This Week in NLP #353
Keep up with what happened in NLP in the week ending Friday 22nd August 2025.
Above the Fold
Overwhelmed? Here’s our pick for five things to know about this week.
DeepSeek has released a 685-billion parameter open-source AI model that challenges American tech giants by offering high-performance capabilities at significantly lower costs. [VentureBeat]
Google has strategically introduced Gemini-powered AI features on Pixel smartphones to compete with rivals and leverage its Android ecosystem’s massive global reach. [CNBC]
Meta is restructuring its AI division into four specialized subdivisions, signaling a strategic reset and aggressive push to compete in the AI development landscape. [ITPro]
Microsoft has integrated an AI-powered COPILOT function into Excel, enabling users to perform generative tasks like summarizing feedback, categorizing data, and retrieving external information directly within spreadsheet cells. [The Register]
The UAE has emerged as a global AI powerhouse, ranking second worldwide in technological capability and demonstrating a comprehensive commitment to digital innovation. [ITP.NET]
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 launched an enterprise subscription for Claude Code, enabling more sophisticated integrations and admin tools for businesses using its command-line coding tool. [TechCrunch]
Apple is reportedly developing an interactive desktop robot, smart display, security cameras, and a mobile robot as part of its AI-driven smart home strategy. [The Decoder]
AWS has introduced revised Kiro AI coding tool pricing that dramatically increases costs. [The Register]
Musk has acknowledged Google DeepMind as the current AI leader due to its superior computational resources and data advantage. [EnterpriseAI.com]
Meta's consumer AI app has struggled with inconsistencies, unreliability, and user frustration since its launch, casting doubt on the company’s ambitious AI vision. [Gizmodo]
OpenAI’s ChatGPT mobile app has generated US$2B in global consumer spending, far outpacing its AI chatbot competitors in downloads and revenue. [TechCrunch]
OpenAI's head of ChatGPT hinted at potential future ads, emphasizing they would need to be thoughtful and carefully implemented. [TechRadar]
OpenAI has quashed rumors about GPT-4o’s imminent deprecation, confirming the model will remain available with advance notice of any future changes. [TechRadar]
ChatGPT Plus offers expanded access to advanced AI features like legacy models, Sora video generation, and higher usage limits for power users willing to pay $20 monthly. [ZDNet]
And OpenAI has launched ChatGPT Go, a budget-friendly subscription tier in India offering 10x higher limits and longer memory at a lower price point. [TechRadar]
Oracle has signed a deal with Google Cloud to offer Gemini AI models through its cloud infrastructure, expanding its multi-vendor AI service. [TechRadar]
Salesforce launched Agentforce for Public Sector, offering AI agents to potentially replace US government employees in various administrative and customer service roles. [The Register]
xAI reportedly lost a potential government contract after Grok’s antisemitic outburst, with the US General Services Administration allegedly dropping the AI model from its offerings. [Ars Technica]
AI Supremacy
DeepSeek's AI model development revealed China’s continued reliance on Nvidia GPUs despite efforts to use domestic Huawei chips for advanced training. [The Economic Times]
Intel may require government intervention to protect US national security and reduce reliance on foreign chip manufacturers, according to an analyst. [CNBC]
OpenAI's Altman warned that export controls won’t effectively prevent China from advancing its AI capabilities and technological ambitions. [Tom’s Hardware]
Sovereign AI
AfricAI, a joint venture of four technology leaders, has established a partnership to develop sovereign, localized AI solutions for African markets, starting with Nigeria. [The Nigerian Guardian]
Ghana and Lesotho agreed to collaborate on AI and digital transformation, creating a regional working group to develop harmonized digital frameworks and strategies. [iAfrica]
Ghana’s Deputy Minister Mohammed Adam Sukparu advocated for developing homegrown technology and nurturing local IT talent to reduce technological dependence on imports. [GhanaWeb]
Google has invested US$200m in AI-driven social projects across Africa, aiming to address challenges like wildfires, hunger, and public health emergencies. [Rest of World]
OpenAI's former executive Jade Leung has been appointed as the UK prime minister’s new AI adviser, focusing on positioning the UK as a leader in AI technology. [Tech.eu]
Telehouse Canada and Vertical Data have formed a strategic partnership to provide scalable, secure AI infrastructure solutions across Canada’s enterprise market. [Business Wire]
The UK’s Council for Science and Technology recommended UK investment in AI chip design, highlighting Arm's potential and the need for strategic focus on semiconductor innovation. [The Register]
This piece argues that Europe must strategically collaborate with US tech firms and invest heavily in open-source technologies to achieve digital sovereignty without sacrificing technological advancement. [CEPA]
Feature Creeps
Anthropic launched learning modes for Claude AI that guide users through step-by-step reasoning, transforming the chatbot into an educational tool aimed at enhancing learning rather than providing instant answers. [VentureBeat]
And Anthropic introduced Enterprise and Teams upgrades for Claude, offering enhanced usage, admin controls, and a Compliance API to help organizations scale AI more effectively. [VentureBeat]
GitHub has launched a new ‘mission control center’ enabling developers to delegate tasks to AI coding agents seamlessly across its platform without disrupting workflow. [ITPro]
Google is preparing a dedicated ‘projects’ feature for Gemini, enabling users to organize work, manage files, and reference documents within AI conversations. [TestingCatalog]
Google launched Pixel Journal, an AI-powered journaling app with personalized prompts, memory suggestions, and privacy features, exclusively for Pixel 10 devices. [TechCrunch]
Google's Pixel 10 introduces Camera Coach, an AI-powered feature designed to help users improve their smartphone photography skills by offering real-time shooting advice. [TechRadar]
Google has introduced an AI-powered audio feature in Docs that allows users to listen to documents read aloud with customizable voices and playback controls. [ZDNet]
And Google's Jules coding agent, powered by Gemini 2.5 Pro, offers free autonomous code generation with a built-in critic feature to improve code quality and prevent vulnerabilities. [ITPro]
OpenAI updated its GPT-5 model to be warmer and friendlier, addressing user complaints about the model’s initial personality. [TechCrunch]
Perplexity expanded its Finance dashboard to include live transcriptions and call schedules for Indian public companies’ earnings reports. [TechCrunch]
Hype Bubble?
Arizona State University researchers revealed that Chain-of-Thought reasoning in LLMs is primarily sophisticated pattern matching, not genuine intelligence, with performance collapsing outside training data distributions. [VentureBeat]
The Commonwealth Bank of Australia abandoned its plan to replace customer service jobs with AI after union pressure and disputes over call volume claims. [Bloomberg]
Legal tech vendors are hyping agentic AI, but the technology is mostly sophisticated automation that won’t actually make autonomous decisions for lawyers. [Above the Law]
Microsoft has leveraged AI tools like GitHub Copilot while simultaneously laying off thousands of employees, raising questions about the true cost and effectiveness of AI technology. [The Register]
MIT’s Project NANDA revealed a ‘shadow AI economy’ where employees widely adopt personal AI tools, outpacing slow corporate initiatives and driving unprecedented productivity gains. [VentureBeat]
OpenAI's Sam Altman acknowledged the potential AI market bubble while affirming AI’s transformative importance and long-term significance. [CNBC]
Research by MIT reveals that despite widespread AI adoption, 95% of enterprise pilots fail to deliver significant revenue growth due to implementation challenges and misaligned strategies. [Fortune]
Big Iron
Amazon has surpassed US$100B in data center spending, exceeding the GDP of several small countries and outpacing its tech rivals’ investments. [TechRadar]
Dell and Hewlett Packard have expanded their AI infrastructure lines with new Nvidia-powered systems designed to help enterprises move from AI experimentation to production. [siliconANGLE]
EdgeConneX has partnered with Lambda to build a 30+ MW AI-focused data center infrastructure in Chicago and Atlanta, featuring high-density, hybrid-cooled facilities. [Business Wire]
Hewlett Packard is preparing to ship Nvidia Blackwell-powered GPU servers amid potential delivery challenges due to high demand from hyperscalers and export restrictions. [Computer Weekly]
OpenAI has considered offering AI infrastructure services in the future, drawing inspiration from Amazon's cloud computing rental model. [Bloomberg]
Oracle has aggressively invested over US$1B annually in AI-focused cloud infrastructure, including a massive Texas data center powered by gas generators. [Gizmodo]
The Chips are Up
Intel is exploring a potential US government investment of around 10% through converting Chips Act grants into equity. [Bloomberg]
A survey by Liquid Web reveals AI teams are increasingly exploring non-Nvidia hardware options due to cost, availability, and cloud adoption constraints. [TechRadar]
Malaysia has imposed a Strategic Trade Permit requirement for US-origin high-performance AI chips, aiming to tighten export controls and align with international regulations. [Electropages]
Nvidia is reportedly developing a cut-down B30A GPU for the Chinese market, potentially circumventing export controls while offering performance rivaling its H100 and H200 accelerators. [The Register]
SoftBank Group invested US$2B in Intel, agreeing to purchase common stock at US$23 per share. [DealStreetAsia]
Warm Bodies
Arm has reportedly recruited Rami Sinno, a top AWS chip designer who previously led Trainium and Inferentia chip development. [The Register]
AWS CEO Matt Garman strongly criticized the idea of replacing junior staff with AI, emphasizing the importance of hiring and developing young talent. [The Register]
Layoffs.fyi tracked over 150,000 tech job cuts across 549 companies in 2024, with more than 22,000 workers already impacted in 2025’s first two months. [TechCrunch]
Meta froze AI hiring after aggressively recruiting over 50 researchers and restructuring its AI organization into four new groups. [TechCrunch]
A survey by Resume Now revealed strong worker support for AI in management, with clear boundaries around emotional intelligence, ethics, and interpersonal skills. [PRWeb]
The World Economic Forum found that AI is most likely to replace jobs in industries with high-quality, accessible data, such as finance and customer support. [Gizmodo]
Consumer AI
Curio's AI-powered plushie Grem failed to impress The New York Times critic, who saw it as a problematic substitute for parental interaction. [TechCrunch]
Deutsche Telekom launched the T Phone 3 and T Tablet 2, featuring a Magenta AI operating system with Perplexity AI integration across multiple European markets. [GSM Arena]
Dex has developed an AI-powered language-learning device for children that encourages interactive, real-world learning through object recognition and storytelling. [TechCrunch]
Fitbit has redesigned its app with an AI-powered Personal Health Coach, offering personalized health guidance and customizable features for Premium subscribers. [Wired]
Google is launching Gemini for Home, a powerful AI-driven voice assistant for Nest devices that promises more natural interactions and advanced capabilities. [The Verge]
Google has unveiled the Pixel 10 series, introducing numerous AI-powered features across photography, communication, and productivity tools powered by its Tensor G5 processor. [TechCrunch]
And Google unveiled the Pixel Buds 2a and Pixel Buds Pro 2 with enhanced Gemini AI features, noise cancellation, and hands-free gesture controls at its recent event. [TechCrunch]
Halo has developed AI-powered smart glasses that continuously record and transcribe conversations, providing real-time information to the wearer. [TechCrunch]
HTC launched the VIVE Eagle, a lightweight AI headset with voice-activated features, local data processing, and Zeiss solar lenses, initially available in Taiwan. [The Decoder]
Meta's next smart glasses might debut at a more affordable $800 price point, potentially boosting early consumer adoption and interest. [TechRadar]
Samsung Galaxy Buds 3 FE arrived with Gemini AI and real-time translation. [TechRadar]
It’s Only a Model
Alibaba's Qwen-Image-Edit, an open-source AI model, enables sophisticated Photoshop-like image editing through text-based instructions with high semantic and visual fidelity. [VentureBeat]
ByteDance's Seed Team released Seed-OSS-36B, a powerful open-source LLM with three variants, offering state-of-the-art performance across math, coding, and long-context reasoning. [VentureBeat]
GPT Proto has launched an affordable Qwen image edit API, offering advanced AI image manipulation and generation capabilities at competitive prices for developers and businesses. [EIN Presswire]
Nous Research found that open-source AI models consume significantly more computing resources than closed-source models when performing identical tasks. [Gizmodo]
The NSF and Nvidia have partnered with AI2 to develop open, domain-specific AI models aimed at advancing scientific research and innovation. [siliconANGLE]
Nvidia has unveiled Granary, a massive multilingual speech dataset, and two advanced AI models that significantly enhance automatic speech recognition and translation for European languages. [Marktechpost Media]
And Nvidia has released Nemotron-Nano-9B-V2, a small language model with hybrid architecture, reasoning toggle, and high performance across multiple languages and benchmarks. [VentureBeat]
OpenAI's gpt-oss open-weights AI model was quickly modified by a researcher to remove reasoning constraints, creating a more uncensored version. [VentureBeat]
OpenAI's GPT-5 appears to have a hidden system prompt that includes the current date and instructions about response verbosity, which cannot be overridden by user-specified system prompts. [Simon Willison]
Tencent has developed X-Omni, an innovative AI image generation model using reinforcement learning to improve hybrid system performance, especially for text rendering. [The Decoder]
Whose Data?
OpenAI removed the option to make ChatGPT conversations publicly searchable after users accidentally shared sensitive queries that could potentially be exposed through data breaches or legal proceedings. [The Register]
OpenAI and other AI companies are forging global partnerships to gather diverse, structured consumer data for improving their AI models’ accuracy and performance. [Rest of World]
Otter.ai faces a class-action lawsuit alleging it secretly recorded and used private work conversations for AI training without participants’ explicit consent. [NPR]
Transcend expanded its ‘Do Not Train’ and ‘Deep Deletion’ capabilities to help B2B AI companies ensure responsible data handling and enterprise trust. [Business Wire]
Veritone has partnered with Newsmax to help the news network monetize and search its extensive 20-year content archive using AI-powered digital media technology. [Business Wire]
xAI‘s Grok chatbot conversations were easily indexed by search engines, revealing users’ requests for potentially harmful or illegal information. [TechCrunch]
The LLM Ecosystem
Fireflies is the first AI meeting tool in Claude’s MCP Directory, enabling teams to instantly query meeting data through Claude with one-click privacy-preserving integration. [Business Wire]
Geek Haus launched TutorFlow Code Editor, an AI-powered programming environment supporting 15+ languages with instant execution and context-aware assistance. [EIN Presswire]
GEPA, a novel AI optimization method, significantly improves language model performance by using natural language reflection and genetic prompt evolution, reducing computational costs and trial runs. [VentureBeat]
Inclusion AI has proposed Inclusion Arena, a novel leaderboard that ranks LLMs based on real-world user preferences through pairwise comparisons in AI-powered applications. [VentureBeat]
Reprise launched an AI-powered demo creation system that automatically captures, maintains, and populates enterprise software demo environments with unprecedented efficiency. [Business Wire]
Workato has launched a free Developer Sandbox, providing developers instant access to its full AI-powered platform for building and testing innovative applications. [Business Wire]
Agentic AI
Confluent has introduced Streaming Agents, a new capability in Confluent Cloud that simplifies building and scaling AI agents using real-time data processing and integration. [Business Wire]
CoreSpeed launched an Agent-native Platform-as-a-Service to simplify AI agent deployment, offering secure, high-performance infrastructure for developers. [EIN Presswire]
Deel launched an AI Workforce platform enabling businesses to create, manage, and deploy specialized AI agents for HR and payroll tasks across 150+ countries. [Business Wire]
Google has developed an agentic security operations center to streamline incident management and enhance AI security capabilities across its product suite. [Computer Weekly]
Javelin launched MCP Security, a defense-in-depth solution that scans and enforces real-time policies to protect enterprises from AI security risks. [Business Wire]
Kasisto launched KAIgentic, an agentic AI platform designed to provide secure, compliant, and intelligent banking experiences across customer and employee interactions. [Business Wire]
Orbitype launched Agentic Cloud OS, enabling seamless AI integration into cloud infrastructures with autonomous agents working alongside humans. [GlobeNewswire]
Paradigm has developed an AI-powered spreadsheet with over 5,000 agents that can crawl the internet and fill out information across different AI models. [TechCrunch]
Striim introduced MCP AgentLink, a solution providing secure, real-time data replicas for enterprise AI agents, enabling safe and compliant data access without risking production systems. [GlobeNewswire]
Workato expanded its ONE platform, unifying agent development, multi-agent orchestration, and enterprise data fabric to enable autonomous, intelligent business execution. [Business Wire]
Workato launched Action Board, an AI management tool enabling enterprises to track, engage with, and control AI agents across different business functions. [Business Wire]
And Workato launched production-grade AI Genies that automate and optimize business functions across enterprises, transforming how companies operate. [Business Wire]
Other LLM Sightings
Aventis launched a generative AI training coach that enables companies of all sizes to create personalized, adaptive corporate learning platforms with minimal setup. [EIN Presswire]
Big Purple Dot released a proprietary AI feature in its CRM that transcribes calls, automates compliance, and transforms workflow for mortgage professionals. [Business Wire]
HootRecruit launched an AI-powered talent sourcing platform that enables recruiters to discover and assess qualified candidates quickly through continuous, interactive refinement. [EIN Presswire]
Industry Navigator has added AI-powered tools to its platform, partnering with Cloverleaf AI to enhance public sector market intelligence and bid review capabilities. [EIN Presswire]
Interact expanded its Employee Experience Platform with AI-powered features designed to enhance internal communications, employee self-service, and organizational efficiency. [GlobeNewswire]
KPMG developed a 100-page prompt for its TaxBot that generates comprehensive tax advice in one day, replacing a two-week human process without job losses. [The Register]
MailGenius launched an AI Email Grader and Rewriter tool that diagnoses and optimizes email performance using advanced AI trained on millions of high-converting campaigns. [EIN Presswire]
NinjaTech AI has partnered with Cerebras Systems to develop SuperNinja Fast Deep Research, delivering AI-powered information analysis up to 5x faster than competitors. [EIN Presswire]
Noded AI has launched an AI-native platform that integrates customer data across tools, enabling Customer Success teams to streamline operations and drive outcomes. [EIN Presswire]
PageUp has launched advanced AI features in its recruitment platform, aiming to streamline hiring processes and enhance candidate experiences through intelligent, time-saving tools. [GlobeNewswire]
PCMS has partnered with Gain Life to provide AI-driven guidance and monitoring solutions that streamline insurance claims processes and enhance customer experiences. [EIN Presswire]
Qlik Answers enables enterprise AI adoption by providing explainable, actionable intelligence from unstructured data through Amazon Bedrock integration. [Business Wire]
RealReports unveiled Research Mode, an AI-powered tool that provides deep, actionable property intelligence for real estate professionals in seconds. [EIN Presswire]
Recruit CRM enhanced its recruitment platform with AI-powered features that streamline candidate sourcing, data enrichment, and application processes. [EIN Presswire]
Trust3 AI unveiled Trust3 IQ, a groundbreaking platform designed to enhance AI accuracy through a Universal Semantic Layer and advanced Context Engine. [EIN Presswire]
The UK government is developing AI-powered virtual agents to help citizens navigate bureaucratic processes, with plans to prototype and potentially roll out these services nationwide by late 2027. [The Register]
Risks and Responses
Anthropic updated Claude’s usage policy to prohibit development of dangerous weapons and address potential cybersecurity risks while loosening restrictions on political content. [The Verge]
And Anthropic introduced a feature enabling Claude AI models to terminate conversations in rare, extreme cases of harmful interactions. [Engadget]
DeepSeek has raised significant cybersecurity concerns among UK professionals, with 60% believing its use could increase organizational cyber attacks due to data sovereignty risks. [ITPro]
A report by Fastly reveals AI crawlers, primarily from Meta and OpenAI, are overwhelming websites with massive, unsustainable traffic volumes, raising concerns about digital infrastructure and content creator rights. [The Register]
Geoffrey Hinton warned that tech industry leaders prioritize short-term profits and immediate research outcomes over long-term AI consequences and potential existential risks. [Fortune]
Google launched Gemini Storybook, an AI-powered app generating personalized children’s stories, which has drawn mixed reactions from parents, tech experts, and authors concerned about its quality and ethical implications. [Publishers Weekly]
Google's AI search results have become vulnerable to scammers inserting fake customer service numbers that trick users into sharing personal and financial information. [ZDNet]
Lenovo discovered a critical vulnerability in its Lena AI chatbot that could allow hackers to inject malicious code and steal sensitive data through cross-site scripting. [ITPro]
Meta faces a potential Senate investigation after leaked documents revealed its AI chatbots were permitted to engage in inappropriate conversations with children. [TechCrunch]
Microsoft's Mustafa Suleyman warns that AI chatbots could convincingly mimic consciousness, potentially tricking users into forming dangerous emotional attachments. [TechRadar]
OpenAI’s ChatGPT 5 demonstrated a significant advancement by admitting uncertainty and saying ‘I don’t know’ when it cannot reliably answer a question, potentially improving AI trustworthiness. [TechRadar]
Perplexity's Comet browser was vulnerable to prompt injection attacks that could manipulate its AI to exfiltrate user account passwords from web pages. [The Register]
Researchers demonstrated that AI chatbots can be easily manipulated through system prompts to covertly extract personal information from unsuspecting users. [The Register]
Standard Bank warned clients about AI-powered spoofing scams using advanced voice cloning and phishing techniques to impersonate bank officials and steal sensitive information. [TechCentral]
Texas Attorney General Ken Paxton launched an investigation into Meta AI and Character.ai for potentially misleading users about their mental health chatbot capabilities and data practices. [TechCrunch]
xAI's Grok chatbot exposed system prompts revealing controversial AI personas, including a conspiracy-theorist mode designed to engage users in fringe beliefs. [TechCrunch]
Regulation
The Australian government is backing away from its AI regulation plans. [Startup Daily]
Tech companies are pressing the White House and Congress to block states from regulating AI through proposed federal legislation. [PYMNTS]
US Senators Gallego and Justice introduced legislation to restrict offshore call center work and mandate AI and location disclosures in customer service interactions. [MultiLingual]
Environmental Issues
EDB's Postgres AI platform dramatically reduced data center emissions and energy consumption by up to 87% and 81% for large enterprise customers. [Business Wire]
Google has released the first detailed report on Gemini AI’s energy and water consumption, revealing lower-than-expected usage per query. [ZDNet]
Conversational AI
8x8 has enhanced its CX platform with AI-driven innovations, enabling businesses to personalize service, streamline operations, and improve customer engagement across multiple channels. [Business Wire]
Agenda Hero launched an AI platform with Magic Chat and Magic Schedules to transform manual calendar work into effortless, automated event creation and sharing. [Business Wire]
BetSymphony has launched a conversational AI sportsbook interface that personalizes, enables multilingual voice interaction, and enhances player engagement. [PRWeb]
Girl Boss Financial launched DontMissThatCall.com, an affordable AI receptionist service enabling solopreneurs to manage calls and leads professionally. [EIN Presswire]
Huawei has unveiled comprehensive AI contact centre solutions, aiming to transform customer service operations through advanced technological innovations. [TechCentral]
IntelePeer has made SmartAnalytics available in the Microsoft Azure Marketplace, enabling customers to access advanced conversational AI analytics and insights through a trusted cloud platform. [Business Wire]
Meera.ai has introduced an Adaptive Agent AI texting tool that enables dynamic, compliant SMS conversations while maintaining precise control over messaging content. [EIN Presswire]
Servion launched JourneyWorCX, a next-generation CX framework that transforms customer interactions from fragmented touchpoints to seamless, AI-powered experiences. [PRWeb]
SoundCurve launched an AI-powered receptionist that handles calls, qualifies leads, and books appointments 24/7 for businesses. [EIN Presswire]
Takeuchi-US partnered with Circuitry.ai to launch an AI-powered Service Advisor on its website, providing instant, accurate product support to customers. [Business Wire]
Workera released Talk for Sage, an AI-driven voice mentor that enables real-time, conversational skills verification through adaptive, nuanced assessments. [Speech Technology Magazine]
Zoom has integrated an AI-powered Virtual Agent concierge with its Phone platform, offering 24/7 personalized call support across multiple industries and languages. [Computer Weekly]
Be Real
Anthropic is exploring the emerging field of ‘AI welfare’, studying potential consciousness in AI models and their rights, despite Microsoft CEO Mustafa Suleyman’s criticism of the concept. [TechCrunch]
A senior Australian lawyer apologized to a judge for submitting AI-generated fake quotes and non-existent case judgments in a murder trial. [Associated Press]
Brother launched an AI-powered virtual salesperson on Taobao that streams 24/7, selling printers and outperforming human salespeople in livestream sales. [Wired]
Character.ai's tech chief predicted AI companions will soon help people navigate life, despite facing lawsuits over potential harmful interactions with chatbots. [The Daily Star]
OpenAI swiftly restored GPT-4o to paid users after an unexpected model replacement triggered widespread emotional distress among users who had formed deep attachments to the AI. [MIT Technology Review]
Voice News
Behavox expanded its Quantum Voice product to support multilingual voice intelligence across nine Asian languages, enhancing compliance monitoring and risk detection. [Business Wire]
CivAI has launched an AI voice game called ‘We Need to Talk’ that allows users to interact with emotionally responsive AI characters to demonstrate AI’s potential impact. [Speech Technology Magazine]
Deepgram has partnered with AWS to expand its voice AI platform, offering enhanced integration, deployment options, and capabilities for enterprise customers. [Speech Technology Magazine]
Google has introduced AI-powered photo editing features in Google Photos, allowing users to edit images through natural language commands on Pixel 10 devices. [TechCrunch]
And Google has unveiled a real-time translation feature using Gemini Nano that deepfakes voices and translates speech accurately across languages on-device. [Gizmodo]
Krisp launched AI Voice Translation v2.0, enabling real-time multilingual communication with advanced features like Synchronous mode and auto-scoring for enhanced call center experiences. [Business Wire]
OpenAI is retiring ChatGPT’s Standard Voice Mode in September, causing user frustration over the loss of its perceived warmth and conversational style. [TechRadar]
ValidSoft has partnered with Reality Defender to provide a multi-layered defense against voice deepfakes using real-time biometric authentication and detection technologies. [Speech Technology Magazine]
Voicing AI has developed a sub-70 millisecond voice response model with high naturalness, speed, and multilingual capabilities for enterprise voice automation. [Speech Technology Magazine]
Document AI
Adobe has launched Acrobat Studio, a comprehensive document management platform integrating AI-powered tools, Acrobat Pro, and Adobe Express to enhance productivity and creativity. [TechRadar]
Ancora Software has released a significant platform update with model-less classification, zero-touch input, and enhanced global automation capabilities for intelligent document processing. [EIN Presswire]
Epiq has partnered with IRIS (Canon) to enhance document processing through AI-powered intelligent capture, workflow automation, and data-driven insights for organizations. [GlobeNewswire]
HuLoop Automation has released version 7 of its Unified Work Intelligence Platform, featuring the new AI assistant HuGO Sidekick to accelerate automation building. [PR Newswire]
MaestroX has launched MaestroX Connect, an AI-powered platform that streamlines title searches, vendor management, and integrations for real estate professionals. [PRWeb]
Offorte has developed AI-powered proposal software that enables sales teams to create and send proposals through voice commands using its innovative MCP technology. [EIN Presswire]
Translation
Campbell Police Department has implemented AI-powered body cameras that provide real-time translation during traffic stops. [NBC Bay Area]
Meta has launched an AI-powered voice translation feature for Facebook and Instagram, enabling creators to translate content into other languages. [TechCrunch]
Signvrse has developed Terp 360, an AI-powered platform translating text and speech into Kenyan Sign Language using realistic 3D avatars. [MultiLingual]
Smartling launched an MCP server enabling seamless translation capabilities within AI platforms, bridging workflow gaps for software teams. [PRWeb]
Timekettle's T1 AI translator is a lightweight, offline-capable device offering global translation for frequent travellers. [The Register]
Search
Ecosia proposed a unique 10-year stewardship of Google's Chrome browser, offering to use potential revenues for climate projects while maintaining Google’s intellectual property. [TechCrunch]
Google expanded its AI Mode to 180 new countries, adding personalized search capabilities and collaborative features for users. [TechCrunch]
And Google has enhanced its AI Mode to help users find restaurant reservations and is working on expanding to concert tickets with personalized search capabilities. [TechRadar]
Google AI Overviews has led to a significant drop in publisher referral traffic, with some sites experiencing up to a 25% decline in search-driven visits. [Digiday]
OpenAI has reduced ChatGPT’s website referral traffic by 52%, with Reddit and Wikipedia now dominating citations through answer-focused content. [Search Engine Land]
Real Chemistry launched HealthGEO, an AI search solution helping life sciences communicators monitor, manage, and shape their representation in generative AI platforms. [Business Wire]
Reddit is developing an AI-powered search engine that synthesizes answers from its platform’s discussions, aiming to become a primary search destination. [Mashable]
SEO.co launched LLM SEO services to help brands gain visibility and citations in AI-generated search results across various platforms. [EIN Presswire]
Writing Assistance
Grammarly launched eight specialized AI agents and a new writing surface to provide targeted, context-aware assistance for students and professionals across various writing tasks. [Business Wire]
AI in Journalism
Pangram has partnered with Qwoted to provide journalists with a reliable AI text detection tool that helps them identify and filter out artificially generated content. [Business Wire]
Tasaka Digital's Guy Tasaka advocates for media companies to adopt AI tools with retrieval-augmented generation to transform content creation and operational efficiency. [Editor & Publisher Magazine]
Wired and other publications removed articles by freelance journalist Margaux Blanchard after discovering they were likely AI-generated fiction with unverifiable sources. [Press Gazette]
Health Tech
Aptarro and aiHealth have partnered to enhance revenue cycle management by combining AI-powered medical coding and charge correction technologies. [Business Wire]
Eleos launched an AI scanner that proactively detects potential Medicaid eligibility changes during client sessions. [GlobeNewswire]
Epic has unveiled AI-powered features for its electronic health records system, including an AI scribe and assistants for doctors, patients, and staff. [STAT]
Inovalon has partnered with AWS to develop an AI solution helping health plans prepare for upcoming Medicare Advantage RADV audit requirements by improving risk score accuracy and documentation validation. [Business Wire]
LMArena and DataTecnica have proposed expanding BiomedArena, a leaderboard to rigorously test AI models’ medical knowledge and performance across biomedical research tasks. [ZDNet]
Talkdesk has been selected by Johns Hopkins Health System to implement its Healthcare Experience Cloud platform for modernizing contact center operations and enhancing patient communications. [GlobeNewswire]
And Talkdesk has integrated its Advanced Dialer with Epic, enabling automated patient outreach and streamlined communication through AI-powered contact center technology. [GlobeNewswire]
The UK’s NHS is piloting an AI-powered discharge platform to help healthcare professionals quickly process patient release documents and free up hospital beds. [Sky News]
A study has found that people trust AI-generated medical advice more than doctors’ responses, even when the AI advice is inaccurate or potentially harmful. [ZDNet]
Legal Tech
AI CERTs has launched a comprehensive Legal Certification Bundle to equip legal professionals with essential AI skills, ethics, and strategic insights for the digital age. [EIN Presswire]
Anthropic’s Claude received high praise from Supreme Court Justice Elena Kagan for its exceptional legal analysis, signaling potential AI usefulness in the legal profession despite ongoing challenges. [ZDNet]
Aracor AI has integrated OpenAI's GPT-OSS 120B open-source LLM to provide secure, customizable document review for deal teams with enhanced privacy controls. [Artificial Lawyer]
Array has relaunched Acumen, its self-service eDiscovery solution, enabling legal teams to independently manage Relativity environments and make real-time decisions. [PRWeb]
Goldman Sachs reported that approximately 17.2% of US lawyers could be at risk of AI automation, representing about 228,000 legal professionals. [Artificial Lawyer]
Harvey has partnered with Ironclad to develop AI-powered contract workflow solutions, leveraging their mutual clients and complementary technological capabilities. [Artificial Lawyer]
Justice AV Solutions unveiled Suite 9, AI-driven courtroom recording software that offers browser-based controls, real-time translation, and instant AI-generated transcripts. [EIN Presswire]
LOWCODEMINDS has launched NeoLegal, a no-code legal matter management solution designed to streamline legal operations and enhance enterprise workflow efficiency. [PRWeb]
Pre/Dicta expanded its legal analytics platform with appellate forecasting, enhanced biographical intelligence, and comparative prediction features across federal and California cases. [LawSites]
A survey by SpotDraft revealed that legal teams could dramatically improve contract efficiency by 73% through AI, yet most continue using manual processes. [Business Wire]
USF’s School of Law has partnered with Accordance and Anthropic to integrate generative AI tools into its curriculum, preparing students for a tech-driven legal landscape. [PRWeb]
Ed Tech
Alef Education has partnered with Liquid AI to develop generative AI solutions that enhance educational workflows. [Business Wire]
CodeSignal launched Cosmo, a mobile AI-powered learning app delivering bite-sized courses across generative AI, coding, marketing, finance, and leadership skills. [VentureBeat]
OpenAI’s ChatGPT is transforming education as teachers cautiously integrate AI tools to personalize learning, manage workloads, and teach critical technological literacy. [Wired]
Funding
Anthropic is nearing a US$170B valuation through an oversubscribed US$5B funding round led by Iconiq Capital and backed by international investors. [Tech Funding News]
Character.ai is considering a potential sale or raising capital to address rising operational costs and challenges in the competitive AI chatbot market. [PYMNTS]
AI coding startup Cognition raised nearly US$500m, led by Founders Fund, boosting its valuation to US$9.8 billion. [The Wall Street Journal]
Databricks secured a US$1B funding round at a US$100B valuation, focusing on developing AI agent database and platform technologies. [TechCrunch]
Firecrawl secured a Series A funding round for its popular open-source web crawler. [TechCrunch]
Genow, a German AI startup, secured €1.65M in seed funding to develop its Wingman knowledge management platform for enterprise knowledge transformation. [Silicon Canals]
Infinity Loop, an AI-native contract intelligence platform, secured US$5m in Seed funding to help enterprises optimize vendor contract spending through advanced AI analysis. [FinSMEs]
Lambda secured a US$480m Series D investment to expand its GPU cloud infrastructure and support AI developers’ growing computational needs. [Tech Funding News]
Marshmallow, a UK AI-native insurance broker, secured oversubscribed funding to revolutionize the traditional insurance model. [Silicon Canals]
Meshed raised £950,000 in a pre-seed round to provide AI-powered digital insurance brokerage services for small and medium-sized businesses. [Tech.eu]
Professional.me raised US$3.1m to develop micro-LLMs that transform hiring by creating personalized, context-aware profiles for employers and professionals. [EIN Presswire]
ShiftUp secured US$3m in seed funding to develop an AI-powered sales intelligence platform that autonomously guides and enhances sales team performance. [Business Wire]
TensorZero has raised US$7.3m in seed funding to develop open-source infrastructure for LLM applications, addressing enterprise AI deployment challenges. [VentureBeat]
TinyFish launched with US$47m to develop enterprise web agents that automate complex business workflows across thousands of platforms simultaneously. [Business Wire]
Upstage raised US$45m to expand its AI-powered document intelligence platform to the US market, targeting insurance industry digitization with advanced document parsing technology. [siliconANGLE]
Yourway Learning secured US$9m to expand its AI-powered K-12 education platform, focusing on classroom technology and professional development. [FinSMEs]
Zipline AI raised US$7m to develop an infrastructure platform that simplifies and accelerates AI application development by unifying complex data workflows. [Business Wire]
Acquisitions
Capacity acquired Call Criteria and Verbio Technologies to enhance its AI-powered support automation platform for contact centers, backed by over US$92m in new investments. [Smart Customer Service]
CitiusTech acquired Health Data Movers, enhancing its Epic implementation capabilities and expanding healthcare technology solutions through strategic integration expertise. [Business Wire]
Folks acquired Glow Talents, an AI-powered recruiting software, to enhance its HR solutions and streamline talent acquisition for small and medium-sized businesses. [GlobeNewswire]
There’s More
Fastly's research has revealed that AI crawlers, predominantly from Meta, constitute nearly 80% of AI bot traffic, raising concerns about web security and infrastructure strain. [Business Wire]
Google has offered its AI services to government agencies for just $0.47 per agency through 2026, raising concerns about potential vendor lock-in and future pricing. [The Register]
Kalshi and other prediction markets have seen a surge in AI-related betting, with traders like Foster McCoy making significant profits by wagering on AI model performance and industry developments. [The Wall Street Journal]
OpenAI’s ChatGPT misidentified a simple duck image as the famous duck-rabbit optical illusion, revealing AI vision models’ tendency to see non-existent patterns in images. [The Register]
The Back Cover
Here’s Justin Johnson with a cover of Fleetwood Mac’s The Chain.
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!