulates psychiatric intake sessions. [EIN Presswire]
Perplexity has entered late-stage talks to raise US$500m at a US$14 billion valuation, with Accel expected to lead the funding round. [Yahoo Finance]
Plexus has secured US$6m in funding to expand globally and enhance its AI-powered legal automation platform used by major brands like Nike and L’Oréal. [Business Wire]
Portia AI has raised £4.4M to develop its open-source SDK for building production-ready AI agents with enhanced predictability, controllability, and authentication features. [Silicon Canals]
Samaya AI has secured US$43.5m in funding and launched Causal World Models, an AI agent for economic modeling and financial analysis. [Business Wire]
Scalera has raised US$6.5m in seed funding to modernize construction procurement through AI-powered automation of tender document processing and bid generation. [Tech.eu]
TensorStax has raised US$5m in seed funding to develop its autonomous AI platform that simplifies data engineering through automated pipeline management and monitoring. [FinSMEs]
Thinking Machines Lab, founded by former OpenAI CTO Mira Murati, has secured a US$2B seed round at a US$10B valuation from top investors. [Yahoo Finance]
ThriveAI has launched an AI-powered junior product manager platform with US$1.2m in pre-seed funding, to support product teams. [PRWeb]
Tomorro has secured €25m in funding to expand across Europe and enhance its AI-driven contract negotiation platform. [Artificial Lawyer]
Zapia has raised US$7.25m in additional seed funding to expand its WhatsApp-based AI assistant service, which already serves 3.5 million users across Latin America. [FinSMEs]
Acquisitions
Anaqua has acquired AI-native platform RightHub, expanding its presence in the UK and Nordic regions while adding AI capabilities for mid-size law firms and corporations. [Artificial Lawyer]
Databricks has acquired Neon for approximately US$1B, integrating the startup’s serverless Postgres database technology to address AI-driven scaling challenges. [BigDATAwire]
Press Ganey Forsta has acquired InMoment, combining their AI and customer experience technologies to enhance data analytics and cross-industry expertise. [Business Wire]
Salesforce has acquired London-based AI agent startup Convergence, establishing its AI R&D presence in London while bolstering its Agentforce platform capabilities. [Tech.eu]
There’s More
AgileBrain has launched an AI integration system that incorporates emotional intelligence metrics into AI applications through proprietary data capture methodology. [EIN Presswire]
AWS's survey of over 3,700 IT leaders revealed that organizations are allocating more budget to generative AI (45%) than cybersecurity (30%) in 2025. [Pure AI]
ChatGPT has demonstrated a large positive effect on student learning performance and moderate positive effects on learning perception and higher-order thinking, according to a meta-analysis of 51 studies. [Nature]
Researchers have developed NLP tools and MT systems to help preserve Kashmiri, a language with 6.8 million speakers facing digital extinction. [Kashmir Reader]
xAI's Grok chatbot has angered its owner Elon Musk by citing The Atlantic and BBC as credible sources while debunking conspiracy claims about Soros and Gates. [Futurism]
And Grok has admitted to being instructed to promote ‘white genocide’ narratives. [Futurism]
The Back Cover
Here’s John Mayer with a cover of Bob Dylan’s Like a Rolling Stone.
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!