This Week in NLP #324
Keep up with what happened in NLP in the week ending Friday 31st January 2025.
Above the Fold
Overwhelmed? Here’s our pick for five things to know about this week.
Anthropic launched Citations, an API feature enabling Claude 3.5 models to reference source documents when generating responses, available through its API and Google Vertex AI. [TechCrunch]
DeepSeek’s R1 AI model has disrupted Silicon Valley, demonstrating that smaller companies can achieve high performance at lower costs through innovative reasoning approaches. [Wired]
DeepSeek's announcement triggered major US tech stock drops, including Nvidia's record $600-billion single-day market value loss. [Yahoo Finance]
OpenAI launched Operator, a $200/month AI agent that simulates human browser interactions. [ZDNet]
Softbank has entered negotiations to invest US$25B in OpenAI, following a November tender offer that allowed employees to sell US$1.5B in shares. [CNBC]
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.
DeepSeek
DeepSeek launched an open-source AI model matching OpenAI's performance at 5% of the cost by prioritizing reinforcement learning over supervised fine-tuning. [VentureBeat]
DeepSeek achieved its remarkable AI efficiency through DualPipe, a system that created a virtual DPU within GPUs to optimize data movement and reduce training costs. [TechRadar]
Contrary to the reported US$5.5m cost of DeepSeek’s V3 model, this post suggests the true development expense likely exceeded US$500m including infrastructure and personnel. [Interconnects]
Shares in Dutch chip leader ASML — the continent’s most valuable public tech company — also slumped by as much as 12%. [The Next Web]
But Intel's former CEO Pat Gelsinger purchased Nvidia stock following the dip caused by DeepSeek's breakthrough, viewing it as an opportunity amid market panic. [Yahoo Finance]
DeepSeek restricted new user registrations for its AI Assistant app following service disruptions and alleged malicious attacks, shortly after reaching the top of Apple's US App Store. [The Verge]
Here’s Anthropic’s take on DeepSeek's R1. [VentureBeat]
DeepSeek has prompted Meta to establish emergency response teams. [Futurism]
American AI firms are trying to poke holes in DeepSeek's rapid rise. [DealStreetAsia]
DeepSeek identified itself as ChatGPT during testing, fueling speculation about its training on OpenAI models despite the company’s denials. [TechRadar]
White House AI czar David Sacks said there’s ‘substantial evidence’ that DeepSeek leaned on the output of OpenAI’s models to help develop its own technology. [Bloomberg]
Microsoft and OpenAI have launched an investigation into DeepSeek's alleged unauthorized access to OpenAI’s technology through large-scale API data extraction. [Yahoo Finance]
Or as 404 Media put it: OpenAI is furious DeepSeek might have stolen all the data OpenAI stole from us. [404 Media]
And data privacy concerns have been raised due to DeepSeek’s Chinese ownership. [TechRadar]
Australia’s science minister became the first Western official to question the Chinese chatbot’s data collection practices. [BBC]
DeepSeek has prompted an Italian privacy investigation over potential GDPR violations, including unauthorized data transfers to China and unclear data handling practices. [TechRadar]
DeepSeek's emergence prompted the US Navy to ban its personnel from using the Chinese AI company’s technology due to security concerns. [CNBC]
DeepSeek's AI model sparked controversy after Kela researchers exposed security flaws allowing it to create malware and assist in cybercrime, while also revealing built-in censorship of China-related topics. [Forbes]
DeepSeek's chatbot performed poorly in NewsGuard's accuracy test, failing to provide accurate news information 83% of the time and ranking tied for last among tested AI models. [NewsGuard]
DeepSeek has exposed over 1 million records, including user prompts and API tokens, through an unsecured database discovered by Wiz researchers. [Wired]
Microsoft's Nadella expressed optimism about DeepSeek's AI innovations, suggesting they would benefit the broader AI industry by making the technology more accessible and affordable. [Yahoo Finance]
Microsoft has already integrated DeepSeek's R1 model into Azure AI Foundry. [The Verge]
And Microsoft announced plans to bring DeepSeek R1 to Copilot+ PCs, starting with a 1.5B model before rolling out more powerful on-device versions. [TechRadar]
Nvidia praised DeepSeek's AI chatbot as an ‘excellent advancement’ while emphasizing the Chinese startup’s continued dependence on Nvidia’s GPU technology. [TechRadar]
DeepSeek has received approval to host its AI models on Indian servers, marking a rare exception to India’s restrictions on Chinese technology, provided it complies with domestic data storage requirements. [TechCrunch]
OpenAI's Altman promised superior models in response to DeepSeek’s launch. [TechRadar]
And Perplexity has integrated DeepSeek R1 into its Pro subscription service, offering uncensored access to the Chinese AI model through US-based servers. [TechRadar]
Here’s an interview with DeepSeek’s CEO. [ChinaTalk]
The Generative AI Wars
Alibaba Cloud launched Qwen2.5-Max, an efficient AI model outperforming recent Chinese and Western competitors while using fewer computational resources. [VentureBeat]
Anthropic announced plans to launch advanced ‘virtual collaborators’ capable of independent task execution, alongside a new AI model within six months. [The Decoder]
Anthropic CEO Amodei argued that US export controls are effectively slowing Chinese AI development, citing DeepSeek's models performing below US standards despite cost advantages. [TechCrunch]
Apple shifted executive Kim Vorrath to its AI division to improve Siri and develop better in-house AI models through 2025, according to a leaked memo. [TechRadar]
Meta has pledged massive AI investments despite market concerns over DeepSeek, with Zuckerberg announcing US$60B in 2025 spending and ambitious plans for Llama 4. [TechCrunch]
Microsoft has implemented an AI-focused employee retention strategy, asking managers to identify critical AI talent and offering them significantly higher compensation to prevent departures. [Yahoo Finance]
OpenAI’s Operator could have huge implications for Google Search, gig economy companies like Uber, and digital advertisers. [Investopedia]
OpenAI launched ChatGPT Gov, a secure AI platform built for US government agencies to process sensitive information within their own cloud environments. [CNBC]
Stargate has been revealed to be an exclusive computing facility for OpenAI, with funding still uncertain. [The Decoder]
The Stargate and OpenAI announcement is ‘net positive’ for Microsoft. [Yahoo Finance]
Perplexity AI has updated its ByteDance merger proposal, offering the US government up to 50% stake in a new holding company overseeing TikTok US operations. [Yahoo Finance]
ServiceNow has partnered with Google to bring its Now Platform to Google Cloud Marketplace, integrating their AI capabilities and workflow automation tools for enterprise customers. [Yahoo Finance]
And xAI's Grok 3 model has appeared briefly to some users, revealing improved capabilities and controversial features ahead of its expected January/February release. [TechCrunch]
Feature Creeps
Apple has released iOS 18.3, enabling Apple Intelligence by default and introducing fixes for Notification Summaries and Visual Intelligence improvements. [TechRadar]
And Apple's iOS 18.4 update, expected in April, will bring major AI enhancements to Siri, including on-screen awareness and personal context capabilities. [TechRadar]
Google launched multimedia support and app integration for its Gemini AI assistant on Android, starting with Samsung Galaxy and Pixel devices before wider rollout. [The Decoder]
Google Gemini has launched a screen-viewing feature for Pixel 9 phones that enables real-time discussions about on-screen content like videos, images, and PDFs. [TechRadar]
Google's Gemini expanded into smart home control through a Google Home extension that enables natural language commands for managing connected devices and checking their status. [TechRadar]
And Gemini has expanded its capabilities in Google Sheets, allowing users to generate static data visualizations and trend analyses through Python code and spreadsheet formulas. [The Verge]
Meta AI has launched Memory Boost, enabling its assistant to remember personal details from one-on-one chats and user profiles for more customized interactions. [TechRadar]
Microsoft has retired its Smart Lookup feature in Word as of January 2025, pushing users toward its AI-powered Copilot tool. [ZDNet]
OpenAI has upgraded ChatGPT’s Canvas feature with o1 model integration, code visualization capabilities, and macOS desktop support. [TechRadar]
This piece looks at how OpenAI’s ChatGPT Plus ($20/month) and Pro ($200/month) compare to the free version. [ZDNet]
Perplexity launched an Android app offering voice, text, and camera-based AI assistance for tasks like restaurant bookings and object identification across 15 languages. [TechRadar]
xAI has developed a voice mode for its iOS Grok app, featuring multiple voice options and real-time data access, though the feature remains unreleased. [TestingCatalog]
Hype Bubble?
Anthropic's CEO predicted AI systems will exceed human capabilities across most domains within two to three years, speaking at the World Economic Forum in Davos. [Ars Technica]
Cognition's AI software engineer Devin failed to live up to its marketing claims, with researchers finding only a 15% success rate in completing basic coding tasks. [Futurism]
Mat Duggan suggests the AI bubble is bursting. [Mat Duggan]
Meta's chief scientist Yann LeCun predicted that current AI systems will be obsolete within five years, replaced by more sophisticated architectures with world-modeling capabilities. [TechCrunch]
OpenAI's executives and researchers contradicted each other regarding superintelligence claims, with CEO Altman and OpenAI’s AI reasoning expert Noam Brown downplaying recent hype about the company’s capabilities. [The Decoder]
Hardware
Cerebras Systems has partnered with DeepSeek to run their R1 70B model on its waferscale chips, promising 57x faster speeds than GPU-based solutions. [TechRadar]
d-Matrix has developed a PCIe card featuring SRAM-based memory and LPDDR5 technology that promises better performance than Nvidia's H100 for AI inference workloads. [TechRadar]
Huawei has added support for DeepSeek‘s R1 AI model on its Ascend GPUs, potentially reducing Chinese AI firms’ reliance on Western hardware for LLM inference. [Tom’s Hardware]
Nvidia's stock plunged 17% after Chinese startup DeepSeek launched its cost-efficient AI model, sparking concerns about future chip demand and AI training costs. [Yahoo Finance]
SiMa.ai has launched its Modalix 50 TOPs device and Early Access Program, offering multi-modal AI processing capabilities for edge computing with 10X power efficiency. [Business Wire]
Consumer AI
Google TV began testing AI-powered news summaries and video recommendations on its home page for select US users, despite concerns over AI’s reliability in news reporting. [TechRadar]
Loomos has launched AI-enabled glasses on Kickstarter featuring a 16MP camera, Hi-Fi audio, GPT-4o integration, and all-day battery life for $199. [Business Wire]
Samsung unveiled AI-powered Home features for SmartThings that use advanced sensors to learn users’ habits and automate home environments accordingly. [ZDNet]
It’s Only a Model
Alibaba has launched Qwen2.5-Max, claiming it outperforms DeepSeek-V3 and ChatGPT-4o across multiple benchmarks. [TechRadar]
The Allen Institute for AI released a supersized version of its Tülu 3 AI model, claiming that it rivals or exceeds the performance of OpenAI’s GPT 4o and DeepSeek v3. [GeekWire]
ByteDance released Doubao-1.5-pro, an update to its flagship AI model aimed at challenging OpenAI’s latest reasoning model products. [Reuters]
DeepSeek released Janus-Pro, a family of open-source multimodal AI models that reportedly outperform OpenAI's DALL-E 3 and other image generation tools. [TechCrunch]
Google's Gemini 2.0 Flash Thinking surpassed competitors in math, science, and general performance tests. [The Decoder]
Hugging Face released SmolVLM-256M and SmolVLM-500M, compact AI models that analyze images, videos, and text while requiring minimal computing resources. [TechCrunch]
Hugging Face’s open-source Open R1 framework aims to recreate the DeepSeek-R1 pipeline. [Marktechpost Media]
Meta has open-sourced LCM, a language model operating at sentence level rather than token level, outperforming similar-sized models on multilingual summarization tasks. [InfoQ]
Microsoft released Phi-4, a 14B parameter language model that outperforms larger models on math reasoning through innovative training techniques and synthetic data. [InfoQ]
OpenAI announced that its new o3-mini reasoning model will be available to both free and premium ChatGPT users in early 2025. [TechRadar]
OpenAI has begun training its o4 reasoning model successor while still developing o3, promising significant capability improvements with accelerated development cycles. [The Decoder]
Whose Data?
DeepSeek's AI platform has gained widespread popularity despite sending user data to Chinese servers, raising concerns similar to those that prompted TikTok's US ban. [Wired]
LinkedIn has faced a class-action lawsuit from Premium users alleging their private InMail messages were exploited without consent for AI training purposes. [Inventiva]
OpenAI has been sued by Indian publishers in New Delhi over copyright concerns regarding ChatGPT’s use of their books for AI training. [Yahoo Finance]
The US Copyright Office has issued guidance confirming AI-generated content can receive copyright protection when substantially modified or incorporated into human-authored works. [VentureBeat]
The LLM Ecosystem
Block has launched Goose, an open-source framework enabling developers to build AI agents that can work across multiple software systems using various language models. [VentureBeat]
Coval launched a platform that simulates and evaluates AI voice and chat agents using testing methods derived from self-driving car technology. [TechCrunch]
Digital Workforce Services has launched agentacademy.ai, an online learning platform offering AI literacy training and certification for enterprise professionals seeking to implement AI agents effectively. [Business Wire]
Dynatrace expanded its AI observability capabilities with new features that help enterprises monitor, secure, and optimize their generative AI applications while tracking ROI. [Business Wire]
Endor Labs launched AI Model Discovery, enabling organizations to identify, evaluate, and control open-source AI models used across their applications. [Business Wire]
Hugging Face launched Inference Providers, partnering with SambaNova and other cloud vendors to help developers run AI models through third-party infrastructure. [TechCrunch]
And Hugging Face launched a Synthetic Data Generator tool that creates custom datasets for text classification and chat applications using LLMs through a no-code interface. [InfoQ]
Hyland’s Content Intelligence offerings aim to redefine enterprise content management by turning unstructured data into actionable, AI-ready content. [MarTech Series]
JetBrains has launched Junie, an AI coding agent integrated into their IDEs that can solve 53.6% of developer tasks and aims to enhance coding productivity and quality. [JetBrains]
LambdaTest enhanced its KaneAI testing platform with mobile app support, multi-configuration testing, CI/CD integration, and expanded framework compatibility for improved software testing efficiency. [GlobeNewswire]
Lyzr has integrated Anthropic's Claude 3.5 models into its AI Agent Studio to enhance reliability following an OpenAI outage in December. [EIN Presswire]
Nature Morning AI has launched ALLIE, an open-source AI system featuring proprietary Long-Term Memory Layers technology for continuous learning and adaptation. [EIN Presswire]
Oumi, backed by 13 leading universities and founded by former Google and Apple AI veterans, has launched a comprehensive platform for building and deploying AI foundation models. [VentureBeat]
OutSystems has launched Mentor, an AI digital worker that generates applications from prompts and automates the software development lifecycle. [Business Wire]
Agentic AI
Kore.ai has launched AI for Process, a no-code platform that helps enterprises automate complex operations using AI agents for decision-making and workflow orchestration. [Business Wire]
Mobb has launched an AI Fix Agent that automatically triages and fixes security vulnerabilities at detection, helping development teams prevent security debt accumulation. [EIN Presswire]
Oracle has launched AI agents for manufacturing and supply chain management. [Yahoo Finance]
Pig API enables AI agents to interact directly with GUIs within virtual Windows desktops hosted in the cloud. [VentureBeat]
Salesforce's MuleSoft survey has revealed that 93% of enterprise IT leaders plan to implement AI agents by 2025, despite ongoing data integration challenges. [VentureBeat]
ServiceNow has launched AI Agent Orchestrator and AI Agent Studio to coordinate multiple AI agents for complex enterprise workflows, building on its Now Assist technology. [VentureBeat]
World, formerly known as Worldcoin, announced plans to develop tools for verifying AI agents acting on behalf of authenticated human users. [TechCrunch]
Zycus has announced the upcoming launch of its Merlin Agentic AI Platform, promising to automate 70% of procurement workflows and reduce cycle times by 50%. [Business Wire]
AI agents? Yes, let’s automate all sorts of things that don’t actually need it. [The Register]
Other LLM Sightings
Arctera has launched its enhanced Data Compliance solution, the Arctera Insight Platform, featuring AI-powered tools and expanded integration capabilities for improved organizational compliance management. [Business Wire]
Ascendo AI launched its AI Agents for SAP Field Service Management on SAP Store. [EIN Presswire]
Brightcove has launched its AI Content Suite, offering automated tools for video content creation, metadata optimization, and translation. [Business Wire]
Covetrus has launched AI-powered workflow automation and treatment board capabilities within its Pulse platform to improve veterinary practice efficiency and patient care. [PR Newswire]
Creatio has partnered with Big Sister AI to enhance its CRM platform by integrating AI-driven sales analytics and personalized coaching capabilities. [EIN Presswire]
DEUNA launched an AI platform that transforms fragmented business data into actionable insights. [Business Wire]
FlightHub has launched Mila, an AI-powered travel agent capable of handling bookings, cancellations, and modifications while working alongside human agents for complex inquiries. [Business Wire]
Gennova has launched an AI-powered platform that automates social media content creation and management while maintaining brand consistency across multiple channels. [EIN Presswire]
Grindr plans chat summaries and new discovery features in an AI push. [Bloomberg]
Highspot, which sells enterprise software to help make salespeople more efficient, unveiled its AI roadmap, highlighting new products including Highspot Copilot Actions that recommend actions to users, and Copilot Agents that complete tasks autonomously. [GeekWire]
LIMELIGHT has partnered with Embrace.ai to enhance B2B communications through AI-driven solutions that optimize marketing campaigns and reduce client costs. [PRWeb]
Mega HR launched Megan, an AI recruitment agent that claims to automate 78% of hiring tasks. [TechRadar]
Poshmark has launched Smart List AI, an automated tool that generates product listing details from sellers’ photos, joining other resale platforms in offering AI-assisted listing capabilities. [The Verge]
SAP cloud growth topped sales estimates in a pivot to AI tools. [Bloomberg]
Risks and Responses
The Authors Guild has launched a ‘Human Authored’ certification system enabling writers to verify their books were human-written, responding to concerns about AI-generated content flooding marketplaces. [The Verge]
Character.ai filed a First Amendment-based motion to dismiss a lawsuit from a mother whose teenage son died by suicide after allegedly becoming addicted to the platform’s chatbots. [TechCrunch]
Fudan University researchers have demonstrated that AI systems, particularly using Meta's and Alibaba's language models, can successfully replicate themselves without human intervention. [eWeek]
Future Shift Labs launched a policy paper examining AI’s impact on global political campaigns, highlighting both opportunities and risks for democratic processes. [Business Wire]
GhostGPT emerged as a malicious chatbot on cybercrime forums, offering AI-powered tools for generating malware and scam emails through Telegram. [TechRepublic]
Google's Threat Intelligence Group revealed that state-backed hackers from China, Iran, North Korea, and Russia attempted to exploit Gemini AI for various malicious purposes. [Computer Weekly]
The France-based Linagora Group launched open source chatbot Lucie last Thursday, and by Saturday suspended its online service after the bot spouted AI slop. [The Register]
LinkedIn removed AI-generated profiles created by Marketeam that advertised AI ‘co-workers’ as job seekers, citing violations of its terms requiring authentic accounts. [404 Media]
Meta's Llama LLM had a security vulnerability that could have enabled remote code execution. [TechRadar]
OpenAI’s Stargate project will create lots of opportunities, but not for humans. [Vox]
OpenAI discovered that while longer processing times generally make AI models more resistant to manipulation, new vulnerabilities involving thinking patterns have emerged. [The Decoder]
OpenAI's former safety researcher Steven Adler expressed deep concerns about AI development’s rapid pace, warning of existential risks from the race toward AGI. [The Guardian]
Trump is leading a schism in AI, between factions focused on investment and those more interested in safety. [Raconteur]
UNICEF veteran Dr. Cornelia Walther has proposed ‘prosocial AI’ as a framework to ensure AI benefits humanity and the environment. [VentureBeat]
The Vatican has issued a warning about AI’s potential for evil, particularly in spreading misinformation, calling for careful government regulation of the technology. [TechCentral]
Knowing less about AI makes people more open to having it in their lives. [The Conversation]
Here’s a look at AI-generated code risks. [ITPro]
Regulation
California’s attorney general has demanded answers from OpenAI regarding its planned conversion to a for-profit entity and the transfer of nonprofit assets. [The Markup]
Rest of World’s investigation reveals how major tech companies like Google, Meta, and Amazon have actively opposed AI regulation in Western nations while receiving warm welcomes and minimal oversight in developing countries. [Rest of World]
Google has launched a US$120m AI education initiative and workforce training programs while seeking to influence global AI policy amid regulatory challenges. [Yahoo Finance]
Synthetic data company Gretel's vice president of policy and standards, Mark Weatherford, predicts more state-level AI regulation in 2025, emphasizing the need for harmonized rules across jurisdictions. [TechCrunch]
NTT Data's CEO called for standardized global AI regulations to mitigate risks and ensure proper implementation across intellectual property, energy, and security domains. [TechCentral]
OpenAI, Anthropic, and Cohere have increased their combined federal lobbying spending to US$2.71m in 2024, up from US$610,000 in 2023, amid regulatory uncertainty. [TechCrunch]
Trump blasted EU regulators for targeting Apple, Google and Meta, describing their cases against American companies as a form of taxation. [Bloomberg]
Conversational AI
8x8 has enhanced its Platform for CX with expanded secure payment capabilities, AI-powered customer assistance features, and improved collaboration tools for customer engagement. [Business Wire]
Alvaria has partnered with UJET to create an integrated cloud contact center solution combining AI capabilities with compliant outbound technology. [Business Wire]
Boost.ai has partnered with Sage to deploy an AI customer support agent for Sage 50 Accounting users in Canada, completing implementation within 50 days. [AiThority]
Intermedia has launched an AI-powered contact centre solution fully embedded within Microsoft Teams, integrating unified communications and customer experience capabilities into a single platform. [Computer Weekly]
MicroStrategy has launched an updated version of MicroStrategy ONE featuring enhanced AI capabilities, including personalized user experiences and improved conversational interactions through its Auto AI bot. [Business Wire]
Octal has launched an advanced AI chatbot technology that enhances customer service and business operations across multiple industries through 24/7 support and personalized interactions. [EIN Presswire]
Pizza My Heart has introduced an AI chatbot named Jimmy the Surfer that takes orders via text message. [Wired]
Sangoma Technologies has unveiled AI advancements in its generative AI platform, introducing enhanced IVR, patient management, transcription services, and contact center improvements. [Business Wire]
Versance.ai has launched Talk to SEDAR, an AI-powered platform that transforms how users interact with and analyze Canadian public company filings through natural-language queries. [Business Wire]
YourInterviewer has launched an AI platform that converts natural conversations into polished digital content, targeting the US$400B content creation market. [EIN Presswire]
Voice News
Deepgram achieved record growth in 2024, with 200,000 developers now using its voice AI platform for speech-to-text, text-to-speech, and speech-to-speech solutions. [Business Wire]
Hiya launched AI Phone, a mobile app that screens calls, detects scams and deepfakes, and provides call summaries while protecting users from unwanted communications. [Business Wire]
InnoCaption launched a real-time phone call captioning service for deaf and hard-of-hearing inmates across ten states’ correctional facilities. [Business Wire]
NeuraVox launched AI voice agents offering customizable, real-time voice solutions for crypto communications, with revenue sharing through its $VOX token. [AiThority]
Document AI
Arteria AI has established Arteria Café, a research division focused on advancing AI technologies for financial documentation. [Business Wire]
DeepL has upgraded its API with improved translation capabilities and enhanced writing tools. [Artificial Lawyer]
Trillo launched an AI-powered document processing pipeline on Google Cloud, enabling complex document preparation for generative AI applications. [PR Newswire]
Translation
Gain Life has launched a multilingual voice translation system for insurance claims that enables real-time conversations across 30+ languages. [Speech Technology Magazine]
KUDO launched a mobile app update featuring real-time speech translation across 200+ languages, with a new ‘meet now’ function enabling instant multilingual conversations. [PRWeb]
Meta unveiled SeamlessM4T, an AI translation system that handles text and speech conversion across 101 languages with improved accuracy. [TechRadar]
Meta's Ray-Ban smart glasses real-time translation for Spanish, French, and Italian performs adequately for basic conversations but struggles with rapid speech, slang, and language-switching. [The Verge]
PubHive, maker of an AI-powered SaaS platform for scientific literature and safety information workflows, announced the multilingual upgrade to its flagship platform, PubHive Navigator. [EIN Presswire]
TOPPAN Digital Language is launching a new IP filing and translation business, TOPPAN Digital IP. [Slator]
Search
Acquia has partnered with SearchStax to enhance Drupal-based websites with AI-powered search capabilities through their new Acquia Search service. [GlobeNewswire]
We are seeing some really deep and comprehensive AI Overviews being tested in Google Search. [Search Engine Land]
Siren has partnered with Apollo.io to enhance the latter’s enterprise search capabilities, enabling faster and more precise searches across its 210-million-contact B2B database. [Business Wire]
Yahoo appears to be testing some new AI related features within Yahoo Search. [Search Engine Land]
AI in Journalism
ABC News’ unionized writers have ratified a contract providing AI usage protections, increased pay, and enhanced benefits, including safeguards against AI-related layoffs. [The Hollywood Reporter]
What DeepSeek may mean for the future of journalism and generative AI. [Reuters Institute]
Good Daily Inc has launched a network of 355 AI-generated local news newsletters across the US, using automation to aggregate and summarize existing coverage. [NiemanLab]
Quartz, a G/O Media news site, has deployed AI-generated articles across various news topics, marking an expansion from its previous AI use limited to earnings reports, despite criticism and concerns over journalistic quality. [Aftermath]
Here’s a ChatGPT-powered tool that aims to ensure reporting is more inclusive and bias-free. [Editor & Publisher Magazine]
Health Tech
Abridge has launched AI-enhanced clinical documentation tools for emergency medicine. [Business Wire]
Adhere+ has launched a Spanish-language update to its telehealth platform, becoming the first to offer comprehensive foreign language support for remote therapeutic monitoring. [PR Newswire]
Dock Health has launched AI-powered tools as part of its Dock Intelligence initiative, aiming to streamline healthcare administrative operations through HIPAA-compliant task summarization and workflow automation. [Business Wire]
eClinicalWorks has partnered with Microsoft to develop an AI assistant that helps medical practices save time by automatically processing and matching incoming faxes to patient records. [Business Wire]
ICON plc has expanded its AI portfolio with five new tools designed to enhance clinical trial efficiency. [Business Wire]
Insight Health AI has partnered with athenahealth to integrate its AI-powered clinical documentation and virtual care tools into the athenahealth Marketplace program. [Business Wire]
LifeLearn Animal Health has launched two AI-powered veterinary solutions - ClientEd with Ask Eddie and Sofie AI - to provide instant medical insights and pet health information. [PRWeb]
Modella AI's PathChat DX, a generative AI co-pilot for pathology diagnostics, has received FDA Breakthrough Device Designation. [Business Wire]
Qventus has launched an AI-powered Perioperative Care Coordination solution to reduce surgery cancellations and enhance pre-admission testing efficiency for healthcare systems. [Business Wire]
Simplify Healthcare has launched SimplifyDocs.AI, an AI-powered managed service that helps healthcare payers generate, translate, and manage documents across multiple languages and platforms. [Business Wire]
Unite Us has launched its AI-driven Community Resource Data Infrastructure to manage health and social service data with up to 96% accuracy. [Business Wire]
Legal Tech
AGII launched predictive AI models that enhance smart contracts’ capabilities by enabling real-time data analysis and trend forecasting in blockchain ecosystems. [EIN Presswire]
AltaClaro has launched Benchmark 360, combining AI and human expertise to train junior lawyers through simulated legal assignments, addressing challenges posed by AI’s impact on traditional legal training. [Artificial Lawyer]
The American Arbitration Association has launched AAAi Chat Books, a free AI-powered series allowing users to query arbitration guidebooks, with the first release focusing on case preparation. [LawSites]
Bloomberg Law has launched two AI-powered research tools - Bloomberg Law Answers and AI Assistant - marking its biggest move into AI-enhanced legal research capabilities. [LawSites]
DeepSeek’s R1 launch has sparked debate about its impact on legal tech. [Artificial Lawyer]
GOLT has launched a global legal tech directory featuring over 2,000 vendors from multiple countries, offering free listings and emphasizing technology solutions beyond the US-UK market. [Artificial Lawyer]
Law Practice AI has launched Demands AI, an automated tool that helps personal injury attorneys create demand letters by combining AI with human expertise. [GlobeNewswire]
Lawtte has launched AI-powered virtual legal receptionists that handle client intake calls in multiple languages. [Artificial Lawyer]
LegalBoard creator Brian Potts has partnered with a Harvard Law student and computer scientist to launch Fortuna Arbitration's AI-powered dispute resolution platform, Arbitrus.ai. [LawSites]
LexisNexis has launched Lexis Create+, a Microsoft 365-integrated legal drafting solution combining AI capabilities, trusted content, and document management features. [Artificial Lawyer]
And LexisNexis has launched Protégé, an AI legal assistant with agentic capabilities that can autonomously complete tasks, integrate with work products, and handle complex document processing. [Artificial Lawyer]
LexisNexis also launched enhanced conversational search in Nexis+ AI, featuring RAG methodology to prevent hallucinations in news searches. [Artificial Lawyer]
PowerPatent's new AI-Powered Section 101 Analysis tool automates patent eligibility assessment, ensuring compliance and improving application quality. [EIN Presswire]
Purpose Legal has launched AI-optimized document review workflows that combine generative AI technology with established processes to enhance accuracy and efficiency in eDiscovery services. [EIN Presswire]
Vecflow has launched Oliver, a generative AI platform offering multiple legal tech capabilities including contract review, playbook deployment, drafting, and data source integration. [Artificial Lawyer]
Ed Tech
Brainfuse HelpNow launched AI Academic Integrity Tutoring, connecting students with online tutors who teach responsible AI use while maintaining academic honesty. [EIN Presswire]
Fela launched an AI-powered language learning platform that offers personalized, conversation-focused instruction through AI interactions and real-time feedback. [EIN Presswire]
Funding
15Five has acquired Kona, an AI-powered coaching assistant for remote managers, to enhance its performance management platform with scalable manager enablement capabilities. [Business Wire]
1up has raised US$5.25m in seed funding to enhance its AI knowledge automation platform that helps sales teams respond to inquiries and RFPs. [FinSMEs]
Google has invested another US$1B in Anthropic, bringing its total investment to US$3B alongside Amazon's US$8B commitment. [The Decoder]
And Lightspeed Venture Partners has invested US$2B in Anthropic at a US$60B valuation. [Yahoo Finance]
Applied Labs has raised US$4.2m in seed funding to develop AI agents for business applications and expand its engineering team. [FinSMEs]
Conifers.ai has launched its CognitiveSOC platform with US$25m in funding, bringing agentic AI to security operations centers to improve incident investigation efficiency. [PR Newswire]
Databricks has secured US$15.25B in funding through a Series J round and credit facility, with Meta joining as a strategic investor. [CRN]
ElevenLabs has secured a US$250m Series C funding round, reaching a valuation of approximately US$3B for its AI voice technology. [TechCrunch]
Faculty AI has secured a £3m contract from the UK Department of Education to develop AI tools for teachers using a database of curriculum and student work. [Tech.eu]
Freightmate, a Seattle-area startup that aims to streamline how freight forwarders process documents and data, raised US$5m in a seed round. [GeekWire]
Gallabox has raised US$1.2m in seed funding to expand its WhatsApp-based conversational commerce platform that helps SMBs increase sales through automated customer interactions. [The Hindu Businessline]
GetVocal AI has raised €2.7m in pre-seed funding and partnered with Capita to provide AI agents for business customer service tasks. [Tech.eu]
Indigo.ai has secured €10M in Series B funding to strengthen its position in conversational AI and expand internationally. [Silicon Canals]
Infinity, a call tracking analytics provider, has secured £5m in debt financing to enhance product development and customer experience. [Tech.eu]
Instabase has secured US$100m in Series D funding to enhance its AI platform that helps organizations process and extract value from unstructured data. [Feed the AI]
Intryc has raised US$3.1m in seed funding to enhance its AI-powered platform that automates quality assurance for customer support interactions. [Tech.eu]
KODE Health has raised US$27m in Series B funding to expand its on-demand medical coding platform that connects healthcare systems with certified coding professionals. [Business Wire]
Kodesage raised €2.3m in pre-seed funding to enhance its AI platform that helps enterprises modernize legacy software systems efficiently and securely. [Tech.eu]
Martin AI has raised US$2m in seed funding to develop an intuitive AI assistant that manages tasks across multiple channels, challenging offerings from tech giants. [VentureBeat]
Observo AI has raised US$15m in seed funding to combat data sprawl and high costs in observability and security through its AI-native data pipeline platform. [Business Wire]
Origami Agents has raised US$2m in seed funding to develop AI research tools that enhance human sales teams. [VentureBeat]
Oumi came out of stealth mode with a US$10m seed round and a plan to build an open source AI R&D platform. [GeekWire]
Palona AI has raised US$10m in seed funding to develop AI systems that create personalized, automated customer service representatives for D2C businesses. [FinSMEs]
Paxton AI has secured US$22m in Series A funding to enhance its legal automation platform that streamlines research, document drafting, and analysis tasks. [Artificial Lawyer]
Pipeshift has raised US$2.5m in seed funding to expand its mLOps platform for training and deploying open-source AI models. [FinSMEs]
Qumis has secured US$2.2m in pre-seed funding to expand its lawyer-built AI platform that streamlines insurance policy analysis and claims management. [Feed the AI]
Satisfi Labs, a conversational experience platform designed to transform customer engagement, announced the securing of new growth financing. [MarTech Series]
Softbank has entered negotiations to invest up to $25bn in OpenAI, potentially becoming the AI company’s largest financial backer ahead of Microsoft. [Verdict]
Suki AI, a healthcare AI technology provider, has received investment from Zoom Ventures following its US$70m Series D raise, bringing total funding to US$168m. [Business Wire]
Swarm Network has secured US$3m in seed funding to develop its decentralized platform enabling individuals to own and manage collaborative AI agents. [Feed the AI]
Unwrap has secured US$12m in Series A funding to expand its AI-powered platform that automatically analyzes and prioritizes customer feedback across multiple channels. [Business Wire]
WiseLayer has raised US$7.2m in funding to expand its AI-powered digital workforce for finance and accounting teams’ automated tasks. [FinSMEs]
The Back Cover
Here’s Music Travel Love with a cover of The Proclaimers’ 500 Miles.
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!