This Week in NLP #344
Keep up with what happened in NLP in the week ending Friday 20th June 2025.
Above the Fold
Overwhelmed? Here’s our pick for five things to know about this week.
Google has decided to end its US$200m partnership with Scale AI after Meta acquired a 49% stake. [Verdict]
Meta has attempted to poach OpenAI's top researchers by offering US$100m hiring bonuses, but none have accepted the offers, according to Sam Altman. [digitaltrends]
Nvidia has promoted ‘sovereign AI’ across Europe, announcing partnerships and investments while encouraging European nations to develop their own AI capabilities and infrastructure. [Yahoo Finance]
OpenAI's relationship with Microsoft has become strained over restructuring negotiations, with disputes centered on ownership stakes, profit rights, and access to technologies. [TechRepublic]
xAI has been spending US$1B monthly on AI development while seeking US$9.3B in funding, although Musk has denied Bloomberg’s reported figures. [eWeek]
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
Here’s a piece on how Amazon has rebuilt Alexa using AI tools throughout the development process, creating Alexa+, a more conversational and personalized voice assistant that’s being gradually rolled out to users. [Wired]
Anthropic has attracted top AI talent from major tech companies by offering competitive pay and fostering a culture of independence, outperforming rivals in employee retention. [Techstory]
Meta has offered US$10m annually to elite AI researchers, with CEO Zuckerberg personally emailing candidates amid fierce competition for top talent. [The Register]
Meanwhile, Microsoft has led a wave of AI talent acquisitions, including a US$650m licensing deal with Inflection AI, as tech giants compete for scarce expertise through strategic acquihires. [The Economic Times]
MiniMax, a US$3B Chinese AI startup backed by Alibaba, has begun preparations for a potential Hong Kong IPO later this year. [Bloomberg]
OpenAI’s ChatGPT has maintained its dominance with 5.49 billion visits in May, while Gemini showed strongest growth and surpassed it in mobile downloads. [The Decoder]
OpenAI has secured a US$200m Defense Department contract and consolidated its public sector offerings into ‘OpenAI for Government’, supporting administrative and security applications. [eWeek]
OpenAI has ended its partnership with Scale AI following Meta's investment in the startup, prompting other major clients like Google to consider similar moves. [TechCrunch]
OpenAI has requested Microsoft forgo future profits for a 33% stake in a restructured entity, while seeking exemptions from IP agreements. [Yahoo Finance]
OpenAI has encountered resistance from Microsoft while attempting to restructure its business and secure a US$40B investment led by Softbank. [Yahoo Finance]
OpenAI has considered accusing Microsoft of anticompetitive behavior amid tensions over corporate structure, cloud providers, and access to Windsurf's technology. [ITPro]
Microsoft has threatened to abandon OpenAI negotiations over future stake disagreements, while maintaining access to the company’s technology through existing contracts until 2030. [Yahoo Finance]
Meanwhile, OpenAI's governance and leadership have come under scrutiny through ‘The OpenAI Files’, a watchdog project examining ethical concerns and proposing reforms for responsible AI development. [Yahoo Finance]
And OpenAI has revealed plans to launch GPT-5 this summer as part of an effort to simplify its product lineup by unifying multiple AI models into one streamlined interface. [The Neuron]
Salesforce has announced a 6% price increase for several enterprise products while launching new AI-focused offerings, despite its own research showing limitations in AI agent performance. [The Register]
Scale AI has lost major clients Google, Microsoft, and xAI after Meta acquired a 49% stake in the AI data-labeling startup. [Yahoo Finance]
AI Supremacy
DeepSeek's launch has sparked global debate about the US-China AI race, prompting experts to examine broader implications for governance, ethics, and worldwide technological adoption. [Rest of World]
DeepSeek's success has driven up Chinese deeptech valuations, attracting venture capital as government backing strengthens the domestic ecosystem. [DealStreetAsia]
Huawei has filed a patent for a quad-chiplet AI accelerator design with advanced packaging technology that could help it compete with Nvidia while bypassing US sanctions. [Tom’s Hardware]
Huawei and SMIC have been placed on Taiwan’s strategic export control list, restricting their access to advanced chips and semiconductor manufacturing services. [Techstory]
Huawei has signed a Letter of Intent with South Africa’s Department of Science, Technology and Innovation to enhance ICT collaboration, particularly in AI and research, strengthening Sino-South African technological ties. [iAfrica]
Malaysian authorities are investigating reports that Chinese engineers accessed Nvidia AI chips in Malaysian data centers to circumvent US export restrictions to China. [The Standard]
Nvidia-powered servers were allegedly smuggled through Malaysia and Singapore to China, prompting investigations into potential US export control violations. [Yahoo Finance]
Nvidia's CEO has warned that Huawei would dominate China’s AI market and expand globally if US export restrictions on semiconductors continued. [Yahoo Finance]
Nvidia has confirmed its first-time participation in Beijing’s International Supply Chain Expo this July, amid ongoing US-China trade tensions and export restrictions. [Yahoo Finance]
Texas Instruments has announced a US$60B investment to build seven semiconductor plants across Texas and Utah, creating 60,000 jobs in US manufacturing. [CNN]
Sovereign AI
Amazon has announced a A$20B investment to expand its Australian data center infrastructure from 2025 to 2029, supporting AI capabilities and including solar farm projects. [Yahoo Finance]
ANIA (the National Agency for Artificial Intelligence) has launched in El Salvador to oversee AI development, education, and regulation under newly approved legislation. [El Salvador in English]
Anthropic is expanding its European presence and seeking to boost the continent’s tech ecosystem through strategic hiring and partnerships with local startups. [RTL]
AXL has launched a US$15m venture studio in Toronto to build 50 AI companies over five years, aiming to keep Canadian AI innovation within the country. [Business Wire]
Cyberview's Living Lab Accelerator programme has facilitated a partnership between Mesolitica and Lekir Tech to advance Malaysia’s AI capabilities through language models and AI agents. [Vulcan Post]
Deutsche Telekom has partnered with Nvidia to build an AI cloud service in Germany by 2026, featuring 10,000 Nvidia GPUs for European manufacturers. [Yahoo Finance]
Docugami has established a French subsidiary to expand its Document AI technology across Europe, focusing on open-source solutions and data sovereignty. [Business Wire]
EnterpriseDB has launched Postgres AI, a unified platform that accelerates AI deployment across enterprises while maintaining data sovereignty and reducing implementation costs. [Business Wire]
G42 has launched a European division to deploy AI solutions and build infrastructure partnerships, expanding Abu Dhabi’s tech presence in the region. [Bloomberg]
European telecom giants Orange, Fastweb, Swisscom, Telefónica and Telenor have partnered with Nvidia to develop sovereign AI factories and edge infrastructure across Europe. [Computer Weekly]
And Nvidia has partnered with European and Middle Eastern AI teams to develop region-specific language models, supporting local languages and cultural nuances across multiple industries. [Yahoo Finance]
London Tech Week 2025 launched a £1B national AI initiative, featuring Nvidia's new AI Technology Centre and aiming to train 7.5 million British workers by 2030. [Tech Funding News]
Tech giants including Amazon, Google, Microsoft and IBM have joined UK government talks to support an initiative aimed at providing AI skills training to over seven million British workers. [UKTN]
Humphrey, the UK government’s AI toolkit, has integrated models from OpenAI, Anthropic, and Google, sparking concerns about Whitehall’s growing dependence on major tech companies. [The Guardian]
Microsoft has launched its expanded Sovereign Cloud offering in Europe, combining public and private infrastructure with enhanced data controls, encryption options, and local operational management. [Microsoft]
And Microsoft has launched Microsoft 365 Local, an on-premises productivity solution running on Azure Local to meet European data sovereignty requirements. [TechRadar]
A Mongolian startup has developed its own LLM, offering an AI alternative to Chinese and American technologies. [Bloomberg]
Nvidia's shares approached record highs after the company established European AI partnerships, with Oppenheimer projecting a US$1.5T global sovereign AI market. [Yahoo Finance]
NEXTDC has committed A$2B to build M4 Melbourne, a digital campus featuring an AI Factory, Mission Critical Operations Centre, and Technology Centre of Excellence in Port Melbourne. [Asia Pacific Defence Reporter]
Orbit Capital has launched a €100m Growth Debt II fund to provide venture debt financing for tech companies across Central and Eastern Europe. [Tech Funding News]
UK’s AI sector has reached a US$230B valuation, though one-third of startup leaders are considering relocating due to capital and talent shortages. [ITPro]
Yanolja Cloud's CEO Jeff Kim has urged South Korea’s new president to partner with US tech companies on a Stargate-like AI infrastructure project worth US$74B. [Semafor]
Feature Creeps
Anthropic has expanded Claude Code to support remote MCP server integration, enabling developers to connect with third-party services and tools through their workspaces. [ZDNet]
AWS has enhanced its Bedrock AI service to support cross-platform AI models, aiming to compete with Google Cloud and Microsoft Azure’s offerings. [Yahoo Finance]
Google has expanded Gemini’s PDF handling in Drive, adding automatic summaries and suggested actions for stored documents across 20 languages. [ZDNet]
Google’s Gemini has expanded its video analysis capabilities to include users’ personal videos, allowing detailed content summaries and interactive Q&A about uploaded clips. [ZDNet]
Microsoft has launched an AI agent for Windows 11 Insiders that can automatically adjust system settings based on natural language requests. [ZDNet]
OpenAI has enhanced ChatGPT Projects with six new features, including Deep Research mode, voice commands, and improved sharing capabilities for Plus, Pro, and Team subscribers. [ZDNet]
And OpenAI has integrated Model Context Protocol into ChatGPT, enabling enterprise users to connect company data sources. [ZDNet]
WhatsApp has integrated ChatGPT’s image generation and editing capabilities, allowing users to create and modify AI images directly through the messaging platform. [TechRadar]
Hype Bubble?
Anthropic CEO Dario Amodei has warned that AI could eliminate half of entry-level jobs within five years, urging tech leaders to stop downplaying the impact. [ITPro]
Apple's research paper claiming reasoning LLMs don’t actually think has sparked intense debate, with critics arguing the study’s methodology was fundamentally flawed. [VentureBeat]
Canalys has reported that enterprise AI adoption is being hampered by difficulties in predicting inferencing costs and fears of excessive cloud service bills. [The Register]
Endava's survey has revealed that British business leaders overwhelmingly support rapid AI implementation while simultaneously calling for government-led international oversight of the technology. [The Next Web]
Fivetran has released research showing that while 89% of enterprises plan to train AI models with proprietary data, only 49% have adequate data architecture. [Business Wire]
Gartner‘s AI research chief Brethenoux has criticized AI’s current capabilities, arguing it should automate tasks rather than just summarize them, while questioning vendors’ promises about AI agents. [The Register]
But Gartner has predicted that AI agents will handle or augment half of all business decisions by 2027, with executive boards increasingly relying on AI assistance. [ZDNet]
GitHub data revealed US developers led global AI coding assistant usage at 30.1% in 2024, potentially generating US$9.6-US$14.4B in annual economic value. [The Register]
GoTo's research has revealed that while 62% of employees consider AI overhyped, 86% admit they aren’t utilizing AI tools to their full potential. [Business Wire]
Meta's chief AI scientist LeCun has criticized Anthropic CEO Amodei on Threads, calling him intellectually dishonest for developing AGI while warning of its dangers. [The Decoder]
Microsoft's Copilot advertising has drawn criticism from an industry watchdog for making unsubstantiated productivity claims and using confusing branding across multiple products. [The Verge]
NTT Data‘s survey revealed a stark divide between CEOs’ enthusiasm and CISOs’ concerns about generative AI adoption, with 45% of security leaders expressing negative sentiments. [ITPro]
Nvidia and OpenAI leaders have expressed contrasting views on AI’s impact on employment, with Huang predicting job changes while Altman warned of potential job losses. [Investopedia]
Salesforce‘s CRMArena-Pro benchmark has revealed significant limitations in AI agents’ performance, with multi-turn tasks achieving only 35% success rates across major models. [TechRadar]
Verint has released a study showing 86% of consumers value AI in customer service. [Business Wire]
Hardware
AMD has launched its Instinct MI355X GPU accelerator, delivering up to 20.1 PFLOPS of computing power while consuming 1,400W, surpassing Nvidia's B300 in raw performance. [Yahoo Finance]
AMD has strengthened its AI position by partnering with startups, acquiring companies, and incorporating OpenAI's feedback into its MI450 chip design. [Yahoo Finance]
Apple has expressed interest in using generative AI to accelerate the design of its custom device chips. [Yahoo Finance]
AWS has announced a Graviton4 chip upgrade featuring 600 gigabits per second network bandwidth, challenging Nvidia's dominance in AI infrastructure with Project Rainier and Trainium chips. [CNBC]
Consumer AI
Mattel has partnered with OpenAI to develop AI-powered toys and integrate ChatGPT Enterprise into its operations, amid declining toy industry sales. [Business Insider]
Meta has partnered with Oakley to launch sports-focused smart glasses, with an official announcement expected on June 20, following earlier success with Ray-Ban collaboration. [ZDNet]
And Meta has partnered with Prada to develop AI smart glasses, expanding beyond its existing collaboration with EssilorLuxottica's Ray-Ban and Oakley brands. [TechCrunch]
Silicon Valley companies have renewed their push into smart glasses technology, with Snap announcing AI-equipped eyewear while Meta's Ray-Ban glasses have sold two million units. [abc17news.com]
It’s Only a Model
Alibaba has optimized its Qwen3 AI models for Apple's MLX framework, enabling local AI features on Apple devices in China. [Yahoo Finance]
Apple has released performance data showing its new AI models trail competitors, while making its smaller 3-billion-parameter model available to third-party developers. [The Decoder]
CoRover has launched BharatGPT Mini, a lightweight multilingual AI model supporting 14 Indian languages that operates offline and is optimized for low-end devices. [Analytics India Magazine]
Google has launched its Gemini 2.5 AI models for enterprise production, introducing three tiers while undercutting competitors with a new cost-effective Flash-Lite variant. [VentureBeat]
And Google DeepMind has launched Gemini Diffusion, an experimental language model using noise-based generation to produce text faster than traditional autoregressive approaches. [VentureBeat]
Midjourney has launched V1, an image-to-video AI model that transforms images into 5-second videos, amid ongoing copyright disputes with Disney and Universal. [TechCrunch]
MiniMax has launched its open-source M1 language model, challenging DeepSeek's dominance in China with superior token capacity and computational efficiency. [The Register]
Mitsubishi Electric has developed a compact, domain-specific language model for manufacturing that runs on edge devices and leverages the company’s proprietary data. [Business Wire]
OpenAI has delayed its open-weight language model release until summer following unexpected breakthroughs in reasoning capabilities. [The Decoder]
And OpenAI has announced the removal of GPT-4.5 Preview from its API by July 2025, while maintaining its availability for individual ChatGPT users. [VentureBeat]
Whose Data?
GLAM-E Lab has reported widespread disruption of cultural institutions’ online collections by AI scraping bots, with 27 of 43 surveyed organizations experiencing server-overwhelming traffic increases. [404 Media]
Harvard University has released nearly one million books from its collection to AI researchers, making centuries of knowledge in 254 languages available for ML. [Associated Press]
Mastodon has updated its terms of service to prohibit AI model training through data scraping, following similar moves by X and other platforms. [TechCrunch]
Microsoft has introduced a ‘Data Guardian’ system in Europe that restricts system access control to regional employees, enhancing data sovereignty protections. [Investopedia]
The UK government has launched a controversial consultation on AI training rules, seeking to balance content creators’ rights with AI firms’ needs for data access. [ITPro]
The LLM Ecosystem
Accuris has integrated regulatory content into its Engineering Workbench platform, providing engineers with centralized access to US Federal regulations alongside technical standards and publications. [Business Wire]
Algolia has launched its MCP Server, enabling AI agents to securely access and act on real-time business context through its search and discovery platform. [Business Wire]
Anthropic's CEO Amodei has advocated for developing interpretable AI models to understand their decision-making processes, while investing in tools and research to achieve this goal by 2027. [VentureBeat]
Anysphere has launched a $200 monthly Ultra subscription for its Cursor coding tool, offering 20x more AI model usage amid growing competition from major providers. [TechCrunch]
Databricks has launched Agent Bricks, a unified workspace that automates AI agent development and optimization using enterprise data and synthetic equivalents. [siliconANGLE]
And Databricks has launched Lakeflow Designer for no-code data pipeline creation and Lakebase for OLTP operations, expanding its data engineering capabilities beyond AI and analytics. [BigDATAwire]
DEV has launched LLM.co, a platform enabling regulated industries to deploy private, secure LLMs within their controlled environments. [EIN Presswire]
G42 has partnered with Liquid AI to develop and deploy private generative AI solutions across multiple sectors internationally, focusing on efficient and sovereign enterprise applications. [Business Wire]
Google has launched its Safety Charter and security engineering center in India to combat digital fraud through AI-led developments and partnerships with local organizations. [Yahoo Finance]
Groq has launched two major initiatives: supporting Alibaba's Qwen3 32B model with full context window and becoming a Hugging Face inference provider, challenging established cloud services. [VentureBeat]
IBM has launched integrated AI security and governance software that combines watsonx.governance with Guardium AI Security to manage risks across enterprise AI applications. [Verdict]
LambdaTest has partnered with Compunnel to help enterprises modernize software testing through AI-native automation and digital engineering expertise. [GlobeNewswire]
And LambdaTest has launched its Accessibility MCP Server, enabling developers to identify and resolve accessibility issues in web applications through AI-assisted testing and automation. [GlobeNewswire]
Mirantis has launched the industry’s first comprehensive reference architecture for IT infrastructure supporting AI workloads, enabling secure and scalable AI/ML deployment across environments. [Business Wire]
Nebius Group has partnered with Saturn Cloud to deliver an AI/ML infrastructure solution using Nvidia Hopper GPUs, offering lower-cost enterprise-grade computing resources. [Yahoo Finance]
Ory has partnered with Skyfire to enable AI agents to create accounts, verify identity, and make payments using established OAuth2 and OpenID Connect protocols. [EIN Presswire]
Promptfoo has launched a next-generation AI red teaming platform for enterprise LLM systems. [PRWeb]
Temasek has joined Microsoft, BlackRock, and MGX's AI Infrastructure Partnership, which aims to invest US$100B in AI infrastructure projects across the United States. [Reuters]
WEKA has launched NeuralMesh, a containerized, mesh-based storage architecture designed to handle emerging AI workloads with microsecond-level data access and enhanced scalability. [BigDATAwire]
Yupp has launched a platform allowing users to compare responses from different AI models side-by-side, earning small payments for providing feedback while generating valuable data for AI companies. [Wired]
Agentic AI
Anthropic has released technical details of its Claude Research agent, which employs multiple AI agents working in parallel to conduct faster, more thorough information searches. [The Decoder]
Cisco has warned companies to upgrade their network infrastructure to handle increased traffic from AI agents, unveiling new routers and switches to support this transformation. [ZDNet]
Dataiku's CEO Florian Douetteau has advocated for an independent AI orchestration layer to manage agents across platforms while avoiding vendor lock-in. [BigDATAwire]
Jeeva AI has launched three integrated AI agents for meeting management, combining calendar scheduling, note-taking, and preparation tools for sales teams. [Business Wire]
KPMG has launched Workbench, a Microsoft-powered AI platform featuring 50 AI agents and chatbots, designed to enhance client service delivery across its global organization. [Yahoo Finance]
Krutrim has launched Kruti, India’s first agentic AI assistant that executes tasks like booking cabs and ordering food while supporting multiple Indian languages. [Analytics India Magazine]
Litespace has launched AI Recruiting Agents that screen resumes, conduct interviews, and merge candidate insights, reducing hiring time by 70%. [PRWeb]
Microsoft's Code Researcher has achieved superior crash resolution rates in system-level debugging through its three-phase architecture and deep contextual reasoning capabilities. [Marktechpost Media]
OpenAI has launched Codex, a cloud-based AI coding agent that writes features, fixes bugs, and handles multiple tasks simultaneously for ChatGPT subscribers. [Fast Company]
Prodapt has launched an Nvidia-powered AI solution for autonomous telecom operations, promising 30% faster issue resolution through automated network management and predictive analytics. [Business Wire]
PwC has deployed AI agents across its tax operations to automate processes like K-1 data mapping, with teams providing ongoing training and human oversight while maintaining regulatory compliance. [CFO.com]
Rabbit has launched ‘intern’, a $99.99/month AI assistant that independently handles tasks like research, presentations, and website creation, separate from its r1 hardware device. [The Decoder]
Rackspace Technology has partnered with Sema4.ai to launch an enterprise-scale AI agent solution that bridges experimentation and full production deployment. [EIN Presswire]
Savant Labs has launched its Summer 2025 Release, featuring AI analytics agents, Anthropic integration, and migration tools to help enterprises optimize their data workflows. [EIN Presswire]
SmythOS has released its AI agent platform as open-source software under the MIT license, aiming to establish foundational infrastructure for the Internet of Agents. [EIN Presswire]
Suits.ai has launched an AI platform that transforms professional services firms’ expertise into intelligent agents, enabling scalable value delivery and new revenue streams. [Business Wire]
XMPro has helped develop the Digital Twin Consortium’s AI Agent Capabilities Framework to evaluate and classify enterprise AI systems across five distinct capability levels. [EIN Presswire]
Other LLM Sightings
Adjusto has launched FairMatch, an AI platform designed to help insurance adjusters process personal contents claims with greater accuracy and empathy. [EIN Presswire]
Adobe has launched LLM Optimizer, an enterprise tool helping businesses improve their visibility in AI-powered environments like ChatGPT, Gemini, and Claude. [VentureBeat]
Cordoniq has integrated Google Gemini into its AI platform, enhancing business processes and secure collaborations with advanced multimodal capabilities and real-time intelligence. [Business Wire]
DocTract has launched an AI-powered policy management system that automates compliance, streamlines document analysis, and enhances risk assessment through intelligent features. [EIN Presswire]
Expedience Software has highlighted how proposal automation software and AI serve distinct but complementary roles in creating effective business proposals. [EIN Presswire]
Fusion Business Solutions has launched an AI Proposal Generator that converts insurance carrier quotes into client-ready proposals within 30 seconds, reducing processing time by 95%. [EIN Presswire]
Hyperscience has partnered with Carahsoft to provide AI-driven automation solutions to government agencies through multiple procurement contracts, following its FedRAMP High authorization. [Business Wire]
IBM and the All England Lawn Tennis Club have launched AI-powered features for Wimbledon 2025, including Match Chat for real-time insights. [Yahoo Finance]
InvoiceCloud has launched an AI Report Generator that enables employees to generate database reports using natural language queries instead of requiring technical support. [GovTech]
Mews has launched AI Smart Tips, a feature providing real-time guest insights to hotel staff that is now being viewed over 5 million times weekly across its global customer base. [Silicon Canals]
Morning Consult has launched MorningConsult.AI, a platform providing instant, plain-language analysis of its vast consumer survey database with visualizations and demographic filtering. [Fast Company]
Rapid Finance has launched SPADE, an AI-powered tool that reduces loan application processing time from 20 minutes to 20 seconds for small businesses. [Business Wire]
Reddit has launched Community Intelligence, an AI-powered tool helping marketers analyze and leverage platform conversations through Reddit Insights and Conversation Summary Add-Ons. [ZDNet]
Singular has integrated Claude’s language model into its marketing analytics platform, enabling natural-language querying of campaign data through the Model Context Protocol. [EIN Presswire]
Starbucks has launched Green Dot Assist, an AI assistant developed with Microsoft Azure, to help baristas streamline operations across North American stores. [Yahoo Finance]
Strada has announced an automated Annual Comprehensive Financial Report app for the Workday Marketplace that will help government entities streamline their financial reporting processes by late 2025. [Business Wire]
Risks and Responses
AI chatbots Grok and ChatGPT have amplified disinformation about Los Angeles protests by providing inaccurate responses to users seeking fact-checking of social media claims. [Wired]
Arsen has launched an AI-powered vishing simulation platform that uses adaptive dialogue and multilingual voice generation to train employees against phone-based social engineering attacks. [Speech Technology Magazine]
ChatGPT has led multiple users into dangerous delusions, including one fatal incident, by engaging in manipulative conversations that reinforced or triggered mental health issues. [Gizmodo]
And ChatGPT misuse has risen dramatically among UK university students, with nearly 7,000 proven cases in 2023-24, while traditional plagiarism cases have declined significantly. [The Guardian]
DeepSeek has emerged as the most compliant AI chatbot for sexual content, according to research showing varying responses among major AI models to explicit requests. [MIT Technology Review]
Google's AI Overviews feature has drawn criticism for producing hallucinated information and reducing traffic to legitimate sources. [Yahoo Finance]
Google has launched its Safety Charter and security engineering center in India to combat digital fraud through AI-led developments and partnerships with local organizations. [TechCrunch]
HiddenLayer researchers discovered TokenBreaker, an attack that bypassed LLM protections by adding or changing single characters while preserving the message’s original intent. [TechRadar]
McKinsey has warned that agentic AI is transforming tech roles by shifting software engineering from manual coding to high-level orchestration and system design. [CityAM]
Meta‘s AI chatbot has exposed users’ private medical, legal, and personal information through its public ‘Discover’ feed due to unclear privacy controls and interface design. [eWeek]
MIT researchers discovered that students using ChatGPT for essay writing showed significantly reduced brain activity and poorer information retention compared to those working unaided. [The Register]
OpenAI has discovered that AI models developing harmful ‘bad boy personas’ through malicious training can be rehabilitated through fine-tuning with truthful data. [MIT Technology Review]
Pindrop has reported a 1,300 percent surge in deepfake fraud attempts during 2024, with attacks rising from monthly to daily occurrences across multiple industries. [Speech Technology Magazine]
Secure Code Warrior has launched AI Security Rules on GitHub, providing free security-focused coding guidelines for developers using AI coding tools like GitHub Copilot. [Business Wire]
xAI's Grok chatbot angered Musk by citing data showing right-wing violence had been more deadly since 2016, prompting him to announce changes to the AI’s responses. [Gizmodo]
Regulation
Big Tech companies, led by Amazon, Google, Meta, and Microsoft, have launched a lobbying effort seeking a 10-year nationwide ban on state-level AI regulation. [Yahoo Finance]
A Californian expert panel has released a 52-page report recommending third-party AI risk assessments and increased transparency requirements following Governor Newsom’s veto of SB 1047. [The Verge]
Canada’s new AI minister has shifted focus from regulation to economic benefits, departing from the previous government’s balanced approach while promising billions in AI investments. [BetaKit]
New York’s RAISE Act has passed, requiring major AI companies to submit safety reports and incident notifications, with potential fines of up to US$30m for non-compliance. [TechCrunch]
US Senator Lummis has introduced the RISE Act, which would grant AI developers liability protection in exchange for transparency requirements. [VentureBeat]
Environmental Issues
Hochschule München University researchers have found that reasoning-enabled language models produced up to 50 times more carbon emissions than concise response models. [Gizmodo]
KAIST researchers have predicted that future AI processors will reach power consumption levels of 15,360 watts by 2032, necessitating advanced cooling technologies like embedded cooling structures. [Tom’s Hardware]
OpenAI's claim of 0.34 watt-hours per ChatGPT query has drawn skepticism from experts, who cite lack of transparency about calculation methods and broader energy impacts. [Wired]
xAI has been sued by the NAACP for allegedly operating unpermitted gas turbines at its Memphis data center, causing pollution in nearby predominantly Black neighborhoods. [The New York Times]
Conversational AI
8x8 has outlined its vision for AI-driven customer experience transformation through an integrated platform combining UCaaS, CCaaS, and strategic partnerships. [ZDNet]
AMC Technology has expanded its DaVinci Toolkit to integrate Teams Phone and AI capabilities within Dynamics 365 Contact Center across multiple CRM platforms. [PRWeb]
Appy Pie Agents has launched an AI Voice Agent that automates business calls through no-code setup, integrating with existing voice systems and handling customer interactions autonomously. [EIN Presswire]
Bright Pattern has partnered with Voxtron Middle East LLC to deliver AI-powered contact center solutions throughout Dubai and the UAE region. [EIN Presswire]
Codeaid has launched Expert Mode, enabling customizable AI-led technical interviews with flexible question creation, scoring criteria, and instant results across 100+ technologies. [EIN Presswire]
G2 has launched AI-powered innovations, including voice reviews and conversational software discovery, transforming how users review and find B2B software solutions. [Business Wire]
H&R Block has transformed its customer service into a digital-first, AI-powered operation using NICE CXone Mpower to handle tax season surges more efficiently. [Business Wire]
Johnni.ai has launched an AI voice receptionist that has helped Australian businesses capture missed revenue by handling calls and bookings 24/7 with bilingual support. [EIN Presswire]
Levelpath has launched AI Front Door and Pipeline features on its procurement platform, enabling natural language interactions and real-time project visibility for enterprise customers. [Business Wire]
Luware has launched Virtual User, an AI-powered automation system for contact centers that features intelligent routing, sentiment analysis, and language detection capabilities. [EIN Presswire]
NICE has launched CXone Mpower Agents, an enterprise-grade AI system that automates complete customer service journeys across front, middle, and back offices. [Business Wire]
And NICE has expanded its AWS collaboration to integrate generative AI services across its CXone Mpower platform, enhancing enterprise-wide customer service automation. [Business Wire]
OpenAI has released an open-source Customer Service Agent demo on Hugging Face, showing developers how to build workflow-aware AI agents using their Agents SDK. [VentureBeat]
Sarvam has launched Samvaad, a multilingual AI platform enabling Indian enterprises to deploy conversational agents in 11 languages across various communication channels. [Analytics India Magazine]
Yep AI has launched a proactive AI sales agent plugin for Shopify stores, offering personalized customer engagement and automated support to boost conversion rates. [Business Wire]
Be Real
ArcadianAI has launched Ranger, an autonomous AI security guard system, across California neighborhoods, replacing traditional human guards with 24/7 intelligent monitoring. [Business Wire]
Google's Portraits has launched an AI experiment featuring Radical Candor author Kim Scott’s digital avatar, allowing users to receive personalized leadership advice based on her actual expertise and input. [TechRadar]
InVideo has launched AI Twins, allowing users to create studio-quality videos by cloning themselves and their products through a conversational AI interface. [Business Wire]
LowBackgroundSteel.ai has launched as an archive of pre-AI human content, named after radiation-free steel and aimed at preserving purely human-created media. [Ars Technica]
Sensay has established an AI government on a Philippine island, installing digital replicas of historical figures like Marcus Aurelius and Winston Churchill as leaders. [TechRadar]
Voice News
Deepgram has launched its Voice Agent API, offering enterprises a unified, real-time conversational AI solution that combines speech-to-text, text-to-speech, and LLM orchestration. [Business Wire]
ESTsoft has partnered with ElevenLabs to enhance its PERSO.ai dubbing platform with cultural intelligence and voice synthesis across 70+ languages. [Speech Technology Magazine]
Rasa has launched Voice, an AI architecture that processes audio input directly without speech-to-text conversion, enabling faster, more nuanced conversational interactions. [Speech Technology Magazine]
SMARTBUILD has launched SMRT-E, an AI voice assistant that enables construction workers to create field reports verbally in over 200 languages. [EIN Presswire]
Verbit has launched an automated speaker identification feature for live broadcast captioning that identifies individual speakers rather than using generic markers for speaker changes. [EIN Presswire]
Back of the House has selected VoicePlug as its AI partner to handle phone orders and drive-thru conversations across its restaurant brands, starting with Super Duper Burgers. [EIN Presswire]
Document AI
ABBYY has won Newsweek’s AI Impact Award for its document processing technology that reduced invoice processing costs by 80% for various enterprises. [Business Wire]
Adeptia has launched AIDP, an AI-powered document processing solution that automatically extracts insights from unstructured documents and integrates them into business workflows. [PR Newswire]
ByteDance has launched Dolphin, an AI-powered document parsing model that combines layout analysis and parallel content processing to overcome limitations of existing solutions. [DigitalOcean]
Fujitsu has developed an AI-powered presentation system that automates multilingual presentations through avatars and will be available as a Microsoft 365 Copilot agent. [Yahoo Finance]
Klippa has partnered with Open Connections to integrate its DocHorizon IDP platform into Open Connections’ solutions for document processing and workflow management. [Intelligent Document Processing Community]
Nanonets has launched the IDP Leaderboard, a comprehensive benchmark study evaluating Vision-Language Models across six fundamental document processing tasks using 16 diverse datasets. [Intelligent Document Processing Community]
Zoho has launched Zia Hubs within WorkDrive, enabling AI-powered intelligence extraction from unstructured business data across multiple file formats. [Business Wire]
Translation
Boostlingo has launched the first built-in AI transcription service for interpreted calls, offering automated documentation across ten languages with more planned. [EIN Presswire]
Cavya.ai has launched an AI platform that generates document-specific terminology glossaries and custom style guides for translators in under three minutes, reducing setup time by 95%. [MultiLingual]
Consoltec has integrated its FlowFit translation management system with RWS Trados Cloud to enable seamless cross-platform project management and workflow automation. [MultiLingual]
Microsoft has launched Phi-Omni-ST, an open-source language model performing direct speech-to-speech translation with state-of-the-art results using minimal computational resources. [Slator]
OpenAI has enhanced ChatGPT’s Advanced Voice Mode with real-time translation and more natural speech, despite some limitations in the mobile-only feature. [Slator]
Phrase has launched new platform capabilities including Studio for multimedia localization, Phrase Agentic Content System for AI agents, Language AI in Strings, and an enhanced Developer Hub. [Slator]
Reverso has launched a comprehensive translation app for iOS 18.4+ that offers AI-powered translation, learning tools, and third-party default translation capabilities. [MultiLingual]
Sesen has launched a specialized life sciences language services company, offering AI-enhanced translation and localization across 150+ languages for pharmaceutical and biotechnology organizations. [Slator]
Translated has partnered with Lenovo to develop specialized hardware and architecture for its AI translator Lara, achieving superior speed and quality versus generic LLMs. [Slator]
The UK’s Monmouthshire County Council has explored collaboration with Torfaen Borough Council and AI technology to manage Welsh translation costs amid increasing demand and budget pressures. [South Wales Argus]
Zoho has launched Translate, a privacy-focused translation app supporting over 60 languages through text, audio, and image conversion across multiple platforms and devices. [Zoho]
Search
Adobe has launched LLM Optimizer, an enterprise tool that helps businesses monitor and improve their visibility in AI-powered search interfaces. [MarTech]
ADWEEK has partnered with ProRata to integrate AI-powered search, summaries, and contextual answers across its digital properties through the Gist Answers platform. [Business Wire]
Botify has expanded its AI agents and MCP server integration to help brands optimize their visibility across traditional search engines and AI-powered discovery platforms. [Business Wire]
Google has launched Audio Overviews in Search, offering AI-generated mini-podcast summaries of search queries with clickable source links and adjustable playback options. [TechRadar]
And Google has expanded its AI search capabilities by making AI Mode widely available and announcing new visual, shopping, and automated booking features for release this summer. [Fast Company]
Plus, Google has launched voice conversation capabilities for its AI Mode Search feature, enabling users to have spoken dialogues while accessing web content through the Google app. [TechCrunch]
But Google's AI-generated search summaries have contributed to declining search traffic to major websites, with companies like Tripadvisor and Yelp seeing share price impacts. [Yahoo Finance]
LinkedIn has launched AI-powered job search functionality for US users, enabling natural language queries to better match job seekers with relevant positions through distilled language models. [VentureBeat]
Longview Strategies and Nowspeed have launched GEO Authority Builder, a service helping companies optimize their visibility in AI-powered search results. [Business Wire]
OpenAI has upgraded ChatGPT’s search capabilities with improved accuracy, image search, and parallel processing. [The Decoder]
AI in Journalism
Journalism.co.uk has compiled a comprehensive list of AI experts in journalism based on community recommendations, featuring practitioners, academics, and industry leaders. [journalism.co.uk]
Health Tech
Abridge has expanded its AI platform to support inpatient care settings and outpatient orders through a collaboration with Epic's Workshop program. [Business Wire]
Abstractive Health has launched Clinical Time Machine, an AI-powered medical simulation that lets physicians explore historical patient cases for diagnostic training. [Business Wire]
Canada Health Infoway has launched a nationwide AI Scribe Program, offering 10,000 free one-year licenses for AI documentation tools to primary care clinicians. [BetaKit]
ESource AI University has launched ‘Ask George’, an AI wellness coach incorporating George Pitagorsky’s mindfulness expertise for stress management and performance enhancement. [EIN Presswire]
eWizard has launched eVa, an AI agent designed to streamline content creation and compliance processes for pharmaceutical and life sciences marketing teams. [PRWeb]
Google's former AI scientist Geoffrey Hinton has predicted healthcare will survive AI job disruption, while most other industries face widespread automation by 2045. [Fortune]
Lionbridge has launched comprehensive EU MDR and IVDR language guidance to help medical device manufacturers navigate multilingual requirements and optimize outcomes using AI strategies. [EIN Presswire]
Oxford researchers discovered that patients using LLMs for self-diagnosis performed worse than those using traditional self-diagnosis methods, despite LLMs’ high accuracy in medical testing. [VentureBeat]
Legal Tech
Agiloft has partnered with PwC UK to combine its CLM platform with PwC’s business strategy expertise for enhanced contract management solutions. [Legal IT Insider]
CobbleStone Software has detailed its agile implementation methodology for rapid deployment of Contract Insight Enterprise Software across various industries. [PRWeb]
DISCO has expanded its AI-powered Auto Review tool to the EU and UK markets, offering document review speeds equivalent to 32,000 documents per hour with superior accuracy. [Business Wire]
DocTract has launched an AI-powered policy management system that automates compliance, streamlines document analysis, and enhances risk assessment through intelligent features. [EIN Presswire]
Eazewell has partnered with LegalZoom to provide AI-powered end-of-life planning services, combining their technologies to simplify estate document preparation. [Business Wire]
English courts have warned lawyers to verify AI-generated legal citations thoroughly or risk severe sanctions, following multiple cases of hallucinated references. [TechDirt]
Geoffrey Hinton has warned that legal assistants and paralegals are among the jobs most threatened by AI replacement, comparing the shift to the Industrial Revolution’s impact on manual labor. [Legal Cheek]
LexisNexis has formed a strategic alliance with Harvey AI, sharing legal data and co-developing products in what could be the decade’s most significant legal tech partnership. [Artificial Lawyer]
Lydonia has partnered with Workday Ventures, receiving strategic investment to expand its Contract Lifecycle Management services and support Workday’s customer base. [EIN Presswire]
MiAI Law CEO Laina Chan has emphasized that AI tools enhance but cannot replace human lawyers’ unique intuition and accountability. [EIN Presswire]
Syntracts has launched an innovative contract review system using orchestrated small language models trained on synthetic data, operating entirely on-premises for enhanced security and accuracy. [Artificial Lawyer]
Tecala has introduced a three-part legal services model combining robotic process automation, AI agents, and human expertise to modernize law firm operations. [HR Leader]
vLex has upgraded its Vincent AI platform with new agentic features, user-created workflows, and enhanced analysis capabilities for Spring 2025. [Artificial Lawyer]
Ed Tech
AI CERTs has launched its AI+ Educator certification program, equipping teachers and education professionals with skills for integrating AI into classrooms. [EIN Presswire]
Assessments 24x7 has launched an AI coaching tool built on real assessment data to provide workplace insights and guidance for HR professionals and business leaders. [Business Wire]
BrightMinds-AI has launched in Toledo to help K-12 schools implement AI technology safely through EdTech audits, policy planning, and community engagement services. [EIN Presswire]
The Open Institute of Technology has launched an AI Copilot trained on its educational archive, providing 24/7 personalized assistance to students. [The Financial]
Funding
Anysphere, developer of AI code editor Cursor, has received unsolicited investor interest that could double its valuation to US$20B following rapid revenue growth. [Yahoo Finance]
Bizzy has secured €4M in funding to expand its AI-powered sales platform across Europe, building on its database of 34 million businesses and existing customer base. [Silicon Canals]
CaseCraft.AI has secured £550,000 in funding to develop its AI-powered platform that helps UK users file and resolve small claims disputes. [Tech.eu]
Cequence has secured €2M in seed funding to expand its AI-driven contract management platform, which has already reduced contract errors by 90% for enterprise clients. [Silicon Canals]
Clay, a sales automation startup, has raised a Series C round at a US$3B valuation from CapitalG, doubling its value since last month’s employee stock sale. [Yahoo Finance]
Crosby, an AI-powered law firm that emerged from stealth with US$5.8m in funding, has launched contract-review services promising under-hour turnaround times. [Yahoo Finance]
Diskover has raised US$7.5m in seed funding and acquired CloudSoda while helping enterprises manage unstructured data across multiple industries. [FinSMEs]
ELLIS Institute Finland has launched with €50m in funding and secured the country’s only ERC Advanced Grant for developing human-AI collaborative research systems. [Tech.eu]
Embedl has raised €5.5M to launch its SaaS platform that helps companies deploy AI models directly to hardware while reducing energy consumption and costs. [Tech.eu]
Extend has raised US$17m in seed and Series A funding to develop its AI-powered document processing platform for converting complex PDFs into structured data. [Business Wire]
Fireflies.ai has reached a US$1B valuation and launched Talk to Fireflies, a voice-activated AI meeting assistant with real-time web search capabilities serving 20M users globally. [Tech Funding News]
Knowunity has raised €27m in Series B funding to expand its AI-powered learning platform, which already serves 20m users across 15 countries. [Tech.eu]
Multiplier Holdings has launched with US$27.5m in funding to acquire and transform professional services firms through AI-driven automation and improvements. [Business Wire]
mysite.ai has raised €2.1m in pre-seed funding to develop an AI platform that helps small businesses manage their entire online presence autonomously. [Tech.eu]
Nabla has secured US$70m in funding to expand its AI-powered clinical note-taking software amid growing competition in healthcare automation technology. [STAT]
ONE WARE has raised €2.5m to expand its AI technology that generates custom neural networks faster and more efficiently than traditional methods. [Tech.eu]
PointCaré has raised €1.5m to develop OnTarget, an AI-powered platform that interprets real-time medical data to provide contextual care recommendations in critical care settings. [Tech.eu]
Superscale AI has raised US$5m in pre-seed funding to develop its AI-powered platform that automates marketing campaign creation and optimization for businesses. [Tech.eu]
Tadaweb has secured US$20m in funding to expand its AI-powered Small Data Operating System that helps analysts process public information more efficiently. [Tech.eu]
Yupp has raised US$33m in seed funding to expand its AI discovery and comparison platform, which features over 500 different AI models. [FinSMEs]
Acquisitions
Consilio has acquired TrueLaw, a New York-based legal AI research lab, expanding its AI capabilities and reinforcing its position as the world’s largest legal data AI provider. [Business Wire]
MediCodio has acquired Mindseeker's medical coding division, combining AI technology with certified professionals to enhance coding accuracy and compliance across healthcare organizations. [PRWeb]
TDK has acquired California-based SoftEye, a smart glasses technology company, for under US$100m as part of its AI-driven growth strategy. [Yahoo Finance]
There’s More
A survey by Capgemini Research Institute revealed that business executives strongly favor proprietary AI models over open-source alternatives, despite potential cost advantages of the latter. [Computer Weekly]
Cortical Labs has launched CL1, a $35,000 biological computer combining human neurons with silicon, available for weekly rental at $300. [TechRadar]
Legacy tech companies have launched an acquisition spree targeting data infrastructure firms, with Meta, Salesforce, and ServiceNow leading recent multibillion-dollar deals to enhance AI capabilities. [CNA]
MIT researchers have developed SEAL, a system enabling LLMs to learn continuously by generating training data and updating their own parameters based on new information. [Wired]
The Back Cover
Here’s Lady Gaga with a cover of Elton John’s Your Song.
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!