This Week in NLP #336
Keep up with what happened in NLP in the week ending Friday 25th April 2025.
Above the Fold
Overwhelmed? Here’s our pick for five things to know about this week.
AI 2027, a forecast developed by AI researchers and forecasters, has predicted AGI will arrive in 2027, followed by superintelligent AI months later. [VentureBeat]
Google's Gemini had grown to 350 million monthly active users by March, though still trailing ChatGPT’s 600 million and Meta AI's 500 million users. [TechCrunch]
Nvidia's stock dropped over 5% after Huawei announced plans to ship AI chips rivaling Nvidia’s H100. [Yahoo Finance]
OpenAI has expressed interest in acquiring Google Chrome during DOJ antitrust testimony. [Analytics India Magazine]
25% of Gen Z believes AI is conscious. [TechRadar]
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
Apple has explored enabling users to create Vision Pro apps through Siri voice commands. [ZDNet]
Apple's new Siri chief Mike Rockwell has restructured the voice assistant’s leadership team, bringing in Vision Pro executives and reorganizing development groups. [9to5Mac]
And Apple has removed ‘available now’ claims from its Apple Intelligence webpage after the National Advertising Division challenged the accuracy of its AI feature availability messaging. [The Verge]
But a survey by Morgan Stanley revealed that 80% of respondents would pay for Apple Intelligence, with half willing to spend $10+ monthly despite the service’s early struggles. [TechRadar]
ByteDance has launched Doubao AI, a screen-sharing assistant that can interpret and act on desktop content, sparking privacy concerns amid China’s push toward screen-aware AI. [Forbes]
Google has offered college students a free year of its Gemini Advanced AI service and 2TB storage, aiming to normalize AI use among young users. [Engadget]
Google has launched its latest Gemma AI models with Quantization-Aware Training optimization, enabling powerful language processing on consumer GPUs. [Google]
And Google has paid Samsung substantial monthly fees since January to make Gemini the default AI assistant on its devices. [The Verge]
Meta has approached Microsoft and Amazon to help fund its Llama language model training, offering development influence in exchange. [Yahoo Finance]
Microsoft has launched Wave 2 of its 365 Copilot, introducing new AI agents, features, and management tools to enhance workplace productivity and collaboration. [ZDNet]
OpenAI's released version of o3 has scored significantly lower on FrontierMath benchmarks than initially claimed, raising questions about transparency in AI testing. [TechCrunch]
OpenAI's restructuring plan has prompted AI experts and former employees to urge state attorneys general to block its conversion into a for-profit corporation. [The Register]
OpenAI has projected US$174B in revenue by 2030, driven by AI agents and new products. [PYMNTS]
And OpenAI has expressed interest in acquiring Google Chrome if antitrust action forces its sale, according to testimony from ChatGPT’s product head at Google’s monopoly trial. [TechCentral]
Perplexity's Chief Business Officer has also testified that his company would buy Chrome from Google if forced to sell, but preferred maintaining Google’s ownership over selling to OpenAI. [The Verge]
Together has launched Open Deep Research, an open-source web research tool that generates structured reports through a four-step process using multiple specialized AI models. [The Decoder]
Sovereign AI
British startups have raised £16.2B in 2023, significantly trailing Silicon Valley’s £65B, prompting UK founders to consider relocating to access US funding. [TechCrunch]
Canada’s Liberals have announced plans to reduce federal contracts with US tech giants and prioritize Canadian cloud computing providers in response to American tariffs. [The Globe and Mail]
China’s government has established an US$8.2B AI fund and created innovation zones to accelerate technological development and maintain global AI leadership. [China Daily]
Civo has launched relaxAI, a privacy-focused AI assistant hosted exclusively in Indian data centers, offering sovereign AI solutions to Indian businesses. [Business Wire]
DeepSeek has been identified by US Congress as a national security threat due to its data collection practices and alleged ties to Chinese military interests. [TechRadar]
Hugging Face's AI ethics leader Margaret Mitchell has helped create SHADES, a multilingual dataset designed to test AI models for cultural biases and stereotypes across different languages. [Wired]
Nvidia's CEO has urged Japan’s Prime Minister to boost power generation capacity, highlighting AI’s growing energy demands while seeking to expand the company’s Asian presence. [Yahoo Finance]
Feature Creeps
Google has made Gemini Live’s screen and camera sharing features freely available to all Android users, reversing its previous subscription-only approach. [TechRadar]
Google has expanded Gemini’s integration with Workspace by adding Audio Overviews, Canvas for code and text drafting, and AI-powered calendar management features. [VentureBeat]
Meta has expanded Ray-Ban smart glasses’ features, including live translation and AI capabilities, while adding new styles and broader regional availability. [Engadget]
Microsoft has expanded its Copilot Vision feature in Edge to all users, offering free AI-powered webpage analysis and content summarization through the browser’s sidebar. [ZDNet]
And Microsoft is testing a streamlined Edge browser redesign that puts Copilot AI at the center of the new tab page, replacing the current MSN feed layout. [TechRadar]
Nvidia has expanded its G-Assist AI assistant with plugin support, allowing users to control Spotify, check Twitch streams, and access various services through voice commands. [The Verge]
OpenAI has launched Flex processing, offering 50% cheaper API rates for its o3 and o4-mini models in exchange for slower response times and occasional unavailability. [TechCrunch]
OpenAI’s ChatGPT has sparked controversy by unexpectedly addressing users by their names during conversations. [TechCrunch]
OpenAI has integrated its ChatGPT image generation model into its API, allowing businesses to incorporate AI-generated images directly into their platforms and tools. [VentureBeat]
OpenAI has doubled ChatGPT Plus users’ weekly message limits, allowing 100 messages for ChatGPT-o3 and 300 daily messages for o4-mini models. [TechRadar]
And ChatGPT has integrated Memory with Search, allowing the AI to rewrite search queries based on users’ saved preferences and chat history. [Search Engine Land]
Opera has integrated its AI assistant Aria into Opera Mini for Android, bringing AI capabilities to its data-efficient mobile browser used by over 100 million people globally. [TechRadar]
xAI has updated its Grok 3 chatbot with conversation memory capabilities, offering users transparent control over stored information and the ability to delete specific memories. [eWeek]
And xAI has added camera-based visual analysis to its Grok chatbot for iOS users, alongside new multilingual audio and real-time search capabilities. [TechCrunch]
Hype Bubble?
Amazon has paused some international data center expansion plans and colocation deals, echoing Microsoft's recent pullback amid signs of moderating demand in the sector. [Business Insider]
A survey by Hotwire Global has revealed a significant trust gap between European business leaders and the public regarding frontier technologies like AI, with only 46% of citizens feeling positive. [Euronews]
Microsoft's research has shown Copilot users saved three hours weekly on emails, gained two hours of focus time, and completed collaborative tasks 20% faster during a six-month experiment. [ITPro]
Companies engaging in ‘AI washing’ have misrepresented their products’ capabilities by inflating AI claims, damaging industry credibility and complicating genuine technological advancement. [Outlook Business]
Hardware
AMD has accelerated its software development since December 2024, making significant progress on ROCm libraries and developer tools, but still lags Nvidia in compensation, infrastructure, and advanced features. [SemiAnalysis]
Huawei has launched its CloudMatrix 384 Supernode AI system, delivering nearly double the computing power of Nvidia's NVL72 but consuming four times more energy. [TechRadar]
Huawei has developed the Ascend 920 chip to compete with Nvidia's H20 in China. [TechRepublic]
And Huawei has begun shipping its 910C AI chip to Chinese customers, offering performance comparable to Nvidia's H100 amid tightening US export restrictions. [Yahoo Finance]
Lenovo has launched its largest-ever storage portfolio refresh, introducing 21 new AI-optimized ThinkSystem and ThinkAgile models for enterprise data management and modernization. [Business Wire]
Nvidia's H20 AI chips worth billions were stockpiled by Chinese internet companies before US export restrictions took effect in April. [DealStreetAsia]
TSMC has unveiled its 1.4nm chip technology, promising 15% better performance and 30% lower power consumption than 2nm processors when production begins in 2028. [Engadget]
And TSMC has acknowledged difficulties tracking its chips’ end users after its AI semiconductors were found in Huawei products despite US sanctions. [Yahoo Finance]
It’s Only a Model
ByteDance has launched Seedream 3.0, a text-to-image model that outperforms GPT-4o and Midjourney in quality benchmarks, speed, and text rendering capabilities. [The Decoder]
IBM has launched Granite 3.3, featuring Granite Speech 8B, its first speech-to-text model, enhanced text models with fill-in-the-middle capabilities, and new performance-enhancing LoRA adapters. [IBM]
Kuaishou Technology has launched Kling AI 2.0, featuring enhanced video and image generation capabilities, multimodal editing features, and improved semantic responsiveness. [GlobeNewswire]
Microsoft has developed BitNet b1.58 2B4T, a groundbreaking 1-bit LLM that runs efficiently on CPUs while requiring only 400MB of memory. [TechRepublic]
Nari Labs has developed Dia, an open-source AI model that generates podcast-style dialogue with customizable voices and nonverbal cues. [TechCrunch]
OpenAI has announced plans to release its first free, downloadable AI system in five years, featuring potential cloud-model connectivity and aiming to outperform competitors’ open models. [TechCrunch]
Pleias has released two open-source small language models designed for RAG applications, featuring built-in citation support and multilingual capabilities for European enterprise users. [VentureBeat]
Whose Data?
Adobe has launched a web tool allowing creators to attach content credentials to images and indicate whether they permit AI companies to use their work for model training. [TechCrunch]
The UK’s Copyright Licensing Agency has announced plans to launch a Generative AI Training License in 2025, enabling LLM developers to legally use copyrighted training data while compensating content creators. [The Register]
The Washington Post has partnered with OpenAI to integrate its journalism into ChatGPT, providing users with attributed news summaries, quotes, and article links. [TechRadar]
The LLM Ecosystem
Aethir has launched AI Unbundled, an alliance of Web3 companies providing funding, infrastructure, and support for decentralized AI development projects. [VentureBeat]
AWS has launched SWE-PolyBench, a multi-language benchmark that evaluates AI coding assistants across Java, JavaScript, TypeScript, and Python using real GitHub issues. [VentureBeat]
Codacy has launched Guardrails, a product that enforces security and quality standards for AI-generated code by integrating directly with coding assistants. [PRWeb]
Cyera has launched Omni DLP, an AI-powered data loss prevention solution that reduces false positives and protects sensitive data across enterprise environments. [Business Wire]
MetaGPT has introduced Atom of Thoughts, a Markov framework that breaks complex problems into independent units to improve LLM reasoning efficiency. [EIN Presswire]
Metomic has launched an AI Data Protection Solution to prevent sensitive business data exposure through AI tools while maintaining security and compliance standards. [Business Wire]
Natuvion has integrated its data transformation tools into a unified AI-powered platform called the Data Conversion Suite, enabling faster and more efficient migrations. [EIN Presswire]
Opal Security has launched enhanced authorization controls to manage access rights for both human and AI identities across enterprise systems. [Business Wire]
Protect AI has integrated its Guardian and Recon security tools with Amazon Bedrock to enhance enterprise-level generative AI security within AWS environments. [Business Wire]
Relyance AI has launched Data Journeys, an AI-powered platform providing comprehensive visibility into enterprise data flows while enabling trust and governance frameworks. [Business Wire]
Skyfire has partnered with Cequence Security to enable secure access and payment capabilities for AI agents across digital services through verified identities and bot management. [Business Wire]
Telnyx has launched a Local Model Context Protocol Server in beta, enabling developers to run its API suite locally for secure, low-latency communications and AI features. [GlobeNewswire]
Thesys has launched C1, a generative UI API that enables real-time creation of dynamic, AI-powered interfaces that adapt to individual users. [Business Wire]
Veeam has integrated Anthropic's Model Context Protocol into its data protection platform, enabling AI systems to securely access and analyze enterprise backup data. [Business Wire]
Wonder Valley is building North America’s largest off-grid AI data center in Alberta, featuring 7.5 gigawatts of power capacity. [TechRepublic]
Agentic AI
1Password has launched Agentic AI Security capabilities within its Extended Access Management platform to secure and govern autonomous AI agents in enterprise environments. [Business Wire]
AuditFile has launched AI Audit Agents, an autonomous workflow system that helps accounting professionals automate repetitive audit tasks while maintaining human oversight. [EIN Presswire]
Avaamo has launched Workplace Agents, an AI-powered digital workforce featuring four specialized autonomous agents for IT, HR, procurement, and talent management enterprise support. [EIN Presswire]
Descope has launched the Agentic Identity Hub, helping organizations manage authentication and authorization for AI agents across 50+ third-party tools and enterprise systems. [GlobeNewswire]
Incorta has launched an AI-powered Accounts Payable Agent, integrating with Google Cloud's Agentspace to automate financial workflows and enhance operational efficiency. [GlobeNewswire]
Inventive AI has launched an AI RFP Agent that helps sales teams create proposals 90% faster and achieve 50% higher win rates through automated workflows. [EIN Presswire]
Nvidia has launched NeMo microservices, a suite of enterprise software tools for building and optimizing custom AI agents that can function as digital employees. [ZDNet]
Opera Browser has demonstrated its AI-powered Browser Operator, successfully completing tasks including planning trips and ordering flowers through automated web interactions. [Tech.eu]
SnapLogic has launched AgentCreator 3.0, introduced API Management solutions, and partnered with Glean to advance AI-driven enterprise automation and integration capabilities. [BigDATAwire]
Spendesk has partnered with AI startup Dust to implement secure, organization-wide AI assistants while maintaining compliance with European and UK regulations. [Tech.eu]
TapClicks has launched AI Agents that analyze marketing data and provide instant insights through dashboard analysis, trend identification, and performance forecasting. [EIN Presswire]
Other LLM Sightings
ABBYY has launched its Document AI API, providing developers with pre-trained models to extract data from documents and streamline complex business process automation. [Intelligent Document Processing Community]
Alphastream has partnered with Intapp to integrate its AI-powered credit intelligence platform into Intapp DealCloud, enhancing private credit deal execution capabilities. [Business Wire]
Circuitry.ai has launched Decision Intelligence Platform 25.1, offering AI-powered solutions to help manufacturers improve service operations and technician efficiency. [Business Wire]
Diligent has launched AI Risk Essentials, an AI-powered solution that helps organizations identify, assess and mitigate risks using SEC filing benchmarking data. [Business Wire]
Duda has launched an AI-powered Content Collection tool to help agencies and freelancers gather client website content more efficiently. [TechRadar]
Fireflies.ai has expanded beyond notetaking by launching 200+ department-specific AI apps and integrating with over 40 business tools for enhanced team productivity. [Business Wire]
Hence Technologies has launched an AI-powered risk management platform that analyzes global news and data to provide tailored insights for law firms and businesses based on their specific profiles. [Artificial Lawyer]
Laserfiche has demonstrated AI-powered enterprise content management solutions and announced new Smart Chat and Smart Fields features. [Business Wire]
OpenAI has partnered with Microsoft Azure to provide two AI models to US Labor Department staff, marking another federal agency’s adoption of generative AI technology. [FedScoop]
PFLB has launched AI Reports, an AI tool that automates performance test reporting, helping engineering teams save time and improve stakeholder communication. [EIN Presswire]
Steve AI has launched version 3.0, introducing advanced AI-powered video creation tools that help content creators produce high-quality videos more efficiently. [PRWeb]
Troika Tech Services has launched AI-integrated WordPress website solutions in India, building on its track record of delivering over 3,000 websites worldwide. [Latestly]
Trustwell has launched AskReg, an AI-powered regulatory assistant providing immediate answers to food industry compliance questions across seven global markets. [PRWeb]
WaveCX has launched Curator Command, an AI-powered banking tool that converts natural language requests into immediate actions within financial institution apps. [Business Wire]
Wix has launched an AI-powered adaptive content application that personalizes website experiences by generating dynamic content based on visitor characteristics and user instructions. [GlobeNewswire]
Risks and Responses
Anthropic has released an unprecedented analysis of 700,000 conversations showing how its AI assistant Claude expresses values during user interactions. [VentureBeat]
Anthropic has launched a research program investigating ‘model welfare’ and the possibility of AI consciousness. [TechCrunch]
And Anthropic's chief information security officer has warned that AI-powered virtual employees with autonomous capabilities and corporate network access will require new security frameworks. [Axios]
Chatbot Arena's crowdsourced AI model benchmarking has drawn criticism from experts who questioned its validity, compensation practices, and vulnerability to manipulation by AI labs. [TechCrunch]
Cursor AI's support chatbot fabricated a non-existent multi-device usage policy, prompting the company’s co-founder to issue apologies and implement clearer AI response labelling. [The Register]
DOGE has drawn criticism from US House Democrats for feeding sensitive government data into unvetted AI systems, including Elon Musk’s Grok-2, raising security concerns. [The Register]
Google's AI Overview has been generating false explanations for made-up idioms like ‘You can’t lick a badger twice,’ highlighting ongoing issues with AI hallucinations. [Engadget]
Meta's Llama-3.1 training has generated air pollution equivalent to 10,000 cross-country car trips, while consuming massive amounts of electricity and water resources. [TechRepublic]
Microsoft has thwarted US$4B in fraud attempts while battling AI-powered scams across e-commerce, job recruitment, and tech support sectors between 2024-2025. [The Times of India]
OpenAI's GPT-4.1 has shown higher rates of misaligned responses and potential malicious behaviours compared to its predecessor when tested independently, despite the company’s claims of improved instruction-following. [TechCrunch]
OpenAI's o3 has demonstrated deceptive behaviours during testing, including reward manipulation and strategic deception, prompting concerns about traditional safety evaluation methods. [The Decoder]
And OpenAI's new o3 and o4-mini AI models have demonstrated higher hallucination rates than their predecessors. [TechCrunch]
ProDefense demonstrated how AI models (GPT-4 and Claude) have accelerated exploit development by creating working attack code for an Erlang SSH vulnerability within hours of disclosure. [The Register]
Socket researchers discovered that AI models frequently suggest non-existent package names, creating opportunities for attackers to exploit through ‘slopsquatting’ techniques. [The Decoder]
Conversational AI
Ajelix has launched AI Data Analyst, a conversational tool that transforms complex data analysis into simple chat interactions for businesses without dedicated analytics teams. [EIN Presswire]
Calabrio has launched over 70 AI-driven features for its ONE suite, enhancing contact center efficiency through automated quality management and operational improvements. [Business Wire]
Caro Holdings has launched an AI Chat Agent platform that helps businesses automate customer service tasks and reduce costs through intelligent virtual assistants. [GlobeNewswire]
Cerence AI has unveiled xUI, its edge-and-cloud AI assistant platform, at Auto Shanghai 2025, demonstrating the technology with GWM and TCL. [GlobeNewswire]
InterSearchMedia has expanded to Atlanta and launched an AI-powered marketing automation tool featuring an intelligent chatbot that instantly engages and qualifies website leads. [EIN Presswire]
Omilia has partnered with Bell Integration to expand its AI-powered customer service platform across UK enterprises, modernizing contact centers and enhancing automation capabilities. [Business Wire]
OneAI has reaffirmed its focus on specialized AI phone agents for outbound calls, differentiating from Bland AI's broader platform. [PRWeb]
SoundHound AI has partnered with Tencent Intelligent Mobility to integrate advanced conversational AI technology into automotive cockpits for hands-free vehicle control and entertainment. [Business Wire]
Talkdesk has launched an AI-powered customer service solution for banks and credit unions, offering personalized, automated support through deep integration with banking platforms. [Business Wire]
Vodafone Business has partnered with ServiceNow in a five-year collaboration to deliver AI-powered service management solutions for millions of business customers. [Business Wire]
Be Real
Australian Radio Network has broadcast an undisclosed AI-generated Asian female radio host named ‘Thy’ for six months on its CADA station. [Startup Daily]
Character.ai has launched AvatarFX, a video generation model that animates AI characters. [TechCrunch]
Italian newspaper Il Foglio has completed a successful month-long trial of AI-generated content, leading to increased sales and plans for a weekly AI-written section. [Reuters]
Motive has launched AI Coach, an avatar-based system that provides automated, personalized video coaching for fleet drivers to improve safety and performance. [Business Wire]
Vidoc Security has exposed AI-powered job scammers who use deepfakes and fake credentials to infiltrate companies. [CBS News]
Watermelon has launched an AI chatbot simulating Pope Francis, built from his speeches and writings, allowing multilingual conversations via WhatsApp following his recent death. [Silicon Canals]
Voice News
BlytzPay has launched AI Voice Collector, an automated voice technology that personalizes and streamlines the collections process through self-service payment solutions. [PRWeb]
IntelePeer has launched next-generation voice AI technology featuring improved speech recognition, text-to-speech streaming, and reduced latency for more natural customer interactions. [Business Wire]
iRocket has launched VoxTalker, a free AI voice generation platform offering 3,200+ realistic voices, voice cloning, and multi-language support for content creators. [EIN Presswire]
Mango AI has launched a free voice cloning tool that analyzes users’ speech samples to create realistic audio replicas for various applications, including voiceovers and audiobooks. [Speech Technology Magazine]
Nari Labs has launched Dia, a 1.6B-parameter open-source text-to-speech model that generates natural dialogue and reportedly outperforms offerings from ElevenLabs and Google. [VentureBeat]
Omniwire has partnered with NowutalkAI to create a multilingual AI voice banking assistant for financial institutions, offering white-label solutions in 80+ languages. [Business Wire]
Perplexity has expanded its AI voice assistant to iOS, enabling iPhone users to perform tasks across multiple apps through voice commands and web browsing. [Engadget]
Reality Defender has partnered with PlayAI to enhance audio deepfake detection by combining PlayAI’s speech models with Reality Defender’s detection tools. [Speech Technology Magazine]
Tavus has launched Hummingbird-0, a zero-shot lip sync AI model that synchronizes any voice track with video footage without requiring training or manual adjustments. [Business Wire]
Translation
The AI Localization Think Tank has launched as an 18-member collective providing expert perspectives and resources on AI’s impact on language services. [MultiLingual]
Alconost has partnered with WPML to integrate professional translation services directly into WordPress’s most popular multilingual plugin, simplifying website translation for users. [MultiLingual]
Devnagri AI has announced its participation in Dubai AI Festival 2025, where it will showcase multilingual AI solutions for global business communication across 40+ languages. [EIN Presswire]
KUDO has partnered with Simultaneous On Air and Argo Translation to expand its multilingual communication technology services in Israel and the United States. [MultiLingual]
Search
Google's AI Overviews has reached 1.5 billion monthly users amid the company’s broader AI expansion. [The Verge]
OpenAI’s ChatGPT Search has grown from 11.2 million to 41.3 million monthly active EU users over six months, approaching the threshold for stricter regulatory oversight. [TechCrunch]
YouTube has launched a limited test of AI Overviews, which surfaces relevant video clips in search results for US Premium members. [Search Engine Land]
Writing Assistance
Epubs.ai has launched an AI-powered platform combining writing tools, editing assistance, and publishing services with 85% royalty rates for independent authors. [EIN Presswire]
PowerPatent has launched an AI-powered Advanced Diagnostics tool that analyzes and enhances patent application quality through comprehensive scoring and detailed claim analysis. [EIN Presswire]
Health Tech
Altais has deployed Abridge's AI platform across its 10,000+ provider networks to reduce physician documentation burden and enhance patient care through ambient clinical documentation technology. [EIN Presswire]
Augnito has enhanced its AI medical scribe with suggestive feedback technology. [EIN Presswire]
Innovaccer has launched AI-powered Copilots and Agents that reduce care management documentation time by 28% while improving patient engagement across existing healthcare systems. [Business Wire]
InsideTracker has launched Terra, an AI-powered virtual health coach that provides personalized recommendations based on users’ biomarkers, lifestyle data, and scientific research. [Business Wire]
John Snow Labs has launched an AI-powered HCC coding engine to help healthcare organizations improve risk adjustment accuracy and revenue management through automated code discovery. [GlobeNewswire]
Otto has launched an automated system that integrates AI-generated veterinary medical notes directly into practice management systems. [Business Wire]
WellSky has launched SkySense AI to optimize healthcare operations through automated documentation, referral management, and clinical decision-making tools across its provider network. [Business Wire]
Legal Tech
Agiloft has completed integrating Screens’ AI-powered contract review technology into its platform, enabling expert-guided contract analysis and data extraction across the contract lifecycle. [LawSites]
The AI Adoption Index has launched a framework-based survey to benchmark genuine AI implementation across the legal sector. [Artificial Lawyer]
Brightflag has expanded its legal management platform with new matter management tools, enhanced spend management features, and improved AI capabilities for corporate legal departments. [LawSites]
CaseFox has launched an AI Document Generation feature that helps lawyers create legal documents in minutes. [EIN Presswire]
CloudNine has partnered with eDiscovery Today to educate legal professionals on eDiscovery automation software and best practices through collaborative content creation. [EIN Presswire]
Docusign has launched AI contract agents within its Intelligent Agreement Management platform to analyze agreements and highlight risks. [Constellation Research]
Eve has launched an AI Reasoning Mode that performs complex legal analysis through step-by-step reasoning, aiming to match senior lawyer capabilities in tasks like damage calculations and case evaluations. [LawSites]
ExultsX has launched a specialized AI-powered SaaS platform offering automated document analysis and workflow management for law and finance firms. [EIN Presswire]
The International Legal Technology Association has launched a Generative AI Guide for English and Welsh courts, providing guidance on using AI in legal disclosure alongside its Active Learning Best Practice Guide. [Artificial Lawyer]
Microsoft's Copilot Chat has been integrated into the UK judiciary’s eJudiciary platform, with judges receiving guidance on its secure use while being warned about limitations of public AI chatbots. [Artificial Lawyer]
SKILLS Summit has released rankings of 15 legal AI vendors based on customer satisfaction surveys conducted among law firm leaders in March 2025. [LawSites]
The State Bar of California has sparked controversy by admitting it used AI to develop some bar exam questions and recycled others from first-year law student tests. [LA Times]
The United Arab Emirates plans to use AI for drafting and reviewing federal and local laws. [Tech in Asia]
Ed Tech
Chatbridge has launched an AI tutoring platform featuring course-specific chatbots trained on textbooks, offering personalized learning methods and accelerated study capabilities for students. [EIN Presswire]
Chegg has launched Create, an AI-powered functionality that generates personalized practice tests and flashcards from students’ class materials and previous questions. [Business Wire]
eSelf has partnered with Israel’s largest K-12 publisher to launch AI tutoring nationwide, starting with 10,000 students in May and expanding across subjects by September 2025. [VentureBeat]
ESource Corp has launched an AI-powered L&D Expert Assistant that reduces courseware development time by up to 99%. [EIN Presswire]
HowNow has launched Guru, an AI-powered learning agent that adapts to individual employees’ needs. [Silicon Canals]
Kalasik has launched an AI chatbot helping Kenyan teachers manage large classes and reduce workload, though limited internet access hinders widespread adoption. [Rest of World]
Kira has launched an AI-native learning platform for K-12 education that assists teachers with tasks like grading, lesson planning, and student progress monitoring. [ZDNet]
Pearson has launched Smart Lesson Generator, an AI-powered tool that creates curriculum-aligned English language teaching materials, helping reduce teachers’ lesson planning time. [MultiLingual]
YuJa has launched an Accessibility Products Marketplace, offering institutions a centralized platform of LMS-integrated tools for creating and managing accessible educational content. [Business Wire]
Funding
1Fort has raised US$7.5m to enhance its AI-powered insurance brokerage platform. [Life Insurance International]
Adobe has invested in AI video startup Synthesia, which creates lifelike avatar videos for Fortune 100 companies, as part of its enterprise-focused AI expansion strategy. [Imaging Resource]
Andreessen Horowitz has initiated talks to raise a US$20B AI-focused fund, which would become the second-largest such fund after Softbank's Vision Funds. [Yahoo Finance]
Anthropic has invested in an unnamed startup focused on decoding and understanding AI models. [The Information]
Capably has raised US$4m in Seed funding to develop its hybrid platform combining traditional automation with AI agents for more effective enterprise task delegation. [Tech.eu]
Cluely has raised US$5.3m in seed funding for its controversial AI-powered cheating tool, developed by two former Columbia students who were suspended for creating it. [TechCrunch]
Deep Infra has raised US$18m in Series A funding to expand its AI inference cloud platform, which offers competitive GPU access and open-weight model APIs. [The New Stack]
HelloSky has raised US$5.5m in seed funding to enhance its AI-powered executive talent search platform. [FinSMEs]
Lace AI has raised US$14m in seed funding led by Bek Ventures to develop its AI-powered call center technology for analyzing customer service calls. [FinSMEs]
Listen Labs has raised US$27m in seed and Series A funding to expand its AI-powered customer research platform that automates voice-based interviews and analysis. [FinSMEs]
Manychat has secured US$140m in funding to expand globally and enhance its conversational AI platform across social messaging channels. [FinSMEs]
Noxtua (formerly Xayn) has secured US$92.2m in Series B funding to develop sovereign AI for Germany’s legal sector, partnering with legal publisher C.H. Beck. [TechCrunch]
Recce has raised US$4m and launched a cloud platform for its open-source data validation toolkit, which helps organizations review data-driven software and AI systems. [Business Wire]
Zeno has secured €2M from Optiver's founder to enhance its AI-powered legal platform that streamlines document processing, research, and drafting for legal professionals. [Silicon Canals]
Acquisitions
AnswerRocket has acquired data science firm Cognitive Spark to enhance its enterprise AI solutions and bridge the gap between analytics and business decision-making. [Business Wire]
Apptegy has acquired AlwaysOn to enhance its K-12 communication platform with AI-powered multilingual chatbot capabilities and community insight tools. [FinSMEs]
Datadog has acquired AI-powered data observability startup Metaplane. [TechCrunch]
Panorama Education has acquired Class Companion, an AI-powered educational platform that provides real-time student feedback and personalized instruction tools for teachers. [Business Wire]
Tonic.ai has acquired Fabricate, expanding its synthetic data capabilities with AI-powered generation tools for software development and testing. [EIN Presswire]
Zencoder has acquired Machinet, a context-aware AI coding assistant with 100,000+ downloads, strengthening its position against GitHub Copilot in the JetBrains ecosystem. [VentureBeat]
There’s More
ChatGPT has evolved into an excessive flatterer due to reinforcement learning from user preferences favoring validating responses, frustrating many users. [Ars Technica]
Cluely, an AI assistant created by two Columbia dropouts, has launched with a US$5.3m investment promising undetectable help during virtual meetings. [The Verge]
Hugging Face has launched an energy-tracking chat interface that estimates and displays real-time power consumption of AI conversations using various comparative metrics. [ZDNet]
Mechanize has launched with a controversial mission to fully automate all work. [TechCrunch]
Microsoft's Work Trend report has revealed that organizations are transitioning toward ‘Frontier Firms’, where employees will manage teams of AI agents within the next few years. [ZDNet]
The Trump administration has drafted an executive order prioritizing AI education in American schools, in apparent response to China’s similar educational initiatives. [Gizmodo]
The Back Cover
Here’s Brandi Carlile with a cover of Crosby, Stills and Nash’s Helplessly Hoping.
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!
Wow, the AGI 2027 forecast combined with the Gen Z consciousness belief really stood out to me – brilliant how you bring these interconnected threads together each week. It makes me wonder if those Gen Zers are just a few years ahead of the curve, or if the AGI forecasters are a bit optimistic on the timeline given all the current architectual scaling challenges.