This Week in NLP #351
Keep up with what happened in NLP in the week ending Friday 8th August 2025.
Above the Fold
Overwhelmed? Here’s our pick for five things to know about this week.
Apple has formed a new team to develop an AI-powered ‘answer engine’ that could enhance search capabilities across its products. [TechCrunch]
Anthropic has released Claude Opus 4.1, achieving top performance in software engineering tasks and expanding its coding market dominance ahead of expected OpenAI competition. [VentureBeat]
Mistral AI is speculated to be a potential acquisition target for Apple, despite its founders’ commitment to European AI sovereignty and independence. [Sifted]
OpenAI has launched GPT-5, a new lineup of LLMs with improved reasoning, reduced hallucinations, and enhanced capabilities across various tasks and domains. [VentureBeat]
And OpenAI has released its first open-weight language models since 2019, offering two sizes under a permissive license to compete with Chinese models and support research needs. [MIT Technology Review]
Now read on for everything else that happened in NLP this week.
Making News:
What makes this newsletter different?
This Week’s Topics:
If you’re reading this in a desktop web browser, you can access a navigation menu at the left-hand edge of this window.
The Generative AI Wars
Apple has signaled readiness to spend more on AI through data center expansion or potential acquisitions to catch up with rivals like Microsoft and Google. [Yahoo Finance]
Apple's CEO Tim Cook rallied employees around the company’s AI potential, emphasizing its late-but-innovative approach to transformative technologies. [Yahoo Finance]
Anthropic revoked OpenAI's API access to Claude after alleging violations of its terms of service, particularly regarding potential competitive model development. [Wired]
Google has launched Jules, an AI coding agent powered by Gemini 2.5 Pro, offering asynchronous code improvements through GitHub integration and tiered pricing plans. [TechCrunch]
Google researchers developed TTD-DR, an AI research framework that uses diffusion and evolutionary algorithms to generate more comprehensive and accurate reports by iteratively refining drafts. [VentureBeat]
Microsoft envisions a future Windows where AI enables multimodal, conversational interactions, moving beyond traditional mouse and keyboard interfaces. [TechRadar]
OpenAI has raised US$8.3B at a US$300B valuation, attracting major investors and surpassing its revenue and user growth projections. [TechCrunch]
And OpenAI is exploring a potential stock sale that could value the company at US$500B, allowing employees to cash out shares. [TechCentral]
OpenAI partnered with the US government to provide its AI models to federal employees at a nominal cost of $1 for the next year. [Wired]
OpenAI’s ChatGPT has approached 700 million weekly active users, marking significant growth and increased engagement since its launch. [TechCrunch]
And OpenAI has partnered with AWS to offer its new open-weight reasoning models on AWS’s Bedrock and SageMaker AI platforms. [TechCrunch]
X is planning to introduce ads in Grok’s AI responses to help support its struggling advertising business. [TechCrunch]
AI Supremacy
DeepSeek's breakthrough in AI sparked a transformative wave of technological innovation and adoption across various sectors in China. [Al Mayadeen]
Nvidia denied Chinese accusations of a hardware ‘kill switch’ in its AI chips, emphasizing the security risks and potential harm to national interests. [CNBC]
Nvidia's RTX 5090 gaming GPUs are being converted into AI accelerators by Chinese manufacturers, bypassing export restrictions through custom PCB modifications. [TechRadar]
Nvidia chips were illegally shipped to China by two Chinese nationals in California, bypassing export restrictions and facing potential legal consequences. [Reuters]
Shanghai has unveiled a US$139m AI subsidy programme, intensifying China’s technological competition with the United States and aiming to reduce dependence on foreign technology. [AI News]
The US is exploring ways to enhance semiconductor location-tracking capabilities to restrict technology transfers to China. [Yahoo Finance]
And Trump’s administration is considering semiconductor tariffs and AI chip export restrictions. [TechCrunch]
Sovereign AI
AWS is launching a European Sovereign Cloud service exclusively operated by EU citizens to address data sovereignty concerns and provide independent cloud infrastructure. [ITPro]
Ecosia has partnered with Qwant to develop an independent European search index, marking a significant step towards digital sovereignty and privacy-first technology. [Tech.eu]
Google has launched a US$37m AI investment in Africa, focusing on grants for food security, language models, and education across key countries. [TechTrendsKE]
The Tony Blair Institute has urged the UK to prioritize AI infrastructure and deployment, rather than competing in AI model training, to remain economically competitive. [The Register]
The UAE and Gulf states have strategically invested billions in AI infrastructure, positioning themselves as potential global technology hubs with US support. [BBC]
Ukraine has launched the AI Factory, a state infrastructure for AI development, aiming to create a sovereign LLM and become a top-3 AI nation by 2030. [Tech.eu]
Feature Creeps
Anthropic has developed persona vectors, a technique to identify, monitor, and control undesirable personality traits in LLMs through automated analysis of their internal activations. [VentureBeat]
Character.ai has launched an innovative, interactive social feed that transforms AI content consumption into a creative, collaborative experience for users. [Character.AI]
Google’s Gemini has introduced a feature enabling users to create personalized, illustrated storybooks with read-aloud narration using AI and personal photos. [9to5Google]
Microsoft may be exploring an AI-powered ‘Taskbar Companion’ in Windows 11. [TechRadar]
OpenAI has enhanced ChatGPT with customizable chat colors, new personalities, improved Google app integration, and Advanced Voice Mode accessible to all users. [ZDNet]
xAI is developing a Teams plan for Grok, enabling collaborative workspaces with administrative controls and separate team-specific chat histories. [TestingCatalog]
And xAI launched Grok Imagine, an AI image and video generator with a ‘spicy mode’ for NSFW content, available to X Premium+ subscribers. [TechCrunch]
Hype Bubble?
AI companies have been pouring unprecedented amounts of money into infrastructure, potentially propping up the entire US economy and risking a massive bubble comparable to the dot-com crisis. [Futurism]
And tech companies are revising AI marketing claims to address regulatory scrutiny and provide more transparent, evidence-based representations of their technological capabilities. [WebProNews]
Balaji Srinivasan and David Sacks have argued that AI requires human guidance and verification, creating new job opportunities rather than wholesale job displacement. [Benzinga]
Gartner's 2025 Hype Cycle report highlights AI agents at peak expectations, emphasizing the need for strategic, precise application and trust management. [ZDNet]
OpenAI’s ChatGPT agent, despite ambitious promises, has failed to deliver reliable performance across complex tasks, revealing significant limitations in current AI agent capabilities. [Substack]
The US Federal Reserve predicted generative AI will significantly boost productivity, though its economic impact will be gradual and fraught with adoption challenges. [Gizmodo]
Big Iron
Foxconn plans to convert a former Ohio electric-truck factory into a cloud computing hardware plant for AI applications. [The Wall Street Journal]
G42's Stargate AI campus highlights the Gulf’s trillion-dollar AI ambitions, which face significant challenges from extreme water scarcity and data center cooling demands. [Rest of World]
Google has agreed to pause AI data center workloads during electricity grid demand spikes to help manage power consumption more efficiently. [ITPro]
Meta has reclassified US$2.04B in data center assets for sale, signaling its strategy to share AI infrastructure costs with external partners. [Reuters]
The Chips are Up
Cerebras Systems supports OpenAI's gpt-oss-120B open-weight model, achieving world-record inference speeds of 3,000 tokens per second on its AI infrastructure. [Business Wire]
Nvidia AI chips find extensive underground repair in China, with dozens of firms fixing thousands of smuggled GPUs annually despite export restrictions. [TechRadar]
Nvidia faces delays in obtaining US licenses to sell AI chips to China due to Department of Commerce bureaucratic challenges. [TechCrunch]
Samsung secured a US$16.5B contract to produce AI semiconductors for Tesla's self-driving vehicles through 2033, bolstering its struggling foundry unit. [TechRepublic]
Warm Bodies
Anthropic has resisted Meta's US$100m poaching offers, with CEO Dario Amodei refusing to match salaries to retain employees. [Yahoo Finance]
Meta has recruited 24-year-old AI researcher Matt Deitke with a US$250m package, highlighting escalating talent wars and concerns about inequality in tech. [Finance Monthly]
Microsoft has been aggressively recruiting top AI talent from Google to bolster its AI capabilities. [The Wall Street Journal]
xAI has aggressively recruited top AI talent from Meta and other tech companies while positioning itself as a merit-based alternative with high growth potential. [Mashable]
Consumer AI
Alibaba unveiled Quark AI Glasses, a smart eyewear device running on its Qwen language model, set to compete in China’s AI wearables market by late 2025. [eWeek]
Amazon has launched Alexa+, a generative AI-powered digital assistant that aims to enhance smart home interactions by offering more advanced scheduling, information retrieval, and potential task-completion capabilities. [TechCrunch]
Amazon CEO Andy Jassy has envisioned integrating ads into Alexa+ conversations, seeing potential for product discovery and revenue generation through the AI-powered digital assistant. [TechCrunch]
Brilliant Labs launched Halo smart glasses with advanced AI capabilities, including a memory system and Noa AI agent, priced at $299. [The Verge]
Google is reportedly developing a Gemini-powered Conversational Photo Editing mode for the Pixel 10 series, allowing voice-controlled image adjustments. [TechRadar]
Google is poised to emphasize AI integration, particularly Gemini, across its upcoming Pixel devices during the August 20 launch event. [Gizmodo]
And Google mocked Apple's delayed Siri AI improvements in a Pixel 10 ad, suggesting users could switch phones if promised features fail to materialize. [The Verge]
Meta has shifted its AI strategy from competing with ChatGPT to focusing on creating engaging, personalized experiences that maximize user attention and connection. [The Verge]
Samsung is upgrading its Bixby TV assistant with generative AI, enabling more natural conversations, enhanced search, and smart home device control. [ZDNet]
It’s Only a Model
Alibaba released Qwen-Image, an open-source AI image generator excelling at accurate text rendering across multiple languages and use cases. [VentureBeat]
Anthropic appears to be preparing Claude 4.1, with internal testing and safety validation suggesting a potential release focused on enhanced problem-solving capabilities. [TestingCatalog]
Cohere launched Command A Vision, an enterprise-focused visual AI model that excels at analyzing complex documents, charts, and images with high accuracy and low computational cost. [VentureBeat]
Deep Cogito launched four LLMs designed to improve reasoning efficiency through an innovative training process that helps models learn more effective thinking paths. [VentureBeat]
ElevenLabs has developed an AI music generator that enables users to create studio-quality tracks instantly using simple text prompts across various genres and styles. [ElevenLabs]
Google launched Deep Think, a high-performance AI model available to Ultra subscribers, demonstrating advanced problem-solving capabilities in mathematics and scientific discovery. [ZDNet]
And Google DeepMind unveiled Genie 3, an advanced world model capable of generating interactive 3D environments and potentially advancing the path to AGI. [TechCrunch]
Google DeepMind also released Perch, an AI model that helps conservationists analyze bioacoustic data from diverse ecosystems by processing complex audio recordings. [Google]
GPT Proto, an AI model platform, has launched OpenAI's GPT-OSS-120B and 20B open weight models, offering developers advanced AI capabilities with competitive pricing and enhanced infrastructure. [EIN Presswire]
Microsoft is making OpenAI's new free GPT model, gpt-oss-20b, available to Windows 11 users via Windows AI Foundry, enabling AI-powered tasks on consumer hardware. [TechCrunch]
Mistral AI launched Codestral 25.08 and a comprehensive enterprise coding stack designed to overcome deployment, customization, and integration challenges in AI-powered software development. [Mistral AI]
Here’s how to run OpenAI's new AI models on your laptop or phone. [TechRadar]
The release of OpenAI’s open-weight models has sparked debate about transparency, with the Allen Institute for AI arguing for more comprehensive openness in AI development. [GeekWire]
Xiaomi released an open-source voice model MiDashengLM-7B, integrating Alibaba's Qwen2.5-Omni-7B to advance its AI and automotive technologies. [Yahoo Finance]
Whose Data?
Australia’s Productivity Commission has suggested a text and data mining exception for AI training could allow smaller institutions to use copyrighted works without compensation. [The Guardian]
Microsoft sought to exclude its new Copilot AI consumer product from a copyright lawsuit filed by news organizations against Microsoft and OpenAI. [GeekWire]
News Corp's chief exec has criticized Trump’s AI stance, arguing that stripping intellectual property rights would undermine American creativity and potentially harm the president’s own book sales. [Publishers Weekly]
A survey by Noyb revealed that only 7% of European Meta users support the company’s AI training on their social media data. [The Register]
OpenAI swiftly discontinued a feature allowing ChatGPT conversations to be searchable online after widespread privacy concerns and potential unintended data exposure. [VentureBeat]
Cloudflare accused Perplexity of circumventing website blocking mechanisms, sparking a broader debate about AI web crawlers’ rights to access online content. [TechCrunch]
But Perplexity denied Cloudflare's accusations, claiming the technical analysis was fundamentally flawed. [ZDNet]
Universal Pictures has started adding copyright warnings to film credits to prevent AI companies from using its movies as training data without permission. [Gizmodo]
Vancouver-based author J.B. MacKinnon has filed class action lawsuits against Nvidia, Meta, Anthropic, and Databricks, alleging unauthorized use of copyrighted works to train AI systems. [Publishers Weekly]
The LLM Ecosystem
Cohere has developed North, an AI agent platform enabling private, secure enterprise deployments that protect sensitive data while providing advanced workplace automation capabilities. [TechCrunch]
Descope launched the Agentic Identity Control Plane, an innovative solution enabling security teams to manage and govern AI agent identities and access. [GlobeNewswire]
Fireworks AI has released benchmarks comparing open-source and proprietary AI models across knowledge, tool-use, and complex reasoning tasks, highlighting Qwen Instruct, Qwen3 Coder, and Claude Sonnet 4’s strengths. [Fireworks AI]
Google’s Gemini CLI GitHub Actions provides an AI coding teammate that automates issue triage, pull request reviews, and on-demand collaboration through customizable, secure workflows. [Google]
Google Cloud has launched AI agents for data engineering, transforming data pipeline creation, notebook workflows, and analytics through NLP and automation. [VentureBeat]
Microsoft has developed CLIO, a self-adaptive AI system that enables controllable, explainable reasoning for scientific discovery by creating internal reflection loops without post-training. [Microsoft]
Monte Carlo launched native Salesforce integrations to help enterprises monitor and ensure the reliability of critical CRM and Data Cloud data for AI applications. [Business Wire]
Skyflow launched an MCP Data Protection Layer to secure sensitive data flowing through AI agent connections while maintaining functionality and compliance. [Yahoo Finance]
SNIA launched Storage.AI, an open standards project uniting industry leaders to solve AI-related data challenges through collaborative, vendor-neutral approaches. [Yahoo Finance]
SuperX has launched an All-in-One Multi-Model Server series pre-configured with high-performance OpenAI language models, aimed at revolutionizing enterprise AI productivity. [EIN Presswire]
Tech Centre launched AI-driven MCP servers that enable Managed Service Providers to automate complex data retrieval and decision-making across multiple integrated platforms. [PRWeb]
YourGPT launched an AI Copilot Builder that enables businesses to automate complex workflows and convert conversations into instant, action-oriented tasks. [EIN Presswire]
Agentic AI
Arcade.dev and Lithic have partnered to develop the world’s first AI e-commerce agents capable of making secure, authenticated purchases. [Business Wire]
Cognigy launched a Mastery Program to certify professionals in building autonomous AI Agents for enterprise contact centers. [Business Wire]
ConCntric launched Amplify, an AI-powered preconstruction management platform that autonomously assists teams by analyzing data, generating insights, and streamlining workflows. [GlobeNewswire]
CyberHeed introduced an AI-powered Compliance Management Multi-Agent system that automates and streamlines cybersecurity compliance processes across various frameworks. [EIN Presswire]
Menos AI has launched Sonαr, an AI-powered research agent designed to help hedge funds and asset managers uncover novel investment insights efficiently. [Business Wire]
NinjaTech AI has disrupted the AI agent market by launching SuperNinja, an autonomous agent with flexible pricing starting at $25 per month. [EIN Presswire]
Outreach launched AI agents that automate and enhance revenue workflows, reducing manual work and providing deeper insights for sales teams. [Business Wire]
PropMix launched Appraisal Assistant, an AI-powered platform designed to revolutionize real estate appraisals by automating tasks and guiding professionals. [EIN Presswire]
Searchlight Cyber released an AI-powered tool that quickly summarizes dark web threat actor profiles, helping investigators efficiently analyze criminal activities. [Business Wire]
Upland Software launched its BA Insight platform on AWS Marketplace, enabling easier discovery and deployment of AI agent solutions. [Yahoo Finance]
Other LLM Sightings
Bellevue has partnered with Govstream.ai to pilot AI tools aimed at streamlining permitting processes and supporting housing and economic development goals. [BellevueWA.gov]
Click Intelligence launched Click Insights, an AI-powered SEO reporting software that provides comprehensive, real-time analytics through an easy-to-navigate dashboard. [EIN Presswire]
Docyt launched High Precision Accounting Intelligence, an AI-powered accounting automation tool trained on 128 billion data points. [Business Wire]
FiscalNote launched an AI-powered social listening feature in PolicyNote to help users detect early policy signals across social media platforms. [Business Wire]
Flashpoint unveiled AI-powered summarization tools for search and investigations, enhancing threat intelligence workflows and analyst efficiency. [Business Wire]
G2 and AWS expanded their partnership to integrate AI-powered, peer-review-based product insights into AWS Marketplace. [Business Wire]
InvestCloud launched AI-enabled solutions with smartKYC and Zocks to enhance advisor productivity and client engagement through intelligent screening and meeting technologies. [Business Wire]
Jeeva AI launched comprehensive AI sales automation solutions across industries, enabling businesses to boost pipeline growth, efficiency, and revenue through tailored, compliant platforms. [EIN Presswire]
Mindbreeze integrated multimodal LLMs into its InSpire product, enabling enterprises to analyze diverse data types with advanced AI capabilities. [Business Wire]
Pivt launched an AI-powered Buddy System to enhance employee connection, belonging, and retention during workplace transitions. [PRWeb]
RadarFirst launched Radar Controls, an AI-powered platform that automates regulatory compliance mapping and helps organizations identify and manage legal requirements. [GlobeNewswire]
Search.co launched an AI platform that autonomously extracts, ingests, and acts on data across enterprise workflows without requiring complex coding. [EIN Presswire]
Technician Find launched Jason Perkins, an AI recruiting assistant trained on seven years of automotive technician conversations to help shops attract qualified candidates more effectively. [EIN Presswire]
Windfall launched an AI Copilot that helps go-to-market teams leverage people data and AI to make faster, more strategic decisions. [PRWeb]
Wrike launched Wrike Copilot, an AI-powered work management assistant that provides real-time insights, collaboration support, and actionable intelligence for enterprise teams. [Business Wire]
Risks and Responses
Anthropic researchers discovered that deliberately activating undesirable AI behavior patterns during training can paradoxically prevent those traits from emerging in LLMs. [MIT Technology Review]
Anthropic has launched automated security review tools for Claude Code, enabling developers to scan code for vulnerabilities and suggest fixes using AI-powered analysis. [VentureBeat]
AWS is expanding Automated Reasoning Checks on Bedrock to help enterprises confidently deploy AI applications by validating responses and reducing hallucinations through neurosymbolic AI techniques. [VentureBeat]
Carnegie Mellon researchers demonstrated that AI can independently plan and execute complex cyberattacks, replicating the Equifax breach without human intervention. [TechRadar]
CrowdStrike revealed hackers are leveraging AI to accelerate attacks, exploit enterprise AI tools, and lower barriers to cybercrime. [TechRadar]
DeepSeek faced renewed US Senate scrutiny over potential national security risks, with Republican senators requesting further investigation into the Chinese AI chatbot’s data security. [TechRadar]
Google's Med-Gemini healthcare AI model erroneously referenced a non-existent ‘basilar ganglia’ in a research paper, raising concerns about AI accuracy in medical diagnostics. [The Verge]
Google's Gemini AI assistant was demonstrated to be vulnerable to prompt injection attacks that could hijack smart devices through manipulated calendar invites. [Gizmodo]
Microsoft Windows 11’s Recall feature struggled to consistently filter out sensitive information like credit card numbers and passport details during testing. [TechRadar]
Microsoft’s Project Ire AI security agent detected only 26% of malware in tests. [The Register]
MultiKol has developed a dynamic biometric voice recognition technology to combat deepfake fraud by analyzing internal voice anatomy and using AI to identify malicious AI-generated scam attempts. [Ynet]
NIST organized a red-teaming exercise revealing 139 novel ways to exploit AI systems, but the resulting report remains unpublished due to potential political sensitivities. [Wired]
Norton launched Deepfake Protection in its mobile apps, enabling users to detect AI-generated fake videos on YouTube across select English-speaking countries. [TechRadar]
OpenAI’s ChatGPT reportedly provided alarming, personalized advice to teen researchers about drugs, self-harm, and suicide. [Associated Press]
A ChatGPT Connectors vulnerability allowed researchers to extract sensitive data from Google Drive through a zero-click, indirect prompt injection attack. [Wired]
OpenAI is updating ChatGPT to discourage unhealthy user behaviors by prompting breaks and avoiding overly emotional or therapeutic responses. [NBC News]
OpenAI has struck a deal with US National Laboratories to explore using AI for nuclear weapon security, amid growing concerns about potential risks and unintended consequences. [Futurism]
And OpenAI has faced criticism from AI experts and celebrities in an open letter questioning its commitment to benefiting humanity amid its transition to a for-profit entity. [Futurism]
Veracode discovered that 45% of AI-generated code contains security flaws across various programming languages, raising significant concerns for software development. [TechRadar]
Wikipedia adopted a new policy allowing speedy deletion of AI-generated articles that include user-directed language or incorrect citations. [404 Media]
Regulation
Brussels secured broad support from tech companies for its AI Code of Practice, signaling smooth implementation of upcoming EU AI regulations. [Politico]
Governor Ron DeSantis has signalled potential AI regulations for Florida, expressing scepticism about technology’s impact while acknowledging its economic importance. [Tallahassee Democrat]
Illinois passed a law banning AI from acting as a therapist, limiting its use to administrative tasks in mental health services. [Gizmodo]
Environmental Issues
Meta is exploring mass timber construction for its datacenters to reduce greenhouse gas emissions by approximately 41 percent. [The Register]
Rocky Mountain Power revealed that a planned Wyoming AI facility will consume more electricity than all residential households in the state, highlighting the rapid growth of energy-intensive AI infrastructure. [TechRepublic]
Conversational AI
8x8 has expanded its AI-powered customer experience platform, boosting engagement through intelligent automation, API growth, and new customer wins. [Business Wire]
AT&T expanded its Office@Hand portfolio with RingCentral's AI-powered contact center and conversational intelligence solutions to enhance business communications. [Business Wire]
Avaamo unveiled CX Agents, a next-generation digital workforce platform that transforms enterprise customer service through AI-powered, multilingual support agents. [EIN Presswire]
Insider Perks launched an AI phone-call agent for campground bookings, enabling instant reservations and multilingual support through Campspot integration. [EIN Presswire]
MakeMyTrip launched a multilingual GenAI Trip Planning Assistant, enabling conversational and inclusive travel booking across multiple Indian languages. [Business Wire]
Nesti launched Nesti.Chat, an AI assistant for real estate agents designed to handle inquiries, qualify leads, and book viewings. [UKTN]
Pega has introduced an advanced self-service solution combining Blueprint and Predictable AI to transform customer service across multiple channels with enterprise-grade intelligence. [Business Wire]
Reiwa Travel launched NEWT Chat, an AI-powered customer service agent designed to support travel and service businesses with 24/7 multilingual assistance. [PRWeb]
Rexpt has launched an AI receptionist service enabling small businesses to create custom automated call-handling agents in under three minutes. [EIN Presswire]
RingCentral expanded its AI Receptionist capabilities, launching AIR Everywhere to provide AI-powered call handling across diverse phone systems. [Business Wire]
SoundHound AI announced its Chat AI Automotive has been deployed by three major global automotive brands across North America, enhancing in-vehicle conversational experiences. [Business Wire]
UC Connect has launched an AI-powered call transcription and summarization feature that transforms recorded calls into actionable insights for businesses. [EIN Presswire]
Be Real
Former CNN anchor Jim Acosta drew widespread criticism for conducting an insensitive AI interview with a Parkland shooting victim’s avatar, sparking debate about the ethical use of AI technology. [Futurism]
Grok Imagine, xAI's iOS image generator, allows users to create potentially inappropriate AI-generated videos of celebrities with varying levels of nudity, particularly targeting women. [Gizmodo]
Ulf Kristersson, Sweden’s prime minister, admitted to seeking governance advice from ChatGPT, sparking concerns about AI’s growing influence on decision-making. [Gizmodo]
Voice News
Attention labs has developed an on-device voice recognition technology that enables natural communication in noisy environments with instant, privacy-preserving AI response. [EIN Presswire]
Cerence has filed patent infringement complaints against Sony and TCL, seeking to block imports and pursue damages for unauthorized use of its voice technology. [Gizmochina]
Howard University and Google launched Project Elevate Black Voices to improve voice-recognition technology’s accuracy for African American English speakers. [Washingtonian]
Krisp launched AI Accent Conversion technology for Filipino and Indian English accents, enhancing communication clarity in global virtual meetings. [Business Wire]
Microsoft is warning users of older Office versions that voice tools like transcription and dictation will lose support by January 2026. [TechRadar]
Nutshell introduced an AI-powered call summarization feature that automatically transcribes and summarizes calls from various platforms, saving sales teams time and effort. [PRWeb]
SignalWire has partnered with AWS to integrate Nova Sonic voice-to-voice AI into its platform, enabling developers to create advanced, intelligent voice applications. [Business Wire]
And SignalWire launched a Developer Voice Toolkit and Sigmond AI agent alpha to help developers build real-time AI communications applications more efficiently. [Business Wire]
Document AI
FELLOWPRO has enhanced DocBits with automated testing, multilingual support, walkthrough guidance, and expanded documentation to improve document processing reliability and user-friendliness. [FELLOWPRO]
fileAI launched its V2 platform, enabling enterprises to access, collect, and structure unstructured business data across diverse file types with high accuracy. [IDM]
LotusPetal AI launched an end-to-end RFP automation platform that helps businesses generate proposals faster and more accurately using AI. [PRWeb]
NetDocuments unveiled AI-powered document management enhancements, including automated metadata extraction and an agentic AI editing tool for Microsoft Word. [LawSites]
PCMI launched Claims Intelligence, an AI-powered framework designed to automate and optimize finance and insurance claims processing through advanced analytics and intelligent workflows. [PRWeb]
PubHive launched AI-Insight, a Gen-AI platform enabling personalized, efficient literature reviews for life science teams across various functions. [EIN Presswire]
Trademo launched TradeScreen, an AI platform that automates trade document digitization, compliance checks, and risk analysis for global trade stakeholders. [EIN Presswire]
Translation
Awtomated has integrated Microsoft Translator into its Translation Business Management System, enabling LSPs to access high-quality MT directly within the platform. [MultiLingual]
Profuz Digital has expanded its LAPIS platform with AI-driven orchestration features that intelligently manage digital assets across broadcast, production, and localization workflows. [MultiLingual]
Search
Brave has introduced AI Grounding with its Search API, providing enhanced search performance and factual accuracy for AI applications through verifiable web sources. [Brave]
Eminent SEO launched AI Optimization Services to help brands gain visibility and authority in AI-powered search ecosystems. [EIN Presswire]
Google denied that AI search features are significantly harming publishers’ traffic, claiming organic click volume remains stable and click quality has slightly increased. [TechCrunch]
Microsoft has aggressively promoted Copilot AI on Bing search results, strategically placing a prominent banner above rival AI service searches to potentially mislead users. [TechRadar]
Perplexity is powering a new AI search engine for Truth Social, Trump’s social media platform. [TechCrunch]
Reddit aims to become a search destination by unifying its search and AI-powered Q&A features, leveraging its unique conversational content. [Yahoo Finance]
The Browser Company has launched a $20 Pro subscription for Dia, its AI-powered web browser, offering unlimited access to advanced AI features. [TechCrunch]
WebFX analyzed 2.37 million US Google searches, revealing AI Overviews now appear in 25.8% of results, rising to 54% for longer informational queries. [PRWeb]
AI in Journalism
This report says Politico hastily implemented error-prone AI tools across its platform, violating editorial standards and union contracts while introducing numerous factual inaccuracies. [TechDirt]
SmartNews launched NewsArc, an AI-powered news app designed to deliver high-quality, contextual journalism through advanced language models and publisher partnerships. [Business Wire]
Norwegian newspaper Verdens Gang has launched FOIA Bot, an AI-powered newsroom assistant that helps journalists efficiently file and respond to Freedom of Information Act requests using retrieval-augmented generation technology. [NiemanLab]
A survey by WAN-IFRA revealed that nearly half of news organizations have adopted AI, despite ongoing ethical concerns about the technology. [Editor & Publisher Magazine]
Health Tech
eClinicalWorks and Sunoh.ai helped Evolve Psychiatry streamline clinical documentation, saving providers over two hours daily through AI-powered ambient listening technology. [Business Wire]
HealthRescue Partners has launched an affordable Cloud EHR and AI-powered Practice Intelligence Platform to revolutionize private practice management. [EIN Presswire]
IntelePeer launched SmartAgent, an AI-powered automation solution designed to improve operational efficiency and patient experiences for dental service organizations. [Business Wire]
PsychAssist.ai launched an AI-native platform for assessment psychologists, automating workflow while preserving clinical reasoning and stakeholder-specific reporting. [PRWeb]
Voxela has launched VCare Pro, an AI-powered eldercare solution that reduces caregiver workload and enhances resident safety through automated incident reporting and remote monitoring. [Business Wire]
Legal Tech
Descrybe joined the National Society for Legal Technology’s curriculum as a core legal research tool, replacing Casetext and validating its AI-powered platform. [Legal IT Insider]
Digest AI launched an AI-powered legal assistant tailored for Filipino lawyers, streamlining research, drafting, and compliance tasks. [Mindanao Times]
Harvey, a legal generative AI pioneer, celebrated its three-year milestone by reaching US$100m ARR and securing over 500 customers across 54 countries. [Artificial Lawyer]
And Harvey has launched an Outlook Add-In to bring AI-powered drafting and summarization directly into users’ inboxes, supporting its broader vision of integrating AI across legal tech stacks. [Artificial Lawyer]
Harvey, Spellbook, Thomson Reuters, and Legora have praised GPT-5’s advanced reasoning, document analysis, and contextual understanding capabilities for legal technology applications. [Artificial Lawyer]
LegalZoom’s products are being leveraged to provide legal guidance within OpenAI’s ChatGPT agent. [Business Wire]
Litera has expanded its Foundation Platform with Microsoft Outlook integration and AI-powered Lito assistant, enhancing firm intelligence and workflow productivity. [LawSites]
The National Association for Law Placement data reveals record law graduate employment rates, with a slight 3% salary drop, potentially hinting at early, subtle AI-driven market shifts. [Artificial Lawyer]
NetDocuments has strategically partnered with Harvey and Legora to expand AI capabilities and ensure secure, seamless integration across legal tech platforms. [Above the Law]
Practice AI has integrated with SmartAdvocate, enabling seamless document and data management for legal professionals through automated workflows. [GlobeNewswire]
Reveal Data is launching aji, a generative AI-powered document review platform for legal professionals. [LawSites]
And Reveal and Clearbrief have integrated their AI-powered platforms to connect legal discovery and drafting with transparent, evidence-based workflows. [Business Wire]
Thomson Reuters has launched CoCounsel Legal, an integrated AI system combining research tools, content, and workflow capabilities for legal professionals. [Artificial Lawyer]
Turnberry Solutions has selected Universal Migrator to accelerate data migrations into Litify, leveraging its sophisticated migration processes for law firms. [LawSites]
UniCourt launched DART, a comprehensive legal analytics platform enabling professionals to research, track, and gain insights from court data across thousands of state and federal courts. [Business Wire]
Ed Tech
eSkilled has launched an AI-Assisted Marking tool to improve assessment consistency, reduce bias, and ensure fairness in vocational education. [EIN Presswire]
And eSkilled has released LLND Quick Assess, an AI-powered tool helping RTOs streamline pre-enrolment language, literacy, and digital skills assessments. [EIN Presswire]
Google has expanded NotebookLM to younger users, offering AI-powered note-taking features with enhanced safety measures for students. [TechCrunch]
Google has announced free AI tools for college students in select countries, offering advanced capabilities like Gemini 2.5 Pro and Guided Learning to democratize access to cutting-edge educational technology. [Forbes]
Instructure has partnered with OpenAI to integrate AI tools into Canvas, aiming to enhance learning management while raising concerns among faculty about educational quality. [Inside Higher Ed]
LearnWise has developed an AI-powered edtech platform that provides context-aware, institution-specific assessment feedback and grading solutions for universities. [Silicon Canals]
Funding
Anaconda secured a significant Series C funding round led by Insight Partners, aiming to accelerate its AI platform’s growth and global expansion. [The Arabian Post]
August has raised US$7m to provide configurable AI workflows that help midsize law firms automate and optimize their legal processes. [Artificial Lawyer]
Blue J raised US$122m to expand its AI-powered tax research platform, leveraging generative AI to provide instant, reliable answers to complex tax questions. [Artificial Lawyer]
Clay, a sales automation startup, raised a US$100m Series C at a US$3.1 billion valuation, with plans to reach US$100m in revenue. [TechCrunch]
Conversion raised US$28m in Series A funding led by Abstract Ventures to expand its AI-native marketing automation platform for B2B businesses. [FinSMEs]
Elion raised US$9.3m to develop an AI-powered platform helping healthcare systems navigate technology vendor procurement and decision-making. [Business Wire]
Fal raised a US$125m Series C round, valuing the AI infrastructure company at US$1.5 billion with investments from Meritech, Salesforce Ventures, and others. [Reuters]
Fundamental Research Labs raised US$33m in Series A funding to develop AI applications across various domains, including consumer assistants and productivity tools. [TechCrunch]
FuriosaAI secured US$125m in Series C funding and gained LG AI Research's endorsement for its energy-efficient RNGD AI inference accelerator. [Yahoo Finance]
Kontext raised US$10m to develop a contextual ad platform that generates real-time, personalized advertisements within AI chatbot interfaces. [Adweek]
Kotoba Technologies secured US$11.83m in Seed 2 funding to advance its AI-powered simultaneous interpretation technology for Asian languages. [Business Wire]
Lorikeet secured US$35m in Series A funding to develop AI concierges that help businesses resolve complex customer service issues across multiple channels. [FinSMEs]
Mistral AI is negotiating a US$1B funding round with investors, potentially valuing the AI startup at US$10B. [Yahoo Finance]
Oxmiq Labs has raised US$20m to develop licensable AI-focused GPU technology. [Yahoo Finance]
SixSense developed an AI platform that helps semiconductor manufacturers predict and detect chip defects in real-time, raising US$8.5m in Series A funding. [TechCrunch]
Streamline AI secured US$8.6m in Series A funding to expand its AI-powered legal intake and matter management platform for in-house teams. [FinSMEs]
Tavily secured US$20m in Series A funding to help enterprises connect AI agents to the web while maintaining compliance and governance. [TechCrunch]
Tracelight raised US$3.6m to develop AI technology that simplifies financial modeling by integrating LLMs directly into Excel workflows. [Tech.eu]
Vast Data is in talks with Alphabet's CapitalG and Nvidia to raise funding at a potential US$30B valuation for its AI infrastructure technology. [Yahoo Finance]
EY reported that VC investment in generative AI surged to US$49.2B in the first half of 2025, exceeding the total for the previous year. [ITPro]
And Abu Dhabi’s MGX is exploring a potential US$25B capital raise to expand its AI investments. [DealStreetAsia]
Acquisitions
Aderant acquired HerculesAI's legal technology assets, including AI-powered products Apollo, Athena, and Verify, to enhance its work-to-cash platform for law firms. [LawSites]
QuMind was acquired by Largo.ai, combining advanced AI-driven market research technologies to enhance consumer insights and decision-making processes. [Tech.eu]
There’s More
Anthropic held a theatrical funeral for Claude 3 Sonnet, complete with eulogies and a resurrection ritual, highlighting the passionate and almost mystical relationship some users have with AI models. [Wired]
Geoff Hinton warned that AI could develop an incomprehensible language, potentially thinking in ways humans cannot understand or track. [Business Insider]
OpenAI's gpt-oss-20b struggled with the 2024 US presidential election results, confidently asserting Biden’s victory despite Trump actually winning. [The Register]
Sapient Intelligence has developed HRM, a brain-inspired AI architecture that outperforms LLMs on complex reasoning tasks with significantly fewer resources. [VentureBeat]
The Back Cover
Here’s The Smith Street Band with a cover of Alanis Morissette’s Hand in my Pocket.
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!