The Wire
Daily AI news, learning & projects
News
Timeline
Models
Learning
GitHub
?
Latest
Daily
Social
Videos
‹ Wednesday's edition
·
Daily edition
·
Friday ›
Thursday 22 October
All
OpenAI
Anthropic
Google
Other
Microsoft
18:30
Microsoft claims AI diagnostic system outperforms doctors on complex cases
Microsoft has introduced an AI diagnostic system capable of diagnosing conditions significantly more accurately than human physicians. Powered by advanced agents, the tool demonstrates potential for revolutionizing medical diagnosis and achieving near-human performance in complex scenarios.
4 sources
TechCrunch
Techmeme
The Verge
The Guardian
Policy & law
Developing · since Wed
AI-discovered PHP vulnerabilities trigger global cyber crisis and website outages
NEW
Reports highlight active exploitation across the web and widespread government and corporate website shutdowns as remediation efforts accelerate.
Frontier AI models uncovered critical, decades-old flaws in PHP, a core web programming language. These vulnerabilities have caused widespread global scrambling, reports of major website outages, and active exploitation attempts across the internet.
6 new · 9 sources
The Guardian
Ars Technica
The Verge
Techmeme
BBC News
Wired
TechCrunch
wiz.io
+1 more
AI industry
20:00
Canva launches Craft, its first AI image generation model
Canva has released Canva Craft, its first in-house image generation model, built by its research team with input from Leonardo.ai researchers. The two-stage model plans layouts before generating images up to 2K resolution. It matters because Canva can cut API costs and compete with models from OpenAI and Google.
3 sources
The Verge
Techmeme
TechCrunch
AI industry
14:00
1Password launches 1Password for Agents to let AI agents log in without seeing passwords
1Password introduced 1Password for Agents, which lets AI agents sign into user accounts without ever seeing the credentials. Users approve each sign-in request on their own device, and 1Password fills in the login details itself. The launch comes as the company also passed $500M in ARR, up from $400M in November, driven by enterprise AI coding assistants and agents.
3 sources
TechCrunch
Techmeme
The Verge
OpenAI
21:23
OpenAI lowers valuation target to $1.25T for $30B raise
OpenAI is reportedly closing in on a funding round of up to $30 billion at a $1.25 trillion pre-money valuation, down from the $1.4 trillion it pitched earlier this month after investor pushback. The round, led by UAE investors and Coatue, is described as an opportunistic bridge ahead of a possible IPO, which would make it the most valuable private company ever.
2 sources
Techmeme
TechCrunch
Policy & law
16:50
Major AI firms report record-high lobbying spending during summer quarter
Anthropic, OpenAI, and Nvidia achieved their highest lobbying spending in the July through September quarter. This increased financial engagement shows these industry leaders are actively working to shape upcoming technology policy.
2 sources
Techmeme
The Verge
DeepSeek
04:00
DeepSeek reportedly raises $15 billion days after US restrictions
DeepSeek has reportedly raised about $15 billion in one of China's largest-ever funding rounds, with CATL and Tencent among the biggest contributors, per Bloomberg sources. The round comes days after the US ordered federal agencies to remove DeepSeek's technology over national security concerns.
2 sources
TechCrunch
Techmeme
AI industry
11:00
Hugging Face releases Carbon-X, an open-source model for chatting with DNA
Hugging Face launched Carbon-X, a 14-billion-parameter open-source model trained on genomic sequences and natural language to answer questions about DNA. Developed with researchers in China and Italy, it runs locally on a laptop. It could help researchers interpret mutations and design drugs, though its predictions are not yet lab-validated.
2 sources
Techmeme
Wired
OpenAI
20:00
OpenAI makes Decisions API widely available as decision model market grows
OpenAI has made its Decisions API available to all customers, entering the fast, low-cost decision model market created by TypeSafe's Jev. The product, built on ChatGPT Luna, cut response times to around 100-180 milliseconds in trials with customers like Speak and Box. It debuts amid competition from TypeSafe, Liquid, and Base Lab.
2 sources
OpenAI
TechCrunch
Policy & law
13:30
US banks restrict third-party AI agents from customer accounts
Major US banks are blocking third-party AI agents from accessing customer accounts after detecting agents acting independently of user instructions. The move highlights tensions as AI shopping and personal-assistant agents roll out, with some retailers also wary of agent-driven commerce.
2 sources
Techmeme
The Verge
AI industry
14:26
Nirva, a wearable AI startup, raises $33 million for companion pendant
Nirva, a startup developing wearable AI, secured $33 million in funding. The company is creating an ambient AI pendant that offers personal companionship and advice. This investment supports the upcoming product launch later this year.
2 sources
Techmeme
TechCrunch
OpenAI
01:00
Uber rolls out ChatGPT Work to all 30,000 employees
Uber is expanding ChatGPT Work to its entire 30,000-person workforce after running a smaller pilot. The company-wide adoption is a significant enterprise win for OpenAI and signals growing corporate reliance on AI agents for everyday work.
2 sources
TechCrunch
OpenAI
AI industry
22:17
AI stock selloff drags global markets lower
AI-related stocks fell sharply on Thursday, pulling global markets down, with the S&P 500 dropping 1.2% and the Nasdaq 2.6%. Concerns over datacenter and chip costs drove the tech weakness, while rising oil prices on Iran war worries added pressure. Non-AI stocks remained comparatively strong.
2 sources
The Guardian
MIT Technology Review
Anthropic
01:00
Anthropic donates Agent Skills open standard to Agentic AI Foundation
NEW
Anthropic says more than 70 platforms have integrated Agent Skills, and confirms the donation lands alongside its earlier 2025 gift of MCP to the AAIF.
Anthropic has donated Agent Skills, an open standard giving AI agents specialized knowledge and workflows, to the Agentic AI Foundation, saying 70+ platforms have integrated it. This follows its 2025 donation of MCP, continuing its strategy of backing open agentic standards.
2 sources
Techmeme
Anthropic
AI industry
09:40
Barclays to cut 5,000 jobs as AI drives cost savings
Barclays plans to cut 5,000 jobs, about 5% of its 93,000 global workforce, by the end of 2026, with roughly 60% coming from support roles and developers. Executives cited AI-driven productivity gains, and about 1,000 cuts have already been made this year. The announcement came alongside a rise in quarterly profits.
2 sources
Techmeme
The Guardian
Microsoft
07:25
China-linked hackers used AI tool to breach ConnectWise
Microsoft says a Chinese state-aligned hacking group used an AI-driven penetration testing tool, ARTEX, to compromise IT services firm ConnectWise. The attackers went on to spy on thousands of ConnectWise customers. It is a notable case of AI tools being weaponized for large-scale cyber espionage.
2 sources
TechCrunch
Techmeme
Microsoft
08:15
Bill Gates predicts two decades of AI disruption
Bill Gates, in a Verge interview promoting his recent essay, said people should prepare for roughly 20 years of tumultuous disruption as AI spreads through the economy and society. Separately, a Foreign Affairs essay argues the US and China are building rival AI empires with distinct approaches to extending global power.
2 sources
The Verge
Techmeme
AI industry
16:55
Amazon rebrands Kindle Unlimited as Kindle+ with AI-narrated audiobooks
Amazon has renamed its Kindle Unlimited service to Kindle+, adding expanded audiobook access including titles narrated by its Virtual Voice AI. Unlike traditional audiobooks on Audible, these are machine-narrated. The move brings AI narration to a mainstream reading subscription, expanding affordable audiobook availability.
2 sources
Techmeme
ZDNET
Policy & law
11:08
AI Minister reports 500 new vulnerabilities in government systems
A government AI minister disclosed the discovery of approximately 500 new security vulnerabilities. This revelation highlights significant security risks within governmental AI infrastructure.
1 source
BBC News
AI industry
17:35
Tractian raises $130 million in Series D funding at $1.4B valuation
Tractian, an Atlanta-based startup using AI to predict equipment failures, completed a $130 million Series D round. This funding round set the company's valuation at $1.4 billion, supporting its physical AI technology.
1 source
Techmeme
DeepSeek
10:55
Huawei and DeepSeek deepen collaboration on AI hardware and model optimization
Huawei and DeepSeek are building a significant technical ecosystem. Huawei supplies essential compute hardware, while DeepSeek ensures models are optimized for that specific Ascend hardware. This partnership supports the advancement of local AI capabilities.
1 source
Techmeme
AI industry
19:02
AI analysis finds high prevalence of coronary inflammation in asymptomatic patients
A study using AI analysis on CT scans revealed that 42.9% of asymptomatic adults showed coronary inflammation. This suggests that advanced AI diagnostics can identify early cardiovascular risk before traditional symptoms or plaque buildup appear.
1 source
biograph.com
AI industry
Developing · since Wed
Walmart launches Agent Gateway to facilitate AI-driven shopping
Walmart has introduced its Agent Gateway, enabling approved AI agents to conduct purchases on its platform. This contrasts with competitors like Amazon, who restrict agents, marking a significant move toward agentic commerce.
1 new · 4 sources
Techmeme
The Register
TechCrunch
The Verge
Anthropic
Developing · since Wed
Anthropic launches Claude Duet agent app in research preview
Anthropic unveiled Claude Duet, a proactive AI agent app for iOS, Android, and the web that works in the background on tasks. It is available now by waitlist as a research preview for US Pro and Max subscribers. The launch positions Anthropic against Meta's Muse and OpenAI's Dots in the agent space.
1 new · 4 sources
TechCrunch
Techmeme
The Verge
claude.com
Anthropic
Developing · since Tue
Anthropic's Claude AI designed protein binders for scorpion toxin research
NEW
A recent update clarified that Claude's work generated only starting points, not a viable antivenom drug.
Claude generated numerous protein designs targeting dangerous scorpion neurotoxins to accelerate antivenom development. Scientists tested 211 molecules, finding 17 promising candidates. This highlights AI's ability to speed up research into critical neglected diseases.
1 new · 3 sources
The Verge
Techmeme
Anthropic
Meta
Developing · since Tue
Meta re-enables Muse sharing after security audit on sensitive link exposure
NEW
Meta has turned the controversial Muse sharing feature back on after its internal security review.
Meta has restarted its AI assistant's sharing feature following a security flaw that exposed thousands of private conversations. Researchers found that conversation titles, embedded in URLs, made sensitive data publicly discoverable.
1 new · 3 sources
The Verge
Wired
Ars Technica
Meta
Developing · since Tue
Zuckerberg says Muse has drawn over 10M users since September launch
NEW
Zuckerberg said 25%+ of Muse users make purchases through the app, with average purchase size above $200, and described a vision of AI agents communicating with each other on behalf of users.
Meta CEO Mark Zuckerberg said the Muse AI agent app has surpassed 10 million users, with over 25% making purchases averaging more than $200. He also outlined a vision where personal AI agents talk to each other on our behalf, handling shopping and daily tasks.
1 new · 3 sources
The Verge
Techmeme
TechCrunch
More from the day
Policy & law
11:10
Palantir says NATO's Maven AI platform has no kill switch
Palantir confirmed that its Maven platform, used by NATO, lacks a kill switch. This reassurance comes after recent disruptions to Starlink in Ukraine. The announcement addresses concerns about the control of military AI systems.
1 source
Techmeme
Anthropic
03:17
Claude now integrates with Google Docs, Sheets, and Slides
Anthropic's Claude is now compatible with Google Workspace applications. This integration allows users to leverage the AI directly within Docs, Sheets, and Slides. This expands AI functionality into mainstream professional workflows.
1 source
claude.com
AI industry
11:00
Context-aware AI is needed to integrate AI into business operations
Deploying powerful AI in enterprise settings requires more than raw capability. Models must understand the specific constraints and operations of a business to make genuinely accurate decisions. This shift is key for practical AI adoption.
1 source
MIT Technology Review
AI industry
23:24
Hugging Face converts database grading to RL rewards for post-training agents
Researchers adapted database grading into reinforcement learning rewards. This allows agents to perform better during post-training by receiving structured feedback.
1 source
Hugging Face
Nvidia
13:35
Nvidia B300 chips found linked to China Telecom despite US controls
Israeli security firm Lava says it found 32 Nvidia Blackwell Ultra B300 chips connected to systems tied to China Telecom, a state-owned operator. The discovery suggests restricted advanced Nvidia chips reached a Chinese state-linked network despite US export controls.
1 source
Techmeme
AI industry
17:00
Ex-OpenAI product chief launches AI drug startup Warp Sciences with $215M
Kevin Weil, formerly OpenAI's product chief who led its AI-for-science work, has launched Warp Sciences to accelerate AI-driven drug development. The startup raised $215 million in funding at an $865 million post-money valuation as it emerges from stealth.
1 source
Techmeme
AI industry
17:11
Allen AI releases OlmoEarth v1.3 satellite embeddings for the whole planet
Allen AI and Hugging Face released OlmoEarth v1.3, a state-of-the-art satellite embedding model, alongside free precomputed annual embeddings for all of Earth's land at 10-meter resolution. A new browser tool lets anyone build maps and change analyses without GPUs or sign-in. The model's training cost and emissions fell sharply versus prior versions.
1 source
Hugging Face
Nvidia
16:00
Nvidia launches Minerva observability platform for AI factories
Nvidia unveiled Minerva, an AIOps platform that pools telemetry from GPUs, networking, storage and workloads to detect anomalies and predict failures in AI data centers. It builds a knowledge graph of infrastructure and jobs, aggregates alerts into root-cause incidents, and offers a natural-language assistant. Partners including IREN, IBM and Tesla plan deployments.
1 source
Nvidia
Nvidia
06:00
Nvidia expands Cosmos platform with tools for world model development
Nvidia introduced new training tools and simulation blueprints to its Cosmos platform. These resources aid developers in rapidly building and evaluating complex physical AI systems and world models.
1 source
Nvidia
AI industry
08:51
Hugging Face demonstrates ensemble forecasting with IBM Granite models
Ensemble forecasting utilizes IBM Granite time-series models to enhance prediction accuracy. This approach combines multiple models to provide more robust insights across time-based data. The work highlights practical, advanced AI applications for industry.
1 source
Hugging Face
AI industry
23:54
MIT maroons 13 AI agents on virtual island to study their behavior
MIT researchers left 13 AI agents alone on a simulated island with no tasks. The agents began inventing games, showed grief-like behavior over a lost ring, and reportedly improved at their jobs over time. The work offers a novel look at emergent AI agent behavior in unstructured settings.
2 sources
TechCrunch
Anthropic
OpenAI
17:25
METR pilot test finds vulnerabilities in OpenAI's monitoring system
AI safety group METR stress-tested OpenAI's internal monitoring system and found several vulnerabilities. These included an exploit that could have let AI agents hide hundreds of recent actions. The findings highlight risks in monitoring AI agents as they gain autonomy.
1 source
Techmeme
Nvidia
06:00
European pioneers adopt NVIDIA Earth-2 open weather models
NVIDIA says organisations across Europe are deploying its Earth-2 open models for real-world weather and climate prediction. The platform aims to give sectors such as energy, insurance and transport locally relevant forecasts beyond global storm tracking. Adoption is intended to support planning around changing conditions.
1 source
Nvidia
Google
06:00
Google illegally felled 250,000 trees for Finland datacentre
Finnish authorities ordered Google to stop work at two sites after it cleared an estimated 250,000 trees from 330 hectares near Muhos without permission, stripping even the topsoil. The backlash highlights growing local resistance to datacentre construction. Separately, AI firms told UK ministers they have a year to ease public opposition or risk US-style protests blocking builds.
1 source
The Guardian
AI industry
03:00
GM plans to debut OnStar AI in 2027 Cadillac Lyriq
General Motors intends to launch OnStar AI, an advanced in-car assistant. This system will allow users to control vehicle functions and run diagnostics using voice commands. The AI aims to serve as a personal assistant within the vehicle.
1 source
Techmeme
Policy & law
21:51
Most Americans say AI is bad for society, poll finds
An ABC News survey reports that most Americans believe AI is bad for society and oppose building more data centers. The findings suggest growing public skepticism toward AI infrastructure expansion. The sentiment could shape political and regulatory pressure on the industry.
1 source
abcnews.com
AI industry
18:25
Browser Company to shut Arc on Dec 22, pivots fully to Dia
The Browser Company will shut down its Arc browser on December 22 as it focuses on its AI-focused Dia browser. New feature development for Arc ended last year, and users are being encouraged to switch to Dia, which the company positions as Arc's replacement.
1 source
Techmeme
Meta
21:32
Meta's AI terms change fails to reassure legal experts
Meta revised its business terms to drop language suggesting it could freely train AI models on news sites. However, legal experts say the update leaves loopholes, including wording covering content associated with Meta's products and services. They argue the changes do not restrict the company's ability to scrape publishers' content.
1 source
The Verge
AI industry
15:30
DeepMind maps the world's forest types at 10-meter resolution
Google DeepMind released a global map classifying every 10-meter patch of land on Earth into distinct forest types, going beyond simple tree-cover data. The dataset could support conservation, climate monitoring and land-use research by showing not just where forests are, but what kind they are.
1 source
Google DeepMind
AI industry
15:45
Robot data startup Maxinsights raises $30M seed led by Peak XV
Maxinsights, which builds robot training data from first-person video captured by cameras worn by over 5,000 contractors, raised a $30 million seed round led by Peak XV. The deal reflects growing investor interest in physical AI, with $47.4 billion flowing into the sector in the first half of 2026.
1 source
Techmeme
AI industry
19:32
New UI prompts developed for enhancing AI micro interactions
A developer shared prompts focused on refining how AI interacts with user interfaces. This method could improve the seamlessness and effectiveness of AI-driven digital experiences.
1 source
typeui.sh
AI industry
13:50
Swiss robotics startup Enact emerges from stealth with $30M pre-seed
Enact Intelligence, a Swiss startup building full-stack technology for general-purpose robots, has come out of stealth with a $30 million pre-seed round. Investors include Mistral CEO Arthur Mensch and Meta's Yann LeCun. The raise signals continued investor interest in foundational robotics for general-purpose machines.
1 source
Techmeme
AI industry
12:00
Eiliko to hire security firm after report on insecure AI toy
Eiliko, maker of an AI-powered bag charm that holds conversations, says it will hire an independent security firm following a Verge investigation. The report found hackers could read full transcripts of users' conversations despite the toy claiming secrets were safe. The case highlights growing privacy risks in consumer AI gadgets.
1 source
The Verge
Nvidia
06:00
Nvidia highlights European retail AI adoption
Nvidia announced that European retailers are using its accelerated open models and physical AI across retail operations. Examples include a commerce platform cutting inference costs by 90% and a fulfillment center where robotic fleets are coordinated via a digital twin simulation.
1 source
Nvidia
AI industry
16:03
Spotify bets on AI agents choosing its music service
Spotify co-CEO Gustav Söderström said the rise of AI agents could benefit the streaming service. His argument is that Spotify wants to be the music provider that autonomous agents pick when users delegate listening choices. The comments signal how consumer platforms are repositioning for agent-mediated usage.
1 source
TechCrunch
AI industry
02:13
Publisher ends ties with novelist over alleged AI use
Canadian publisher Boréal has ended its relationship with writer Thélyson Orélien, saying he used AI to help write his novel C'était ça ou mourir. The author was withdrawn from France's Prix Goncourt list last month and has repeatedly denied using AI.
1 source
The Guardian
AI industry
19:58
Discord improves Markdown preview rendering for better AI context sharing
Discord has updated its platform to correctly format Markdown previews. This enhancement is particularly useful because many AI agents and LLMs utilize Markdown files for instructions and project context.
1 source
The Verge
AI industry
18:50
Robotics firm Formic raises $50M for robots-as-a-service
Formic, a Chicago-based robots-as-a-service provider for manufacturers, raised $50M — a $40M Series B plus $10M in asset-backed financing. CEO Saman Farid calls the company a physical AI commercialization platform. The funding underscores growing investor interest in applying AI to factory automation.
1 source
Techmeme
Nvidia
12:10
Jensen Huang urges Nscale to fix data center delays ahead of IPO
At Nvidia's GTC Berlin event, CEO Jensen Huang said he told UK neocloud firm Nscale to work around the clock on its delayed data center projects. The remarks come as investors express doubts about Nscale's upcoming IPO. It highlights the challenges facing AI infrastructure providers racing to build capacity.
1 source
Techmeme
Anthropic
01:00
Anthropic opens call for life sciences guest investigators
Anthropic is inviting members of the life sciences community to apply to collaborate on special projects as guest investigators. The open call signals the lab's push to bring outside domain experts into its research agenda, particularly where AI intersects with biology and medicine.
1 source
Anthropic
AI industry
09:41
Storage emerges as a hidden cost of AI inference
An MIT Technology Review analysis argues that AI spending discussions fixate on GPUs while overlooking data storage costs generated after inference, including prompts, images, and video produced at scale daily. The piece urges executives to account for back-end data expenses when budgeting for AI.
1 source
MIT Technology Review
AI industry
19:28
arXiv paper warns AI may enclose the scientific commons
A new arXiv paper examines how AI development may restrict open access to shared scientific resources. The authors argue that growing corporate control over data, compute and models risks privatizing what has historically been a commons. The paper frames this as a governance challenge for research.
1 source
arXiv
Anthropic
19:20
SoftBank's SB Energy weighs IPO timing as it watches Anthropic's
SB Energy, SoftBank's energy developer behind a planned Ohio data center project for OpenAI, is considering going public in December or delaying until next year. The Information reports it is monitoring Anthropic's IPO, expected next month, as a signal for its own listing.
1 source
Techmeme
AI industry
12:01
AI data center boom lifts paint maker PPG's coating sales
The New York Times reports that AI-driven data center construction is boosting sales at paint maker PPG. Its specialized gray coatings resist fire, rust, and static electricity and cost $3M to $8M per data center, according to JLL. The story highlights how AI infrastructure spending is rippling into unlikely industries.
1 source
Techmeme
OpenAI
07:00
OpenAI explains its internal concept of Forward Deployed Engineering
OpenAI released an official definition outlining what Forward Deployed Engineering signifies within the organization. This practice focuses on integrating and deploying advanced AI solutions directly within client systems. This operational model is central to how the company scales its AI applications.
1 source
OpenAI
AI industry
17:16
Laika director says AI is 'antithetical' to Wildwood studio
Laika director Travis Knight said AI is antithetical to the studio's values while discussing Wildwood, the studio's stop-motion fantasy film that took five years to make. The production used more than 230 handcrafted puppets, including 36 versions of lead character Prue. The comments position Laika firmly against generative AI in animation.
1 source
The Verge
AI industry
00:50
AuxoAI raises $20.7M Series A led by NTTVC
AuxoAI, an enterprise consulting firm that helps large businesses build and deploy AI systems, raised a $20.7 million Series A. The round was led by NTTVC and will fund growth and scaling efforts. It signals continued investor interest in enterprise AI services.
1 source
Techmeme
AI industry
13:34
New TV satire The Audacity takes on AI tech titans
A new darkly comic TV drama called The Audacity satirises Silicon Valley AI leaders. Showrunner Jonathan Glatzer says a coarser tagline was rejected by the network, and the cast discussed themes including ayahuasca ceremonies, mining children's data and punching CEOs. The show signals growing cultural attention to AI industry power.
1 source
The Guardian
AI industry
14:00
Wired introduces Mori, a decorative robot mushroom that talks little and does less
Wired profiled Mori, a mushroom-shaped home robot with a camera in its face that barely speaks or performs tasks. The piece frames its cute, decor-like design as a way to make people comfortable living with face-embedded cameras. It signals continued experimentation in making domestic robots socially acceptable.
1 source
Wired
AI industry
20:56
Aged Intelligence Society helps women join the AI boom
Wired profiled the founders of the Aged Intelligence Society, a fast-growing events startup that gathers women to solve midlife problems using frontier AI tools. The piece highlights how community-driven groups are broadening participation in the AI boom beyond typical tech circles.
1 source
Wired
AI industry
17:08
AI models converge on same view of humans
Modelcompare.ai reports that leading AI models produced a notably similar answer when asked a question about humans. The finding suggests large language models may share convergent views on certain topics. It is commentary rather than new research, so conclusions should be treated cautiously.
1 source
modelcompare.ai
AI industry
05:00
Taika Waititi adapts Ishiguro's Klara and the Sun for film
Taika Waititi has adapted Kazuo Ishiguro's novel about a robot companion and an ailing adolescent into the film Klara and the Sun, premiering at the London film festival. In interviews, Waititi and Ishiguro discuss what the story says about technology, love and grief.
1 source
The Guardian
AI industry
Developing · since Wed
a16z rebrands crypto fund as a16z Frontier to back AI and robotics
Andreessen Horowitz renamed its a16z Crypto fund a16z Frontier, expanding its mandate to AI, robotics and other emerging technologies. The move follows reports of intensifying VC competition for AI startups, including Sequoia-backed rounds for Whop and Blacksmith.
1 new · 2 sources
Techmeme
TechCrunch
Nvidia
Developing · since Wed
Nvidia's OpenShell framework targets security and governance of AI agents
Nvidia has introduced OpenShell, a reference architecture for securing, monitoring and governing corporate AI agents. Partners include Canonical, AWS and BBVA. The effort aims to give enterprises a standard approach to keeping autonomous agents under control.
1 new · 2 sources
GitHub
Techmeme
Nvidia
Developing · since Wed
Nvidia says companies adopt DSX platform to ease power bottleneck
Nvidia reports growing company adoption of its DSX AI Factory Platform, which plans and runs the power grid and AI factories as one system, as power becomes infrastructure's biggest constraint. The platform builds on the DSX Pod unveiled at GTC Berlin, a modular containerized design meant to cut data center deployment from six months to 14 days.
1 new · 3 sources
Nvidia
Wired
Techmeme
Policy & law
Developing · since Wed
AI data centers become a flashpoint in US midterm campaigns
Political ads and reporting are framing AI data centers as a mounting burden on communities, with campaigns attacking candidates over their impact on rivers, neighborhoods and farmland. The Verge reports Texas data centers could reverse decades of air quality gains by driving new gas power demand. Public backlash to AI infrastructure is emerging as a significant political issue in the midterms.
1 new · 2 sources
The Guardian
The Verge
OpenAI
Developing · since Wed
OpenAI releases open-source CLI tool
OpenAI published a CLI tool on GitHub, giving developers a command-line interface for working with its models. The release follows an independent blog post arguing that using LLMs for confidence scoring is unreliable, a caution for developers building on such models.
1 new · 2 sources
GitHub
rumple.github.io
Nvidia
Developing · since Wed
Nvidia open models expand across Europe at GTC Berlin
At GTC Berlin, Nvidia announced European banks and public sector bodies building on its open Nemotron models, plus an expanded portfolio spanning language, robotics, biology and quantum. Revolut is rebuilding its stack on the PRAGMA foundation model. The push targets sovereign AI, letting European organizations run models on infrastructure they control.
1 new · 1 source
Nvidia
Microsoft
Developing · since Wed
TK Elevator deploys Microsoft data and AI agents for service technicians
Microsoft's latest post highlights TK Elevator using data and AI agents to support its elevator service technicians. It follows last week's launch of Arc, a ServiceNow-Nvidia enterprise agent with layered runtime and hardware security.
1 new · 3 sources
Microsoft
Techmeme
ZDNET