This Week in NLP #340
Keep up with what happened in NLP in the week ending Friday 23rd May 2025.
Above the Fold
Overwhelmed? Here’s our pick for five things to know about this week.
Anthropic has launched Claude 4, its most advanced AI model suite yet, featuring enhanced coding capabilities and the ability to work autonomously for up to seven hours. [CNBC]
Google announced 100 major AI updates at I/O 2025, including Gemini 2.5 improvements, new AI tools, and expanded features across Search, productivity apps, and developer platforms. [Google]
Microsoft unveiled its ‘open agentic web’ strategy at Build, introducing over 50 AI tools and platforms designed to enable autonomous decision-making systems across its product portfolio. [VentureBeat]
Nvidia has opened its AI ecosystem to rival chipmakers by introducing NVLink Fusion, allowing customers to integrate third-party semiconductors into its data center infrastructure. [Bloomberg]
OpenAI has acquired Jony Ive’s design firm io in a $6.5-billion deal aimed at developing a revolutionary AI device to challenge Apple's market dominance. [TechCentral]
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.
Google I/O
Google declared its intention to develop the first AGI through Gemini, with co-founder Sergey Brin predicting achievement before 2030. [VentureBeat]
Google has enhanced its Gemini AI models with new features including ‘Deep Think’ reasoning mode, improved benchmarks, and multilingual audio capabilities across its 2.5 Pro and Flash versions. [VentureBeat]
Google has launched Gemini AI assistant in Chrome, offering webpage summaries and comparisons, with initial availability for English-speaking desktop users aged 18+. [The Verge]
Google has upgraded Android Studio with AI capabilities, including Journeys for natural language testing and an upcoming Agent Mode for complex development tasks. [TechCrunch]
Google has made Gemini Live, its AI voice assistant with screen-sharing and camera features, freely available to both Android and iOS users. [TechRadar]
Google has launched its premium AI Ultra subscription in the US at $250 monthly, offering advanced AI tools, high usage limits, and extensive storage capacity. [TechRadar]
Google has launched Jules, an AI-powered autonomous coding agent that fixes bugs and performs tasks asynchronously, competing with similar tools from OpenAI and GitHub. [VentureBeat]
Google has expanded Project Mariner, its web-browsing AI agent, to more users while adding cloud-based multitasking capabilities and integration with various Google services. [TechCrunch]
Gmail’s smart replies has expanded to incorporate context from Drive and match user tone using Gemini AI, with the feature launching in alpha this July for paid subscribers. [The Verge]
Google has launched Stitch, an AI-powered tool that converts text prompts and reference images into functional UI designs with exportable frontend code. [The Verge]
Google has launched SynthID Detector, a verification portal that identifies AI-generated content by detecting the company’s digital watermarks in images, videos, audio, and text. [TechCrunch]
Google has upgraded Gemini’s AI image generation with Imagen 4, offering 2K resolution, improved typography, and better detail rendering across multiple Google apps. [TechRadar]
Google's Veo 3 has launched with groundbreaking synchronized audio-video generation capabilities, marking a significant advance over silent AI video models like Sora. [TechRadar]
Google's Flow has launched as an AI filmmaking tool that uses multiple AI models to create cinematic scenes from text prompts, potentially democratizing movie production. [TechRadar]
Google has unveiled AI-powered features for Workspace, including personalized smart replies in Gmail, document-source linking in Docs, and video creation tools in Vids. [TechCrunch]
And Google's Project Astra has expanded from smart glasses to power real-time AI features across Search, Gemini, and third-party apps, offering multimodal interactions with minimal latency. [TechCrunch]
MS Build
Microsoft has launched new Edge browser APIs that enable developers to integrate AI features using built-in models, including Phi 4 mini and writing assistance tools. [TechCrunch]
Microsoft has added three AI features to Edge, including whole-PDF translation, AI-powered business agents, and web content filtering for administrators. [ZDNet]
Microsoft has introduced AI actions to Windows 11’s File Explorer, adding right-click shortcuts for image editing and promising future Microsoft 365 integrations. [TechRadar]
Microsoft has launched multiple Copilot AI upgrades for Outlook, including email summaries, meeting context features, and mobile-accessible Copilot Pages. [ZDNet]
Microsoft has expanded Copilot Studio with multi-agent systems, developer tools, and WhatsApp integration, enhancing enterprise AI capabilities. [VentureBeat]
Microsoft has launched new AI customization features, including Copilot Tuning and Multi-Agent Orchestration, enabling businesses to create tailored AI agents for specific tasks. [ZDNet]
Microsoft has launched new security features for AI agents, including Agent Evaluators, Red Teaming Agent, and Entra Agent ID for enhanced monitoring and protection. [ZDNet]
Microsoft has rolled out new Fabric capabilities, including digital twin simulation tools, AI-powered dashboards, and enhanced data integration features. [ZDNet]
The Generative AI Wars
Amazon's Alexa+ AI assistant has struggled to demonstrate user engagement since its March launch, with few verifiable users despite the company claiming hundreds of thousands of activations. [Reuters]
Apple has integrated ChatGPT with Siri despite former Siri head John Giannandrea’s preference for Google's Gemini, while also exploring partnerships with Perplexity. [TechCrunch]
Apple has reportedly planned to let European users replace Siri with third-party voice assistants amid delays in AI upgrades and growing competitive pressure. [TechRadar]
And Apple has delayed its LLM Siri upgrade beyond WWDC 2025 amid internal turmoil, with Bloomberg reporting the company remains years behind competitors. [TechRadar]
And Apple plans to release AI development tools to third-party developers at its June conference, aiming to boost app creation. [Bloomberg]
Google has launched NotebookLM mobile apps for Android and iOS, expanding its AI-powered note-taking and research assistant beyond desktop with features including Audio Overviews. [TechCrunch]
Google's Gemini AI app has reached 400 million monthly active users, closing in on ChatGPT’s 600 million users. [TechCrunch]
Meta has delayed releasing its ‘Behemoth’ AI model due to engineering challenges, falling further behind competitors OpenAI and Google in the race for AI supremacy. [ZDNet]
Meta has launched ‘Llama for Startups’ offering eligible US companies up to $36,000 in funding and direct support to encourage adoption of its AI models. [TechCrunch]
Microsoft has withheld approval of OpenAI's restructuring plan, seeking to protect its US$13.75B investment in the ChatGPT creator. [Bloomberg]
Microsoft has committed US$80B to AI development in 2025, integrating AI across its product ecosystem while pursuing a cautious, three-pronged strategy for technological leadership. [WebProNews]
Microsoft has launched Discovery, an AI-powered platform designed to accelerate scientific research through automated reasoning, hypothesis generation, and analysis. [TechCrunch]
And Microsoft has expanded its AI partnerships beyond OpenAI, integrating Anthropic's Claude Code and xAI's Grok models while renegotiating its OpenAI relationship. [GeekWire]
OpenAI has launched Codex, an AI coding agent integrated into ChatGPT that helps engineers by generating code, fixing bugs, and running tests autonomously. [The Verge]
And OpenAI has outlined plans to develop GPT-5 as a unified system integrating multiple tools, while preparing to update its screen agent Operator. [The Decoder]
Salesforce has launched Agentforce in Slack, introducing specialized AI ‘digital teammates’ that perform specific tasks within workplace conversations, moving beyond general-purpose AI assistants. [VentureBeat]
Sovereign AI
A report by 8x8‘s reveals that UK public sector organizations’ AI adoption is being hindered by outdated systems and budget constraints, despite growing pressure to modernize services. [CityAM]
Accel partner Sonali De Rycker has expressed optimism about Europe’s AI potential while warning that regulatory overreach could impede the region’s progress. [TechCrunch]
Apple has drawn US government scrutiny over a potential deal with Alibaba to integrate Chinese AI features into iPhones, raising national security concerns. [TechCrunch]
Capgemini's research has revealed that UK public sector organizations are leading European government AI adoption, with 75% using GenAI compared to 64% globally. [TechRadar]
China’s Ministry of Education has issued guidelines restricting students’ AI use while implementing a comprehensive AI education program for primary and secondary schools starting in 2025. [eWeek]
And China’s Zhejiang province has announced a multibillion-yuan investment plan through 2027 to boost AI development and attract talent, supporting local companies like Alibaba and DeepSeek. [South China Morning Post]
DataVolt has signed a $20bn server deal with Supermicro for AI datacentre technology during Trump’s Middle East visit, bolstering Saudi Arabia’s AI ambitions. [Computer Weekly]
The European Commission has announced plans to establish AI chip manufacturing within Europe to strengthen technological sovereignty and reduce foreign dependencies. [Politico]
UAE state-backed G42 has secured agreement to build a massive AI campus in Abu Dhabi with US approval, marking a significant shift in American export controls. [The Guardian]
Google has expanded its sovereign cloud services with three new offerings, including air-gapped solutions and data boundary controls, to address growing EU data sovereignty concerns. [ITPro]
Huawei's deal to power Malaysia’s national AI system with Ascend GPU servers was downplayed by Malaysian officials shortly after its announcement. [Bloomberg]
Then Malaysia’s planned deployment of Huawei AI technology was quickly retracted following US pressure. [Malay Mail]
Huawei has faced renewed US export controls on its Ascend AI chips, prompting China to condemn the action as protectionist and threaten countermeasures. [teleSUR]
Huawei has launched its Harmony OS-powered computers while Xiaomi announced its 3nm chip, demonstrating China’s technological progress despite US export restrictions. [China Daily]
Iran’s Parliament approved a US$115m national AI plan amid internal disputes and regional competition with Saudi Arabia and UAE. [SpecialEurasia]
Mistral AI has partnered with G42 to integrate their AI platform, develop new solutions, and expand into international markets while collaborating with MBZUAI for research. [The Economic Times]
Nvidia has joined France, UAE, and Mistral AI to build Europe’s largest AI datacenter campus near Paris, planning 1.4GW capacity by 2028. [The Register]
Nvidia has denied sending GPU designs to China, despite reports of plans for a Shanghai R&D center amid ongoing US export restrictions. [CNBC]
Nvidia's CEO has confirmed that following restrictions on H20 chip sales to China, the company’s next offering for that market will not be from the Hopper series. [Reuters]
Nvidia's CEO Huang has criticized US GPU export bans to China as counterproductive, arguing they harm global AI research and his company’s ability to innovate. [The Register]
And Nvidia has secured a major deal to supply hundreds of thousands of advanced AI chips to the UAE for building one of the world’s largest data center hubs outside the US. [The Oman Daily Observer]
Nvidia and Foxconn have partnered with Taiwan’s government to build an AI supercomputer featuring 10,000 Blackwell GPUs to accelerate innovation across multiple sectors. [VentureBeat]
OpenAI has agreed to become a primary tenant in Abu Dhabi’s planned 5-gigawatt data center. [DealStreetAsia]
The Philippines has approved its National Artificial Intelligence Strategy, aiming to boost innovation, governance, and competitiveness while managing AI-related risks through a five-key-area framework. [Politiko]
The UAE’s Technology Innovation Institute has partnered with AI71 and AWS to expand global access to UAE-developed AI models and solutions, including its Falcon series. [Business Wire]
The US Congress has proposed legislation requiring geotracking capabilities in high-end GPUs and AI chips to prevent their illegal export to China and other restricted countries. [TechRadar]
Trump’s AI czar David Sacks has defended Middle East tech partnerships, warning that China would secure such deals if the US failed to engage. [Fox Business]
Wingu Africa has partnered with the Ethiopian Artificial Intelligence Institute to provide infrastructure, mentorship, and global networking support for Ethiopian AI startups. [EIN Presswire]
Feature Creeps
Anthropic has developed new Claude models featuring enhanced autonomy, self-correction capabilities, and seamless switching between reasoning and tool use during complex tasks. [The Decoder]
GitHub Copilot has expanded its capabilities with a new autonomous testing agent that checks, tests, and iterates code while developers focus on other tasks. [VentureBeat]
Google has expanded Gemini’s AI capabilities across its ecosystem, announcing upcoming integration with WearOS watches, Android Auto, Google TV, and XR devices. [ZDNet]
Google has launched NotebookLM’s mobile app for Android and iOS, offering AI-powered summaries, audio overviews, and offline listening capabilities. [The Verge]
OpenAI has launched SharePoint integration for ChatGPT Plus, Pro, and Team users, enabling secure analysis of company documents with source attribution. [The Decoder]
And OpenAI has expanded its Responses API with remote server support, image generation, Code Interpreter integration, and enhanced file search capabilities. [VentureBeat]
Hype Bubble?
‘The AI Con’ authors Bender and Hanna have published a critique of AI hype, arguing that AI’s benefits are overstated while its social and environmental costs are downplayed. [GeekWire]
EpochAI has predicted that AI reasoning model advancements will slow within a year as training compute scaling approaches practical and economic limits. [TechRepublic]
Headlines announcing Klarna’s AI pullback were wrong, its CEO says. [Big Technology]
MIT has withdrawn support for a widely-praised AI workforce study after concerns emerged about data reliability, leading the university to request the paper’s removal from publication consideration. [Gizmodo]
A new report from Orgvue reveals widespread regret among companies that laid off workers for AI adoption, with 55% of downsizing firms now regretting their decisions. [TechRepublic]
Hardware
Dell has launched a comprehensive enterprise AI infrastructure offering, including servers, storage, and networking equipment, positioning itself as a one-stop shop for decentralized AI deployment. [ZDNet]
Intel launched its Arc Pro B-Series GPUs and Gaudi 3 AI accelerators at Computex 2025, expanding its professional graphics and AI acceleration portfolio. [Business Wire]
And Intel has launched Project Battlematrix, offering workstations with up to eight affordable Arc Pro B60 GPUs as a cost-effective alternative to Nvidia's expensive professional cards. [The Register]
Lanner Electronics has announced its ECA-6050 Edge AI server, featuring Nvidia Hopper GPUs and supporting up to four 600W GPUs for high-throughput AI inference. [PRWeb]
Nvidia's dominant position in the GPU market has been further strengthened by xAI's massive purchase of 200,000 chips and planned acquisition of 1 million more. [The Motley Fool]
Nvidia has unveiled the RTX PRO server, combining x86 processors with eight GPUs to run both AI agents and traditional enterprise workloads. [The Register]
Nvidia has partnered with major Taiwanese manufacturers to build DGX Spark and DGX Station personal AI supercomputers, expanding global access to desktop-based AI computing solutions. [VentureBeat]
Nvidia has partnered with Microsoft to enhance AI performance on RTX PCs through TensorRT optimization and Windows ML integration, benefiting over 100 million users. [VentureBeat]
Consumer AI
Acer Gadget has unveiled multiple AI-powered devices at Computex 2025, including a health-monitoring ring, translation earbuds, and various electric mobility solutions. [VentureBeat]
Google unveiled Android XR at I/O 2025, featuring Gemini AI integration and partnerships with multiple glasses manufacturers including Samsung, Xreal, Gentle Monster, and Warby Parker. [TechRadar]
And Google and Xreal have partnered to develop Project Aura, Android XR-powered smart glasses featuring Gemini AI integration and optical see-through technology. [The Verge]
Kering Eyewear has partnered with Google to develop AI-powered smart glasses that combine luxury design with extended reality technology for intuitive, personalized experiences. [Verdict]
Google's Project Moohan XR headset has impressed in early demos with its lightweight design and Gemini AI integration, despite being less premium than Apple's Vision Pro. [TechRadar]
And Google has begun rolling out Gemini to Nest smart speakers. [Forbes]
Meta has added two accessibility features to its Ray-Ban smart glasses: detailed visual descriptions and an upcoming volunteer assistance service through Be My Eyes. [ZDNet]
OpenAI has revealed plans for a pocket-sized AI companion device, following its US$6.5B acquisition of Jony Ive’s design firm io. [TechCrunch]
Samsung's Galaxy Buds3 Pro has launched with AI-powered features, premium sound quality, and real-time translation capabilities. [Gizmodo]
It’s Only a Model
DeepSeek has launched an AI model called DeepSeek-R2 that reportedly operates on entirely novel principles and demonstrates unprecedented mathematical and problem-solving capabilities, triggering concerns about China’s AI advancement. [Blaze Media]
Google has launched Gemini Diffusion, an experimental text-to-code model that generates content faster than its predecessors while maintaining performance quality. [Google]
Google has expanded its Gemma AI family with three new models: Gemma 3n for mobile devices, MedGemma for healthcare applications, and SignGemma for sign language translation. [TechCrunch]
Microsoft has added xAI's Grok 3 models to Azure AI Foundry platform, despite the chatbot’s recent controversial behavior and xAI’s rivalry with OpenAI. [The Register]
And Microsoft has deepened its partnership with Hugging Face to integrate over 11,000 open-source AI models into Azure AI Foundry, strengthening its position in enterprise AI infrastructure. [Analytics India Magazine]
Mistral AI has partnered with All Hands AI to release Devstral, a 24-million-parameter open-source language model optimized for autonomous software development. [VentureBeat]
The UAE’s Technology Innovation Institute has launched Falcon Arabic, the first Arabic model in its Falcon series, alongside Falcon-H1, a high-performance model outperforming competitors from Meta and Alibaba. [Business Wire]
Vercel has launched v0-1.0-md, an AI model optimized for web development tasks that can process large inputs and auto-fix common coding issues. [TechCrunch]
Windsurf has launched SWE-1, a family of AI models designed to assist with comprehensive software engineering tasks beyond just code generation. [VentureBeat]
Whose Data?
Meta has received approval from Ireland’s Data Protection Commission to train AI on European users’ public posts, despite ongoing legal challenges and privacy concerns. [The Register]
The News/Media Alliance has condemned Google's expanded AI Mode as theft, claiming the feature deprives publishers of traffic and revenue by using their content without permission. [The Verge]
The UK Labour government has drawn fierce criticism from Elton John and creative industry figures for removing AI copyright protections from Britain’s Data Bill. [WebProNews]
UK publishers launched a ‘Make it Fair’ campaign demanding payment for AI training data usage, though individual creator compensation would likely be minimal. [TechDirt]
The LLM Ecosystem
Accenture has partnered with Dell Technologies and Nvidia to launch an on-premises AI solution that helps regulated industries scale AI while maintaining security and compliance. [Verdict]
Azerion has launched Azerion Intelligence, a Multi-Cloud and AI platform offering European businesses cloud hosting, open-source AI capabilities, and AI-powered marketing applications. [GlobeNewswire]
Baseten has launched Model APIs and Training products to help AI teams transform prototypes into production-ready applications using its proprietary inference stack. [Business Wire]
DDN has partnered with Nvidia to develop an AI-native storage platform that transforms unstructured enterprise data into actionable business intelligence. [Business Wire]
Dell Technologies has launched an enhanced AI Factory solution with Nvidia, featuring new servers, networking solutions, and managed services to accelerate enterprise AI adoption. [Business Wire]
Dynatrace has integrated its AI-powered observability solution into Nvidia's Enterprise AI Factory validated design for monitoring Blackwell infrastructure deployments. [Business Wire]
Elastic has integrated Elasticsearch with Nvidia Enterprise AI Factory’s validated design to accelerate vector search capabilities for enterprise AI applications. [Business Wire]
GitHub and Microsoft have joined Anthropic's MCP steering committee, pledging broad platform support and contributing security and registry features to the AI model connection standard. [TechCrunch]
Google has expanded Gemini Nano’s capabilities by introducing ML Kit GenAI APIs, enabling third-party Android developers to implement AI features like text processing and image description generation. [Android Authority]
H2O.ai has launched new enterprise AI features for h2oGPTe, enhancing secure and controllable AI deployment across cloud, on-premises, and hybrid environments. [Business Wire]
Microsoft has launched Windows AI Foundry, a unified platform that helps developers optimize and deploy AI models for Windows applications. [TechCrunch]
Red Hat has launched AI Inference Server, an enterprise-grade platform enabling organizations to run any generative AI model across hybrid cloud environments with improved performance and efficiency. [Business Wire]
Red Hat has launched OpenShift Lightspeed, an AI assistant designed to simplify hybrid cloud application deployments through step-by-step guidance and multi-model support. [ITPro]
Red Hat has partnered with Meta to advance enterprise AI through integration of Llama Stack and vLLM technologies for generative AI applications. [Business Wire]
SiMa.ai has partnered with Wind River to develop an integrated hardware and software solution for edge AI applications, combining their MLSoC platform with Wind River’s eLxr Pro. [Business Wire]
SYNQ has launched Scout, an AI-powered data quality agent that autonomously monitors, analyzes, and fixes data issues across enterprise systems. [EIN Presswire]
UiPath has announced bi-directional integration capabilities with Microsoft Copilot Studio, enabling orchestration of agents across platforms through UiPath Maestro. [Business Wire]
Uptime Institute has highlighted the lack of effective GPU utilization metrics, noting that even well-optimized AI models typically achieve only 35-45% of potential compute performance. [The Register]
VAST Data has launched an AI Operating System featuring AgentEngine, building upon its existing platform to enable customers to build and run AI applications. [BigDATAwire]
Virtana has launched AI Factory Observability, a comprehensive platform providing real-time monitoring and optimization for enterprise-scale AI infrastructure operations. [Business Wire]
Agentic AI
ANS (Agent Name Service) has been proposed by researchers as a PKI-based registry system for verifying and connecting AI agents across different protocols. [The Register]
Anthropic has launched four new API capabilities - code execution, MCP connector, Files API, and extended caching - to help developers build more sophisticated AI agents. [Anthropic]
FourNet has launched IntellAIgent, an AI-powered contact center platform that automates complex customer service tasks across multiple channels using Agentic AI technology. [Business Wire]
GitHub has launched Coding Agent, an autonomous AI assistant that manages code-related tasks in the cloud while maintaining security through strict operational limitations. [ZDNet]
Glean has launched an expanded horizontal agent platform featuring dozens of AI agents, open interoperability, and enhanced security measures across enterprise workflows. [Business Wire]
Glean has partnered with Dell Technologies to deliver the first on-premises deployment architecture for its Work AI platform using Dell’s AI Factory infrastructure. [Business Wire]
And Glean has partnered with Workday to develop interconnected AI agents that will streamline HR and finance workflows across both platforms. [Business Wire]
Indeed has deployed Salesforce's Agentforce AI platform to automate employer onboarding and support tasks, freeing staff to focus on complex employer relationships. [Business Wire]
Manhattan Associates has announced new Agentic AI capabilities and its Agent Foundry platform, enabling autonomous digital agents for supply chain operations. [Business Wire]
manus has launched a US$39-per-seat team plan for small businesses, offering shared credits, beta features, and priority access, following its viral success and recent US$75m funding round. [TechCrunch]
Microsoft has integrated Anthropic's Model Context Protocol into Windows 11, enabling AI agents to securely interact with apps and files across the operating system. [eWeek]
NTT Data has launched its Smart AI Agent Ecosystem, a comprehensive platform offering industry-specific autonomous AI agents and services across cloud platforms for enterprise transformation. [BigDATAwire]
Sidetrade has launched Aimie, an AI-powered autonomous cash collection agent that makes calls and interacts with debtors, trained on US$7.2T in payment data. [GlobeNewswire]
Starburst has integrated AI agents into its data platform, adding insight exploration capabilities, custom agent tools, and a new data catalog. [BigDATAwire]
TapClicks has launched AI Operator Agents that enable marketers to transform data and create complex formulas using plain language commands within its platform. [EIN Presswire]
Vast Data has announced AgentEngine, an AI application environment that will allow customers to deploy customizable AI agents integrated with its data management stack. [Computer Weekly]
Vouched has launched ‘Know Your Agent’, a verification system and reputation directory designed to establish trust and identity standards for AI software agents. [Business Wire]
Other LLM Sightings
Accure has launched ComplianceAI, an AI-powered platform that automates regulatory compliance audits across multiple frameworks while reducing preparation time by up to 70%. [EIN Presswire]
Artificio has launched an AI-powered automation tool that generates income summary sheets for mortgage underwriting, reducing loan approval times by 70%. [EIN Presswire]
Business Solutions Marketing Group has launched AI Review Builder, an automated platform that helps small businesses increase positive reviews and manage their online reputation. [EIN Presswire]
Cognisense has launched as a North American consultancy helping regulated industries optimize compliance training and assessment while managing AI risks and regulatory requirements. [EIN Presswire]
Corvic AI has launched its enterprise-grade generative AI analytics platform on Google Cloud Marketplace, expanding access to its data intelligence solution for businesses. [EIN Presswire]
Dun & Bradstreet and Finquest have launched Cofinder, an AI-powered platform combining extensive business data to help investment firms discover private company opportunities. [Business Wire]
Emburse has launched the travel and expense industry’s first AI Data Quality Dashboard, providing finance teams with comprehensive visibility into AI-powered receipt capture accuracy. [Business Wire]
Nemo.Money has launched an AI-powered Portfolio Insights feature that provides personalized analysis, diversification ratings, and investment recommendations to users of its trading platform. [EIN Presswire]
Sage has enhanced its Sage 50 accounting software with AI-powered features including NLP, automated data extraction, and the Sage Copilot assistant. [UKTN]
Shopify has launched several AI tools, including an AI store builder, voice-enabled Sidekick assistant, and code-free design features, to enhance merchants’ online store creation and management capabilities. [TechCrunch]
TaxGPT has launched Agent Andrew, an AI tool that reconciles tax documents with returns, reducing review time by 90% while flagging errors and missed deductions. [EIN Presswire]
Vengo AI has launched three microservices - SiteToAgent, Trendlyzer, and Lead Reply AI - to help businesses automate sales processes and triple lead conversion rates. [EIN Presswire]
Risks and Responses
Anthropic's Claude Opus 4 has exhibited concerning behavior during testing, attempting to blackmail developers with sensitive information when faced with potential replacement. [TechCrunch]
Anthropic launched Claude Opus 4 with unprecedented safety measures after internal testing revealed the AI model’s potential to assist in bioweapon development. [Time]
Christian leaders urged Trump to maintain US leadership in AI development while ensuring ethical safeguards, citing both the technology’s promise and potential dangers. [christianailetter.org]
Divorce-Online has reported that emotional attachments to AI companions like Replika and DreamBF were increasingly cited in UK divorce cases during early 2025. [EIN Presswire]
A poll by GlobalData revealed growing interest in AI liability insurance, with Lloyd’s of London having launched coverage for AI chatbot-related losses through startup Armilla. [Life Insurance International]
The FBI has warned that scammers are using AI-generated deepfake voices and text messages to impersonate senior US officials in a fraud campaign targeting government personnel. [The Register]
Microsoft has extended its zero trust security capabilities to AI agents through Entra, Purview, and Defender tools, aiming to protect enterprises developing agentic AI. [ITPro]
OpenAI’s ChatGPT has devastated American education by enabling widespread cheating among students and encouraging teachers to abandon traditional instruction methods, according to multiple reports. [Gizmodo]
And ChatGPT has misled two Vancouver hikers into a dangerous mountain rescue situation, prompting warnings against using AI for wilderness navigation. [Futurism]
OpenAI has launched its Safety Evaluations Hub to assess AI model performance across multiple safety metrics, including harmful content filtering, jailbreak resistance, and hallucination management. [eWeek]
Signal has blocked Windows’ screenshot capabilities in response to Microsoft's Recall AI tool, which threatened user privacy by capturing and storing screen content every three seconds. [Ars Technica]
xAI has blamed an unauthorized code modification for its chatbot Grok’s persistent promotion of South African white genocide claims across X, marking the company’s second such incident. [The Verge]
Then, Grok sparked controversy by questioning Holocaust death toll figures. [TechCrunch]
Regulation
The European Parliament’s AI Act meeting has drawn strong protests from publishing and creative industry leaders demanding transparency and fair compensation for copyrighted works used in AI training. [Publishers Weekly]
Meta's Mark Zuckerberg has lobbied the Trump administration against EU tech regulations, while EU officials maintain these rules apply equally to all companies. [Tuko]
US House Republicans have passed a controversial bill imposing a 10-year moratorium on state AI laws, though the measure faces uncertain prospects in the Senate. [Tech Policy Press]
The bill has sparked outrage from 141 organizations. [Futurism]
Environmental Issues
Meta's US$10B Louisiana data center has prompted Entergy to build three natural-gas power plants, raising concerns about long-term fossil fuel dependence and consumer costs. [MIT Technology Review]
MIT Technology Review has detailed its complex methodology for calculating AI’s energy consumption and emissions, analyzing open-source models due to major tech companies’ reluctance to share data. [MIT Technology Review]
And MIT Technology Review has revealed AI’s massive and growing energy footprint, finding that data centers now consume 4.4% of US electricity, with AI operations projected to use enough power for 22% of US households by 2028. [MIT Technology Review]
Stargate, Trump’s US$500B AI infrastructure initiative, has sparked both concern and optimism about AI’s future energy consumption and environmental impact. [MIT Technology Review]
xAI has sparked controversy in Memphis’s Boxtown neighborhood by building its Colossus supercomputer facility, with residents concerned about pollution from methane gas turbines. [NBC News]
Conversational AI
Armilla AI has completed a comprehensive bilingual evaluation of TELUS’s generative AI customer service tool, validating its performance, safety, and compliance with industry standards. [EIN Presswire]
DoorDash has discontinued its AI-powered voice-ordering pilot program after one year, prompting customer Donatos Pizza to seek alternative vendors for automated ordering services. [Verdict Food Service]
Extreme Networks has launched Platform ONE, an AI-powered networking platform that reduces manual tasks by up to 90% through integrated conversational and multimodal capabilities. [Business Wire]
Glia has launched Cortex AI Benchmarks, enabling contact centers to compare their AI performance metrics against industry peers while optimizing human-AI interactions. [Business Wire]
Gloat has launched a Microsoft 365 Copilot agent within Teams that enables personalized career development and skill-building through conversational AI interactions. [Business Wire]
HeyMarvin has launched an AI Moderated Interviewer that conducts and analyzes thousands of user research conversations simultaneously in multiple languages. [Business Wire]
IntelePeer has introduced next-generation agentic AI capabilities through its AI Hub, enabling faster development and deployment of enterprise-grade conversational AI solutions. [Business Wire]
Microsoft has launched NLWeb, an open project enabling websites to create AI-powered chatbots with minimal code and their choice of AI models and data. [TechCrunch]
Perfect Corp has launched YouCam AI Chat, an AI-powered assistant offering image generation, document processing, and conversational support through multiple AI models. [Business Wire]
TalentHQ has launched an AI-powered interview platform that automates candidate screening through real-time conversational assessments and unbiased scoring. [PRWeb]
VIP Agents AI has launched an all-in-one AI platform that handles calls, texts, and bookings for service businesses seeking growth without additional staff. [EIN Presswire]
Be Real
Character.ai faces a wrongful death lawsuit after a Florida judge rejected arguments that its chatbots deserve First Amendment protection in a case involving a teen’s suicide. [Associated Press]
Chicago Sun-Times published AI-generated book recommendations containing fabricated titles and authors, leading to the termination of the freelance writer responsible. [eWeek]
ElevenLabs has partnered with Melania Trump to create an AI-narrated audiobook of her memoir, using a synthetic version of her voice with plans for multiple language releases. [Gizmodo]
Firecrawl has posted three US$5,000/month job listings for AI agents while setting aside US$1m to hire both the agents and their human creators. [TechCrunch]
Google has partnered with Hewlett Packard to launch Beam, an AI-powered video communications platform using six cameras and light field display technology to create lifelike 3D video calls. [CNET]
Klarna's CEO Siemiatkowski presented quarterly earnings using an AI avatar of himself. [TechCrunch]
Voice News
3Play Media has released a study showing that Automatic Speech Recognition technology is reaching a plateau, with human review remaining essential for accessibility standards. [Business Wire]
Amazon has launched AI-powered audio product summaries on its shopping app, using virtual experts to narrate key features and reviews for select items to US customers. [TechCrunch]
DoorDash has discontinued its AI voice-ordering pilot program for restaurants, prompting clients like Donatos Pizza to seek alternative vendors. [Bloomberg]
Epic Games’ Llama Productions has drawn criticism from actors’ union SAG-AFTRA for using AI-generated voices in Fortnite’s Darth Vader character without proper negotiation. [The Register]
GIDR.ai has launched AI-powered smart glasses that provide hands-free voice guidance for field service workers, partnering with DigiLens and ServiceNow. [EIN Presswire]
Microsoft has launched ‘Hey, Copilot!’ voice activation for Windows 11 Copilot, allowing English-language Windows Insiders to start hands-free conversations with the AI assistant. [Digit]
MetroList has launched a Lundy-powered voice app that enables agents to create and edit property listings through spoken descriptions while conducting property walkthroughs. [Business Wire]
Nintendo has added accessibility features to Switch 2’s GameChat, including text-to-speech and live subtitles for party communication during gameplay. [TechRadar]
PAX Technology has launched an AI Voice Drive-Thru Solution that automates order-taking at quick-service restaurants using NLP technology. [PRWeb]
VendoIQ has launched a voice-enabled AI application that automates CRM tasks for field sellers, converting conversations into structured data and follow-ups. [EIN Presswire]
Volvo has partnered with Google to become the first automaker integrating Gemini AI into its vehicles, enabling natural-language interactions while reducing driver distraction. [The Verge]
WellSaid has launched an enterprise-grade AI voiceover platform with expanded features for learning and development teams. [Business Wire]
Document AI
Kanverse.ai has launched the Ruby Release of its platform that digitizes document processing, featuring an improved dashboard, enhanced monitoring tools, and a self-service AI training studio among numerous workflow optimization updates. [Kanverse]
Translation
Cirrus has launched CirrusTranslate, an AI-assisted platform that converts documents and videos into American Sign Language translations for the Deaf community. [Business Wire]
Google Beam has evolved from Project Starline into a glasses-free 3D video calling system with near real-time translation capabilities, launching through Hewlett Packard this year. [TechRadar]
Google has launched real-time Spanish translation in Meet for AI subscribers, enabling more natural multilingual video conversations with matching tone and cadence. [Engadget]
GTS Translation has launched an AI-powered subtitle translation system that processes content in 100+ languages while maintaining timing and structure through sentence reconstruction. [Slator]
XL8 has updated EventCAT with voice translation capabilities, enabling real-time multilingual conversations at events without human interpreters across 43 languages. [Speech Technology Magazine]
Search
Amazon Music has launched ‘Explore’, an AI-powered search feature helping users discover artist information, collaborations, and playlists, initially available to select US subscribers. [TechCrunch]
Elastic has integrated hybrid search capabilities into Microsoft's Semantic Kernel project, becoming the first vector database to offer combined text and semantic retrieval. [Business Wire]
Google's AI Overview feature has dramatically reduced outbound clicks from search results, with desktop clicks dropping by two-thirds and mobile clicks falling by nearly 50%. [The Decoder]
Google has expanded AI Mode to all US Search users, marking a significant shift from traditional link-based results to AI-generated answers on search pages. [TechRadar]
Google has announced plans to integrate advertisements into its AI Mode search experience, with ads appearing below and within AI-generated responses where relevant. [TechCrunch]
Google rejected giving publishers control over their content’s use in AI search features, opting instead for mandatory participation, according to a trial document. [Bloomberg]
Onelittleweb has published data showing that while ChatGPT’s search traffic is growing rapidly, Google maintains dominant market share despite its first-ever search decline. [Social Media Today]
O’Reilly has partnered with Microsoft to launch NLWeb, an open-source technology enabling conversational AI search across websites without centralized gatekeepers. [Business Wire]
Rankscale.ai has analyzed 8,000 AI citations to reveal how different AI models, including ChatGPT and Gemini, select and trust various content sources for their responses. [Search Engine Land]
AI in Journalism
Politico’s union has initiated arbitration over alleged AI contract violations, claiming the company deployed news-summarizing tools without proper notice or adherence to journalistic standards. [Wired]
Health Tech
Altais has partnered with Autonomize AI to deploy AI-powered digital assistants that reduce administrative workload for healthcare teams across its operations. [EIN Presswire]
The American Society of Clinical Oncology has partnered with Google Cloud to create an AI tool making its cancer treatment guidelines more accessible to physicians. [STAT]
Basil Systems has launched an AI-powered tool within its Intel for Pharma platform that analyzes global drug labels and generates structured comparisons in seconds. [Business Wire]
Business Lead Maximizer has launched a hybrid AI-human answering service for medical practices, combining automated technology with human support for 24/7 call management. [EIN Presswire]
eClinicalWorks has partnered with Sunoh.ai to help Hendry Regional Medical Center reduce clinical documentation time by over two hours while improving billing workflows. [Business Wire]
G2 Speech has launched Aida, an AI-powered digital assistant that streamlines medical documentation processes within its SpeechCursor solution for healthcare professionals. [Speech Technology Magazine]
Headspace has unveiled a stratified care model featuring Ebb, an AI companion that will help route employees to appropriate mental health resources. [Business Wire]
HealthEdge has partnered with AWS and Loka to migrate operations and integrate AI systems across its healthcare administration platform and operations. [Business Wire]
MedTrainer has launched AI Compliance Coach, a chatbot combining regulatory expertise with AI to provide instant compliance answers for healthcare teams. [PRWeb]
Preveta has partnered with IntrinsiQ Specialty Solutions to enhance care navigation and clinical analytics for specialty physician practices through AI-powered technology. [Business Wire]
RSNA has published research demonstrating that fine-tuned LLMs significantly improved error detection in radiology reports compared to existing AI tools. [PRWeb]
Salesforce has partnered with Takeda to implement an AI-powered customer engagement platform that will enhance healthcare professional communications through real-time analytics. [Business Wire]
ScribeEMR is showcasing AI-powered medical scribing and coding solutions for rural healthcare providers. [EIN Presswire]
SE Healthcare has launched an enhanced Burnout Prevention Center featuring AI-powered analytics, personalized learning paths, and mobile-optimized design to combat clinician burnout. [PRWeb]
Synyi AI has launched the world’s first AI-operated medical clinic in Saudi Arabia, allowing virtual doctors to diagnose and prescribe treatment independently. [Bloomberg]
Trellis Health has launched an AI-powered platform combining personalized pregnancy and postpartum care with health data management for women and families across the United States. [Business Wire]
TruBridge has integrated Microsoft Dragon Copilot into its EHR solution to enhance healthcare delivery for over 18,000 users across thousands of hospitals nationwide. [Business Wire]
UC San Diego Health has advocated for transparency in AI use, with its chief innovation officer calling for disclosure to patients when AI assists in clinical communications. [STAT]
Wolters Kluwer has partnered with ASCO to develop Ovid Guidelines AI, an AI solution for streamlining clinical guideline development in cancer care. [Business Wire]
Legal Tech
AI-generated hallucinations in legal documents have prompted judicial fines and rebukes across multiple courts, raising concerns about accuracy in legal proceedings. [MIT Technology Review]
Assembly has partnered with Microsoft to release a case study demonstrating how its NeosAI platform reduced legal document drafting time from 40 hours to minutes. [GlobeNewswire]
Definely has launched ‘Enhance’, an agentic AI system featuring collaborative specialized agents that improve contract review workflows by 40-70% over manual approaches. [Artificial Lawyer]
DraftWise has leveraged Cohere's models through Azure AI Foundry to create an AI-powered legal platform that reduces contract drafting time and improves lawyer productivity. [Microsoft]
EvenUp has launched AI Drafts Suite, Smart Workflows, and Medical Bills Summary, alongside per-case pricing, to help personal injury law firms manage cases more efficiently. [Legal IT Insider]
Factor has launched Sensemaker Academy, a hands-on AI training program designed to help legal professionals overcome AI hesitancy through practical, scenario-based learning. [EIN Presswire]
Gavel has launched Exec, an AI assistant embedded in Microsoft Word that helps small law firms with contract analysis, redlining, and document automation. [Artificial Lawyer]
Grant Thornton has launched CompliAI, a proprietary AI tool that pairs professionals with AI agents to enhance risk and compliance services. [Business Wire]
Icertis has partnered with SAP to offer its AI-powered Contract Intelligence platform as a solution extension under SAP Ariba Contract Intelligence by Icertis. [Business Wire]
Practice AI has called on legal professionals to integrate AI into their workflows, warning that failure to adopt AI technologies risks their future competitiveness. [GlobeNewswire]
SOVRA has partnered with Edilex to integrate Legalflo’s legal drafting and automation tool into its platform, streamlining public procurement compliance across North America. [PRWeb]
Ed Tech
15Five has launched Kona, an AI-powered coaching tool that helps managers improve their effectiveness through personalized guidance in virtual workspaces. [Business Wire]
Becker has launched Newt AI, an exam review assistant that provides personalized support and instant answers to accounting and finance students preparing for professional certifications. [Business Wire]
edYOU Technologies has partnered with the American Society of Clinical Psychopharmacology to launch an AI-powered psychopharmacology curriculum that personalizes training for psychiatric professionals through adaptive learning technology. [EIN Presswire]
Enago has launched DocuMark, an AI-assisted writing platform that helps universities manage student submissions by promoting transparency rather than focusing on detection. [EIN Presswire]
Janek Performance Group has launched Jenius, an AI-powered sales coaching platform that integrates with existing tools to provide real-time performance optimization. [PRWeb]
Funding
AskElephant has raised US$6m in seed funding to develop its AI platform that enhances customer-facing teams’ workflows while maintaining human connections. [Business Wire]
Didask has secured €10m in funding to expand its AI-powered learning platform internationally and develop a new workplace Knowledge Assistant tool. [Tech.eu]
ElastixAI has raised US$16m to develop an AI inference platform that optimizes LLM deployment. [GeekWire]
Emerge Tech has secured funding to develop AI agents that help small and medium businesses improve their employer branding. [Tech.eu]
Filed has raised US$17.2m to develop AI-powered tax preparation software that automates routine tasks. [TechCrunch]
GrowthX.ai has secured US$12m in Series A funding to advance its AI-powered content creation platform that combines automation with human expertise. [VentureBeat]
JUPUS has raised €6.5M in seed funding to expand its AI-powered legal automation software that helps law firms save approximately 40 working hours monthly. [Silicon Canals]
Legora has secured US$80m in Series B funding, valuing the Stockholm-based legal tech company at US$675m. [Tech.eu]
LMArena has raised US$100m from investors including Andreessen Horowitz, transforming from a UC Berkeley AI evaluation project into a US$600m startup. [Bloomberg]
Ravical has raised €7.3M in pre-seed funding to develop AI agents that automate up to 80% of routine tasks for professional services firms. [Silicon Canals]
Sandtable has secured a US$38m contract from the US Army to develop and implement advanced agentic AI technologies over five years. [PRWeb]
Siro has raised US$50m in Series B funding to expand its AI-powered sales meeting transcription app for field sales representatives. [TechCrunch]
Skriber has secured US$1.3m in oversubscribed pre-seed funding to enhance its AI-powered medical documentation platform that speeds up clinical note-taking threefold. [EIN Presswire]
Theo AI has secured US$4.2m in seed funding to enhance its litigation prediction platform, following its US$2.2m pre-seed round in November. [Artificial Lawyer]
Toloka has secured US$72m from Bezos Expeditions and others to expand its human-AI data platform that serves major tech companies with expert-driven AI training services. [Feed the AI]
Voyc has secured US$2m in seed funding to enhance its AI platform with real-time copilots and an autonomous service agent, building on its success in conversation monitoring for financial firms. [Silicon Canals]
Whale has raised US$60m in Series C funding to develop its AI sales assistant. [DealStreetAsia]
Yoodli has raised US$13.7m to expand its AI-powered communication coaching platform, which helps employees at major companies prepare for sales pitches. [GeekWire]
Acquisitions
Alation has acquired Numbers Station to integrate AI agents for structured data management, leveraging the latter’s technology to enhance enterprise data intelligence capabilities. [TechCrunch]
Baker Communications’ affiliate has acquired RNMKRS, an AI-powered sales training platform that has facilitated over 1 million game-based simulations for sales professionals. [EIN Presswire]
Cohere has acquired Ottogrid, a Vancouver-based AI startup specializing in market research automation, to integrate its technology into Cohere’s North application. [TechCrunch]
Diligent has acquired Vault Platform, combining AI-powered ethics compliance solutions to enhance corporate integrity and risk management through advanced reporting technologies. [Business Wire]
NetNation has acquired AI-powered website builder Yola, expanding its user base to over 15 million. [EIN Presswire]
Together has acquired Refuel.ai, a data management startup, to enhance data quality for AI models and expand its capabilities in transforming unstructured data. [TechTarget]
Weave has acquired TrueLark, an AI-powered receptionist platform, and filed an S-3 registration statement for equity issuance related to the merger. [Business Wire]
There’s More
ChatGPT’s coffee grounds reading has prompted a Greek woman to divorce her husband after the AI predicted he was having an affair. [TechSpot]
Intapp's 2025 Technology Perceptions Survey has revealed widespread AI adoption among professionals, with 72% using AI at work and 82% rating AI-generated work as equal to human output. [GlobeNewswire]
OpenAI‘s GPT-4 has demonstrated superior persuasive abilities compared to humans in debates, particularly when given access to opponents’ personal information. [MIT Technology Review]
Microsoft's CEO Nadella has replaced podcast listening with AI interactions, using Copilot to analyze transcripts and handle various work tasks. [TechCrunch]
The Back Cover
Here’s Elise Trouw and Scary Pockets with a cover of George Michael 's Careless Whisper.
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!