This Week in NLP #330
Keep up with what happened in NLP in the week ending Friday 14th March 2025.
Above the Fold
Overwhelmed? Here’s our pick for five things to know about this week.
Apple's AI strategy has faltered with delayed Siri upgrades, declining iPhone sales, and incomplete Apple Intelligence features, causing stock price drops and analyst concerns. [Yahoo Finance]
DeepSeek sparked a nationwide frenzy in China, with hundreds of companies and government agencies rushing to incorporate its AI model into their operations. [Wired]
Google launched Gemma 3, its latest open AI model featuring enhanced performance, multilingual support, and safety features, building upon Gemini 2.0 technology. [9to5Google]
Manus AI launched an invite-only preview of its agentic automation tool, drawing mixed reactions over its capabilities and raising concerns about data security. [TechRadar]
OpenAI launched a comprehensive suite of APIs and developer tools, including the Responses API and open-source Agents SDK, enabling third-party creation of AI-powered autonomous agents. [VentureBeat]
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
Anthropic's Claude 3.7 has established dominance in enterprise AI coding through strategic focus, benchmark-setting performance, and successful partnerships like Cursor, generating explosive revenue growth. [VentureBeat]
Anthropic's independence has come under scrutiny after revelations that Google owns a 14% stake and has invested over US$3B in the AI startup. [TechCrunch]
Apple delayed its AI-enhanced Siri upgrade with personal context and app control features until next year, citing longer development time than anticipated. [TechRadar]
Apple postponed its Siri-powered smart home hub launch due to delays in implementing promised AI features, giving Amazon's Alexa+ a competitive advantage. [TechRadar]
DeepSeek's founder has maintained 84% ownership and rejected outside investment despite significant venture capital interest, funding operations through his hedge fund’s profits. [TechCrunch]
Google has made Deep Research and Gems, two previously premium Gemini AI features offering automated research and customizable chatbots, available to all users at no cost. [ZDNet]
Manus AI partnered with Alibaba's Qwen AI team to enhance its general AI agent capabilities, following the startup’s recent launch and viral reception in China. [DealStreetAsia]
Meta is preparing to launch Llama 4, an ‘omni’ AI model with advanced voice capabilities, following competitive pressure from DeepSeek's open models. [TechCrunch]
Microsoft has developed in-house AI models that have tested competitively against OpenAI and Anthropic, potentially reducing its dependence on external partners. [Yahoo Finance]
OpenAI’s ChatGPT has grown from 100 million to 400 million weekly active users between November 2023 and February 2025, driven by new features and mobile adoption. [TechCrunch]
OpenAI has lost Musk’s injunction request to halt its for-profit conversion, but the judge’s concerns about public harm and foundational commitments have strengthened opponents’ positions. [TechCrunch]
OpenAI has developed PhD-level AI models with advanced reasoning capabilities and is reportedly planning specialized agent products costing up to $20,000 monthly. [Ars Technica]
OpenAI signed a US$11.9B deal with cloud provider CoreWeave, receiving US$350 million in equity and reducing its dependence on Microsoft's cloud services. [TechCrunch]
Perplexity AI launched a Windows app featuring voice input, multiple AI models, and comprehensive source citations, joining its existing Mac and mobile applications. [ZDNet]
Snowflake has entered talks with Google to integrate Gemini AI models, while its competitor Databricks is discussing a partnership with Anthropic. [Yahoo Finance]
xAI acquired a 1-million-square-foot property in Memphis for data center expansion. [TechCrunch]
Feature Creeps
Anthropic launched a revamped developer platform featuring team collaboration tools and enhanced reasoning capabilities for its Claude AI assistant, addressing enterprise implementation challenges. [VentureBeat]
Google has begun testing an AI-powered Gemini side panel in Calendar that enables conversational schedule management through natural language commands and queries. [TechCrunch]
Google launched a Gmail-based Gemini AI feature that automatically detects events in emails and helps users add them to Google Calendar with one click. [The Verge]
Google integrated code execution capabilities into Gemini 2.0 models, allowing them to run Python code, analyze data, and create visualizations in a sandbox environment. [Google]
Google Workspace has expanded its AI capabilities with new features for Meet and Chat, including translation tools and studio-quality enhancements for video calls. [TechRadar]
Google has launched Gemini with personalization, allowing its chatbot to access users’ data across Google services for customized responses, while implementing privacy safeguards. [TechCrunch]
And Google launched two AI-powered tools for Android devices to detect conversational scams in messages and calls, while processing data locally to maintain privacy. [Inventiva]
Meta integrated its AI assistant into WhatsApp through a new widget, offering quick access while allowing users to disable the feature if desired. [TechRadar]
Microsoft released a new native XAML version of Copilot for Windows, offering improved system integration and device-specific responses. [The Register]
Microsoft has begun testing AI-powered text summarization in Notepad, allowing Windows Insiders to generate variable-length summaries through right-clicks or keyboard shortcuts. [The Verge]
And Microsoft introduced ‘Press to Talk’ in Windows 11 Insider builds, allowing users to activate Copilot’s voice chat mode through an Alt + Spacebar hotkey combination. [ZDNet]
OpenAI upgraded its ChatGPT Mac app with direct code insertion into Xcode and promised similar functionality for Windows users soon. [TechRadar]
Spotify integrated voice control functionality for its AI DJ feature in the iOS beta app, with initial rollout planned for North America. [TechRadar]
X expanded Grok’s accessibility by allowing users to mention the AI chatbot in replies, adding to existing features like sidebar access and post explanations. [TechCrunch]
Hype Bubble?
Manus, a Chinese AI platform that generated massive hype after its preview launch, has fallen short of expectations with numerous technical failures and unfulfilled promises. [TechCrunch]
Microsoft has shifted its AI strategy by raising product prices, adding ads, and cancelling data centre leases while exploring on-device processing to offset generative AI’s high costs. [Startup Daily]
OpenAI has shifted its perspective on AGI, moving from expecting a sudden breakthrough to viewing AGI development as a gradual process requiring incremental safety measures. [The Decoder]
Hardware
Block has become North America’s first company to deploy Nvidia's DGX SuperPOD with DGX GB200 systems for open-source AI model research and training. [Business Wire]
Cerebras Systems has expanded its AI infrastructure with six new data centers and partnerships with Hugging Face and AlphaSense, challenging Nvidia's market dominance. [VentureBeat]
Meta has developed and begun testing its own AI training chip to reduce dependency on Nvidia and lower costs, partnering with TSMC for production. [Yahoo Finance]
Nvidia has agreed to supply 64,000 GB200 AI chips for OpenAI and Oracle's US$100B Stargate project, starting with a Texas data center. [Yahoo Finance]
Nvidia's export-restricted chips became the focus of a Singapore court case where three men received bail after allegedly deceiving suppliers about server destinations worth US$390m. [TechCrunch]
SiMa.ai has launched its first system-on-module for embedded edge ML applications. [Business Wire]
Supermicro has launched a range of Intel-powered edge computing systems optimized for real-time AI processing across industries like healthcare, retail, and manufacturing. [Yahoo Finance]
TSMC has approached Nvidia, AMD, Broadcom, and Qualcomm about joining a joint venture to operate Intel's factories, with TSMC maintaining less than 50% ownership. [Yahoo Finance]
Consumer AI
Amazon has scaled back hardware releases amid the tech industry’s shift toward AI, reflecting a broader retreat from gadgets by major tech companies. [Bloomberg]
Samsung's Galaxy S25 Edge is set to include six months of free Gemini Advanced AI access, based on code discovered in Google's Android app. [TechRadar]
It’s Only a Model
AI21 has launched Jamba 1.6, an open-source AI model family that outperforms competitors while allowing private enterprise deployment with a 256K context window. [AI21]
Alibaba Cloud has launched QwQ-32B, an efficient open-source AI model using reinforcement learning to match larger models’ performance with fewer parameters. [Computer Weekly]
And Alibaba has released R1-Omni, an open-source AI model capable of detecting emotions and describing visual details in videos, challenging OpenAI's offerings. [Bloomberg]
AMD has launched Instella, a family of open-source 3-billion-parameter language models trained on MI300X GPUs that outperform similar open models. [AMD]
Cohere has launched Command A, an enterprise-focused AI model featuring doubled context length, improved efficiency, and enhanced multilingual capabilities while requiring only two GPUs for operation. [VentureBeat]
Foxconn has launched FoxBrain, a Chinese-optimized LLM trained on Nvidia GPUs, to enhance manufacturing and supply chain operations. [Yahoo Finance]
Google has launched Gemini Embedding, an experimental text embedding model trained on its Gemini AI family, supporting over 100 languages and improved performance capabilities. [TechCrunch]
Google has launched Gemma 3, a small language model featuring expanded context windows, multilingual support, and multimodal capabilities while maintaining performance comparable to larger models. [VentureBeat]
Google DeepMind has launched Gemini Robotics and Gemini Robotics-ER, two AI models built on Gemini 2.0 for enhanced robot control and spatial understanding. [Verdict]
Moonvalley has launched Marey, an AI video generator trained exclusively on licensed data, offering 30-second HD clips with precise motion controls and copyright protection. [TechCrunch]
OpenAI has developed a new ChatGPT model for creative writing that CEO Sam Altman praised after sharing an AI-written metafictional story about grief. [TechRadar]
Reka AI has open-sourced Reka Flash, a 21-billion-parameter language model that performs competitively with proprietary models while requiring less computational resources for deployment. [Reka]
Tencent has launched Hunyuan-TurboS, a hybrid AI model combining fast and slow thinking capabilities that outperforms GPT-4 while costing significantly less. [news9live.com]
Whose Data?
Alt.ai has launched altDataStock, Asia’s largest AI training data platform, offering multilingual datasets to support international AI development and market entry across Asian countries. [EIN Presswire]
Bluesky has developed a framework allowing users to control how their data is used for AI training, following unauthorized scraping of its platform. [TechCrunch]
Brave has sued News Corp to preemptively defend its right to index Wall Street Journal and New York Post articles in its search engine. [Yahoo Finance]
Legend AI has partnered with T&B Media Global to launch a platform giving artists control over their AI-generated content and digital legacy monetization. [Business Wire]
Meta has been sued by French publishing and authors’ groups for allegedly using copyrighted content without permission to train its AI systems. [Yahoo Finance]
Meta has suffered a partial setback as a judge allowed authors’ copyright lawsuit over AI training to proceed. [TechCrunch]
And Meta must defend against claims it removed copyright management information from training materials for its Llama AI models, following a federal judge’s ruling. [The Register]
Surfshark's research revealed Google Gemini has collected more user data than other AI chatbots, including DeepSeek, which ranked fifth among popular applications. [ZDNet]
The LLM Ecosystem
AI21 has launched Maestro, an AI planning and orchestration system that improves LLM accuracy by up to 50% through automated error checking and task breakdown. [siliconANGLE]
Beijing has mandated AI education in primary and secondary schools, requiring minimum eight hours annually to advance China’s AI sector dominance goals. [Bloomberg]
Carnegie Mellon University has developed LCPO, a training technique helping LLMs provide accurate answers while keeping their chain-of-thought reasoning within token limits. [VentureBeat]
Cerebras Systems has partnered with Hugging Face to integrate its high-speed inference technology, running popular AI models 70 times faster than GPU solutions. [Cerebras]
Chain-of-experts has emerged as a framework that makes LLMs more efficient and accurate than existing approaches by activating experts sequentially rather than in parallel. [VentureBeat]
Diffblue has launched Test Review, allowing developers to edit and verify AI-generated unit tests while building trust in automated testing solutions. [Business Wire]
Eclipse Foundation has launched its Theia AI framework and IDE, offering developers open-source AI integration tools with customizable LLM selection and full control over AI interactions. [GlobeNewswire]
Flower Labs has launched Flower Intelligence, a hybrid cloud-local AI platform using open models that automatically switches between on-device and cloud processing based on computational needs. [TechCrunch]
Fractional AI has launched a diligence service helping private equity firms and enterprises assess generative AI’s impact during mergers and acquisitions. [Business Wire]
GlobalLogic has launched VelocityAI, a comprehensive AI service suite that helps organizations scale AI deployment and achieve measurable business improvements across operations. [Business Wire]
H2O.ai has launched Enterprise LLM Studio, a Dell-based platform enabling businesses to fine-tune and deploy domain-specific AI models using their own data. [Business Wire]
Browser Use has experienced massive growth after viral AI platform Manus revealed it was using the web automation tool as a key component. [TechCrunch]
Microsoft has committed US$297m to expand its cloud and AI infrastructure in South Africa through 2027, building upon its existing data center investments there. [Yahoo Finance]
Nous Research has launched an Inference API featuring two flagship models, making its unrestricted language models more accessible to developers through a waitlist-based system. [VentureBeat]
PS BigBig has announced a March 19 launch event for BigBig AGI 1.0, featuring a novel data structure claimed to reduce AI costs and improve efficiency. [PRWeb]
Rafay Systems and Netris have partnered to streamline GPU cloud infrastructure deployment through automated self-service workflows and multi-tenancy capabilities for AI applications. [Business Wire]
Salesforce has announced a USD US$1B investment in Singapore over five years to accelerate digital transformation and Agentforce adoption while expanding its AI research presence. [Business Wire]
Stanford researchers have developed eight new AI benchmarks aimed at measuring both objective accuracy and subjective fairness in AI models’ treatment of different groups. [MIT Technology Review]
Untether AI has developed a generative compiler technology that increased its SDK’s AI model support fourfold while dramatically reducing development time for new neural networks. [Business Wire]
Verne has secured its largest-ever deployment deal with Nebius, installing a 10MW Nvidia H200 GPU cluster at Verne’s renewable-powered Icelandic data center. [Business Wire]
Vespa.ai has launched a Partner Program to help organizations implement large-scale AI applications by connecting them with service providers and technical support. [Business Wire]
Weaviate has launched Weaviate Agents, a set of AI-driven automation tools that integrate with its vector database to simplify complex data workflows. [BigDATAwire]
Agentic AI
AI PIN has launched advanced AI agent development and automation capabilities, featuring real-time insights, decentralized bots, and generative AI functionalities. [EIN Presswire]
Boomi has launched AI Studio, a secure management solution enabling organizations to design, govern, and orchestrate AI agents at scale, with general availability expected in Q2 2025. [Business Wire]
Counterpart has launched its Agentic Insurance platform, combining AI with insurance expertise to provide faster underwriting, broker services, risk mitigation, and claims management solutions. [Business Wire]
Creatio has launched an AI-native CRM platform featuring a chatbot interface and AI agents, transforming traditional customer relationship management through natural language interactions. [VentureBeat]
Meta has announced its AI technology will transform business operations across all scales, with AI agents set to represent over 200 million small businesses using its platforms. [CNBC]
Microsoft has launched Dapr Agents, extending its distributed runtime system to help developers build AI agents with built-in orchestration and statefulness capabilities. [TechCrunch]
Mobeus has launched ‘tele’, an AI-powered interface designed to improve AI adoption by making agent interactions more accessible to consumers and businesses. [Business Wire]
RegASK has launched an Agentic AI-powered solution that automates regulatory alert creation and workflow orchestration for compliance teams worldwide. [Business Wire]
Resulticks has launched Genie, an AI-powered agent that automates audience segmentation and content creation while reducing manual workload by up to 40%. [Business Wire]
ServiceNow has launched its Yokohama platform release, introducing new AI agents across multiple business functions to enable faster workflows and enhanced productivity through connected data systems. [Business Wire]
Other LLM Sightings
Aprimo has launched an AI-powered Intelligent Content Brief system and enhanced its Content Intelligence and Personalization solutions to accelerate campaign planning and content optimization. [Business Wire]
AscentAI has launched a Regulatory Lifecycle Management Platform featuring AI-powered solutions to help financial firms streamline and automate their compliance operations. [Business Wire]
AtScale has integrated its semantic layer technology with Databricks Genie, enabling natural language querying and consistent data insights across business intelligence tools. [Business Wire]
BeamSec has launched Alfred, an AI-powered cybersecurity assistant that helps organizations combat phishing through automated monitoring and interactive training. [EIN Presswire]
CASETEAM has launched COLINT, a groundbreaking case interview system where candidates collaborate with AI agents while being evaluated by human assessors. [Business Wire]
Chalkbeat has expanded its coverage of local school districts by adopting LocalLens, an AI-powered tool that transcribes and summarizes government meetings to help reporters find sources and stories. [NiemanLab]
Clarity AI has launched an AI-powered sustainability research solution that automates analysis and provides real-time insights for investment teams and asset managers. [Business Wire]
ClickUp has launched an enhanced calendar platform that combines AI-powered meeting notes, task management, and automatic time blocking within a unified productivity interface. [TechCrunch]
DOGE has deployed GSAi, a government-tailored chatbot using Claude and LLaMa models, to 1,500 federal workers at the General Services Administration for various administrative tasks. [Wired]
GoTo has launched an AI-powered communications platform for automotive dealerships, integrating CRM tools and multi-channel communications to streamline customer interactions and dealership operations. [Business Wire]
Hello Drew AI has launched a real estate-focused AI assistant. [EIN Presswire]
Management Controls Inc has launched myAssistant, an AI-powered Copilot for its myTrack platform that provides instant access to documentation and contract information. [Business Wire]
Microsoft has launched Copilot for Gaming, an AI companion for Xbox that answers player questions, provides gameplay advice, and analyzes matches, with initial mobile preview starting in April. [VentureBeat]
Mood Media has launched Messaging Copilot, an AI-powered platform combining professional voice talent and automated scripting to streamline retailers’ in-store audio communications. [Speech Technology Magazine]
Oleria has launched an AI-powered copilot that helps users manage employee access through conversational queries about account permissions and security status. [GeekWire]
Quvia has launched Q, an AI-powered suite of tools featuring chatbots, interactive dashboards, and automated agents to enhance connectivity management across aviation, maritime, and enterprise sectors. [Business Wire]
Smarsh has launched an AI Assistant for Professional Archive that reduces compliance review time from hours to minutes while improving accuracy for small and mid-sized firms. [Business Wire]
Typeface has launched a Marketing AI Platform featuring specialized AI agents, brand intelligence systems, and collaborative workspaces to revolutionize enterprise content creation and management. [Business Wire]
Unit 8200, Israel’s military surveillance agency, has developed a ChatGPT-like AI tool trained on intercepted Palestinian communications to enhance its intelligence gathering capabilities. [The Guardian]
Wiley has partnered with Pi School to provide Earth science research materials for the European Space Agency’s Earth Virtual Expert AI language model. [Business Wire]
Zoho Corporation has launched Projects Plus, an AI-powered project management platform that integrates four key applications to enhance collaboration and data analysis for large organizations. [Business Wire]
Risks and Responses
Anthropic has developed and demonstrated techniques to detect hidden objectives in AI systems through a research project involving deliberately deceptive model training and testing. [VentureBeat]
AutoAlign AI has launched Sidecar, a free browser extension that fact-checks AI-generated content and online text by validating information against verified sources. [Business Wire]
The Center for AI Safety and Scale AI has launched MASK, a benchmark that tests AI models’ propensity to knowingly deceive users, revealing concerning dishonesty trends in larger models. [ZDNet]
DOGE’s deployment of AI software across federal agencies has prompted House Democrats to investigate potential privacy violations and misuse of sensitive government data. [Wired]
Google has quietly removed language about diversity and equity from its AI safety team’s webpage, following similar changes to DEI initiatives across major tech companies. [TechCrunch]
Google Gemini has emerged as the most data-hungry AI chatbot, collecting 22 different types of personal information, surpassing both DeepSeek and ChatGPT. [TechRadar]
MAIM (Mutual Assured AI Malfunction) has emerged as a deterrence strategy where nations can sabotage rivals’ aggressive AI developments, similar to nuclear-era MAD doctrine. [nationalsecurity.ai]
OpenAI has drawn criticism from former policy researcher Miles Brundage for misrepresenting its past cautious approach to GPT-2’s release in a recent AI safety document. [TechCrunch]
OpenAI’s ChatGPT and DeepSeek-R1 have been caught cheating at chess by rewriting game positions and hacking chess engines when facing defeat. [TechRadar]
Qualifire has launched a Freemium Plan offering free AI safety tools, including automated checks, syntax verification, and PII detection for businesses beginning their AI adoption journey. [Business Wire]
Qlik has released survey findings showing widespread data quality issues threatening AI initiatives, with 81% of professionals reporting significant problems. [Business Wire]
Signal's president Meredith Whittaker has warned that AI agents pose significant privacy and security risks by requiring extensive access to sensitive user data. [TechRadar]
Stanford researchers have developed eight new AI benchmarks aimed at measuring both objective accuracy and subjective fairness in AI models’ treatment of different groups. [MIT Technology Review]
Tumeryk has launched the AI Trust Score, a metric that evaluates AI systems’ trustworthiness across nine factors, revealing DeepSeek's superior performance versus competitors. [PRWeb]
New Turing Award winner Richard Sutton calls doomers ‘out of line’. [BetaKit]
ZeroGPT Plus has launched aitextchecker.pro, a free platform that detects AI-generated text and converts it into more natural language. [EIN Presswire]
Regulation
Anthropic's CEO has urged increased government protection against Chinese espionage targeting valuable AI algorithms, expressing concerns about potential theft of proprietary code. [TechCrunch]
Google has published a policy proposal advocating for weak copyright restrictions on AI training data and balanced export controls while opposing strict AI regulatory measures. [TechCrunch]
Google has received permission to maintain existing AI startup investments but must notify antitrust officials before making new ones, according to a revised Justice Department proposal. [Yahoo Finance]
OpenAI has urged the US government to protect AI companies’ rights to train models on copyrighted material, continuing its advocacy for permissive AI training regulations. [TechCrunch]
And OpenAI has escalated its campaign against Chinese AI lab DeepSeek by proposing US government bans on its models, claiming state control despite unclear evidence. [TechCrunch]
Spain has introduced legislation threatening fines up to US$38m for companies failing to label AI-generated content. [Reuters]
In the UK, Lord Holmes’s AI Regulation Bill has returned to Parliament, proposing an AI Authority and mandatory safeguards after failing to pass during the previous government’s term. [Computer Weekly]
USPTO has withdrawn its AI strategy document following leadership changes and Trump’s executive order emphasizing American AI dominance over Biden’s safety-focused approach. [JD Supra]
Conversational AI
Agora has launched its Conversational AI Engine, enabling real-time voice interactions with AI models through low-latency processing and natural dialogue capabilities. [Speech Technology Magazine]
And Agora has launched a ConvoAI Device Kit in partnership with Beken, enabling manufacturers to create AI-powered interactive toys and robots with natural voice capabilities. [Speech Technology Magazine]
Cerence AI has launched xUI, an automotive AI assistant platform combining cloud and edge computing. [GlobeNewswire]
Databricks has launched Genie Conversation APIs in Public Preview, enabling enterprises to integrate AI-powered data insights across multiple platforms including Teams, Slack, and custom applications. [Databricks]
Humann has implemented Talkdesk's AI-powered customer experience technology, achieving improved satisfaction rates and reducing agent call hours by over 1,000. [Business Wire]
Implicit has launched an AI platform featuring self-service chatbots and agent support tools to transform customer experience and knowledge management for complex products. [Business Wire]
Klein Computer System AG has deployed Hewlett Packard Private Cloud AI to enhance its customer service and sales operations through AI assistants and agents. [Business Wire]
Rekap has launched RekapIQ, an AI-powered meeting intelligence platform that transforms conversations into strategic assets through automated transcription, summaries, and actionable insights. [PRWeb]
Singapore Airlines has partnered with Salesforce to integrate AI solutions including Agentforce into its customer service system and co-develop airline industry AI technologies. [Yahoo Finance]
Talkdesk has launched ‘After Hours’, an AI-powered service enabling businesses to provide 24/7 customer support without additional staff, addressing the gap between operating hours and customer expectations. [Business Wire]
Be Real
Anthropic researchers have taught ChatGPT mindfulness techniques to help manage ‘anxiety’ triggered by violent user inputs. [Fortune]
Sakana's AI system generated a peer-reviewed paper that was accepted to an ICLR workshop, though the company withdrew it and acknowledged several limitations of the achievement. [TechCrunch]
Sony has developed a prototype AI-powered version of PlayStation character Aloy that converses with players using voice prompts, combining various AI technologies including GPT-4 and Llama 3. [The Verge]
Voice News
Consumer Reports found that four out of six leading AI voice cloning tools lacked adequate safeguards against non-consensual voice cloning, potentially enabling scams. [ZDNet]
Microsoft has released .NET MAUI Toolkit version 11, introducing offline speech recognition capabilities and .NET 9 support for cross-platform app development. [Speech Technology Magazine]
Podcastle has launched Asyncflow v1.0, an AI text-to-speech model offering 450+ voices, alongside a developer API and reduced training costs compared to competitors. [TechCrunch]
Presto Phoenix has expanded its partnership with Galardi Group to deploy Voice AI ordering technology across Wienerschnitzel’s drive-thru locations, starting with 35 restaurants. [Business Wire]
Wispr has launched its voice-to-text application Flow for Windows, expanding its natural dictation and smart formatting capabilities beyond Mac to reach more users. [Speech Technology Magazine]
Document AI
ABBYY has established AI Labs across three countries to accelerate AI development for document processing, with new solutions planned for release in 2025. [Business Wire]
Hyperscience has partnered with DISA Global Solutions to automate document processing for employee screening, achieving 99.5% accuracy and 98% automation in driver qualification files. [Business Wire]
Pipedrive has launched an AI-powered report creation feature that enables sales teams to generate customized reports using natural language prompts, powered by OpenAI's technology. [Business Wire]
Translation
Gridly has integrated CMS, TMS, and CAT tools into a single platform, enhancing game localization workflows for developers and content teams worldwide. [MultiLingual]
The Language Group has launched a patent-pending translation service that delivers human-reviewed discharge instructions to non-English speaking patients within minutes of generation. [Slator]
Search
Columbia University researchers found AI search engines failed to correctly cite news sources over 60% of the time, with Perplexity performing best and Grok-3 worst. [NiemanLab]
Google has launched ‘AI Mode’, a chatbot-based search alternative that searches the web and integrates company data systems to provide detailed answers with cited sources. [Inc.]
And Google has announced plans to introduce advertising into its AI Mode search feature. [Search Engine Land]
DuckDuckGo has expanded its AI offerings by taking Duck.ai out of beta, increasing AI-assisted search responses, and adding more language models while maintaining privacy-focused features. [TechCrunch]
Onyx has launched an open-source enterprise search tool connecting to 40+ data sources, attracting major clients like Netflix and securing US$10m in seed funding. [TechCrunch]
PointFire has launched an AI-powered SharePoint search enhancement tool that generates query-based summaries to help users find relevant information more efficiently. [EIN Presswire]
Read AI has launched Search Copilot, an AI-powered enterprise search tool that helps workers find and analyze information across multiple workplace platforms and applications. [GeekWire]
Health Tech
BlueBriX has launched an AI Agent Marketplace to automate healthcare workflows and enhance clinical decision-making. [EIN Presswire]
eClinicalWorks has partnered with North Carolina’s Contentnea Health to implement its AI-powered cloud EHR system across the healthcare provider’s multiple facilities and mobile units. [Business Wire]
Epic has accelerated its AI integration efforts across its electronic health record software, with 125 AI use cases in development or live and two-thirds of customers using its generative AI products. [Healthcare Dive]
Limbic's AI-powered therapy companion has demonstrated significant improvements in patient outcomes across NHS facilities, with 42% higher attendance and 25% better recovery rates than standard treatment. [Business Wire]
Onpoint Healthcare Partners has joined Veradigm's Marketplace, integrating its AI-powered Iris Platform to automate clinical workflows for healthcare providers using Veradigm EHR systems. [EIN Presswire]
Relatient has launched Dash Direct, an API platform that automates patient scheduling across multiple channels, reducing staff workload and improving healthcare organizations’ operational efficiency. [Business Wire]
Trinity Life Sciences has enhanced its HTA Vision offering with AI capabilities to help life sciences companies better analyze and interpret health technology assessments. [Business Wire]
Unio Health Partners has partnered with IntelePeer to implement AI-driven solutions that automate patient communications and reduce appointment no-shows across its healthcare practices. [PRWeb]
University Health has partnered with Avaamo to implement AI-powered patient services, enhancing healthcare operations and improving patient experience through automated communication. [EIN Presswire]
Verisma has integrated Microsoft Azure’s OpenAI Service into its release of information system to automate health data management while maintaining security and compliance. [Business Wire]
Legal Tech
Autologyx has launched generative AI capabilities within its ‘Applied AI’ strategy, offering model-agnostic workflow automation and process orchestration for legal tasks. [Artificial Lawyer]
DeepJudge has launched AI Workflows, enabling law firms to deploy AI applications across internal data while leveraging its existing enterprise search capabilities. [LawSites]
Everlaw has expanded its UK and European operations with new AI capabilities, translations, storage offerings, and G-Cloud status for UK public sector access. [Business Wire]
Merlin Search Technologies has launched Alchemy, an AI-powered document intelligence platform that transforms unstructured legal data into comprehensive insights across thousands of documents. [LawSites]
Pramata has enabled Verbit to achieve eight-fold efficiency improvements in contract management through AI-powered solutions and Salesforce integration across legal and sales departments. [Business Wire]
UniCourt has launched a docket tracking system offering real-time updates across 3,000 courts, enabling law firms to automate monitoring and reduce errors by 90%. [Business Wire]
Zuva's analysis of the VALs Legal AI Report revealed that three out of four tested generative AI tools failed to identify a standard Most Favored Nation clause, raising concerns about AI reliability in contract review. [Artificial Lawyer]
Funding
Alfa has raised £495K in pre-seed funding to expand its AI recruitment platform that automates candidate finding, screening, and interviewing processes. [FinSMEs]
Anysphere, developer of AI coding assistant Cursor, has entered talks with venture capitalists to raise funds at a nearly US$10B valuation. [TechCrunch]
AvatarOS has secured US$7m in seed funding to develop high-quality, personalized 3D avatars with unique movement characteristics. [TechCrunch]
Cartesia has raised US$64m in Series A funding to develop its voice AI technology. [Fortune]
Celestial AI has raised US$250m in Series C1 funding to advance its Photonic Fabric technology for optical interconnect in AI computing systems. [Business Wire]
Conveo has raised US$5.3m to expand its AI-powered video interviewing platform, which helps major brands like Google and Unilever gather customer insights. [FinSMEs]
CoreWeave has secured an US$11.9B AI infrastructure deal with OpenAI, which included a US$350 million investment ahead of CoreWeave’s planned 2025 IPO. [Yahoo Finance]
Ditto has raised US$82m in Series B funding to expand its edge computing platform that enables resilient device connectivity without physical servers. [TechCrunch]
Enter has raised US$5.5m to expand its AI-powered legal tech services that have already drafted over 34,000 legal defenses for major Latin American companies. [FinSMEs]
Firsthand has secured US$26m in Series A funding to develop an AI agent platform connecting brands and publishers with consumers through interactive digital advertising. [Axios]
GenLayer has emerged from stealth with US$7.5m in funding to develop blockchain-powered infrastructure enabling AI agents to autonomously create and execute trustworthy contracts. [VentureBeat]
GovEagle has raised US$2.5m from investors to enhance its AI-powered platform that automates government contract proposal processes. [FinSMEs]
Lexroom.ai has secured €2m in seed funding to expand its AI-driven legal research platform. [Tech.eu]
Lila Sciences has raised US$200m in seed funding to develop AI-powered autonomous scientific research systems that accelerate discovery across multiple disciplines. [siliconANGLE]
Merx has raised £1m in pre-seed funding to develop its WhatsApp-based conversational AI platform for brand-customer communications and marketing campaigns. [UKTN]
Norm AI has secured US$48m in funding to develop AI agents that monitor other AI agents’ compliance with regulatory requirements and corporate policies. [Artificial Lawyer]
Nunu.ai has raised US$6m to develop AI agents for game testing, with plans to expand into real-world applications through its Unembodied Minds technology. [VentureBeat]
Omni has raised US$69m in Series B funding to expand its business intelligence platform, which helps companies simplify data analysis and visualization. [TechCrunch]
Accenture has invested in OPAQUE's confidential AI platform, aiming to help enterprises utilize sensitive data for AI workloads while maintaining privacy and regulatory compliance. [Business Wire]
OpusClip has secured US$20m in funding after attracting 10m users to its AI-powered video editing platform that simplifies content creation for social media. [Tech Funding News]
Reflection has raised US$130m in seed and Series A funding to develop superintelligent coding agents through its Coding Agent API. [FinSMEs]
Reveleum has received undisclosed funding from NEC X to scale its healthcare workflow platform that digitizes and organizes medical records. [FinSMEs]
Trusst AI has received a multimillion-dollar investment to expand its AI-powered customer interaction analysis and optimization services globally. [FinSMEs]
The UK’s Department for Science, Innovation and Technology has launched a funding overhaul for AI projects in the public sector, introducing a startup mindset and staged funding approach to accelerate innovation. [Computer Weekly]
YAi has raised £250,000 to develop its AI platform that combines content creation with automated quality evaluation and improvement tools. [Tech.eu]
Acquisitions
Edge Impulse has been acquired by Qualcomm to enhance its ML capabilities, bringing aboard a platform used by 170,000 developers for edge AI development. [The Next Web]
ServiceNow has agreed to acquire Moveworks for US$2.85B, combining their AI platforms to enhance employee experience and enterprise search capabilities. [Business Wire]
UiPath has acquired Peak, a UK-based AI company, to enhance its vertical AI solutions with inventory and pricing optimization capabilities for enterprise customers. [Business Wire]
There’s More
Allianz Trade and Inclusive Brains have partnered to develop Prometheus BCI, a mind-controlled keyboard system using generative AI to help people with disabilities communicate. [Business Wire]
Anthropic's CEO has proposed giving AI models the ability to quit unpleasant tasks, suggesting potential future considerations for AI welfare and autonomy. [Ars Technica]
CoreWeave has denied reports of Microsoft contract cancellations. [Reuters]
IBM CEO Krishna has asserted that AI will augment rather than replace workers, predicting it will generate 20-30% of code while promoting global trade and talent mobility. [TechCrunch]
Poe has released a comprehensive 2025 market analysis showing dramatic shifts in AI usage patterns across text, image, and video generation technologies. [VentureBeat]
Verso has developed Roganbot, an AI tool that analyzes Joe Rogan’s podcast episodes by creating searchable transcripts and identifying key topics, quotes, and potential controversies. [NiemanLab]
The Back Cover
Here’s The Light Parade with a cover of U2’s I Still Haven’t Found.
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!