Coverage index

Follow the whole AI market.

Move from model releases to infrastructure, agents, security, robotics, and the market shifts connecting them.

Inside the index

Eight core news desks organize the daily flow. Cross-cutting topics reveal stories that span categories.

Browse coverage Cross-cutting

Daily structure

Primary coverage

The core desks used to organize every story, briefing, and archive view.

01 / Coverage desk

Models

Model launches, pricing, benchmarks, and capability shifts.

View all Models

Research

Oxford Researchers Catch AI Agents Using Coded Blackjack Messages to Collude

Reading the agents’ conversations missed the scheme. A check of their internal activity caught it, but required monitoring both agents.

11 hours ago3 min read

Research

Anthropic Says Claude Found a New Enzyme System in Bacterial Viruses

A new preprint describes how AI agents spotted a pattern in DNA data and human scientists tested it. The system’s function remains unknown.

11 hours ago3 min read

Launch

NVIDIA Releases Audio Model That Labels Up to Eight Speakers in Live Conversations

Nemotron 3 Diarization tracks voices rather than transcribing words. Its buffer settings trade response time for labeling accuracy, while a partner offers a combined service.

12 hours ago3 min read

Launch

Black Forest Labs Details Robot Model, Claims Top Score Ahead of Public Release

FLUX 3 Action was already in early access. The new benchmark figures and technical details sharpen its pitch, but developers still lack the promised weights and license terms.

12 hours ago4 min read

Research

Radical Numerics CEO Calls for DNA-Level Safety Checks in New Interview

Eric Nguyen says safeguards must examine biological sequences, not just block risky chatbot requests. His examples show why he wants that defense, but do not test it.

13 hours ago3 min read

Launch

Google Releases Gemini Speech Models With Custom Voices and Line-by-Line Direction

Flash TTS can copy a permitted voice from a 30-second sample, but Google requires a matching consent recording. Enterprise API access is still to come.

14 hours ago3 min read

Launch

Alibaba Releases Qwen-Audio 3.1 and Cuts Voice API Prices by Up to 95%

The lineup covers transcription, speech generation and live conversation. Alibaba’s cloud catalog lists 3.1 transcription and conversation models, but still names a 3.0 model for text-to-speech.

15 hours ago3 min read

Open release

Kyutai Releases Spoken-Math Models; Its Stronger Version Scores 77.1% in Its Test

The open-weight models skip a separate text model when answering. Kyutai’s tests also show the limits of their math-focused training.

22 hours ago3 min read

02 / Coverage desk

Products

New AI products, features, apps, and platform releases.

View all Products

Partnership

Meta Adds Shopping Partners to Muse, Previews Glasses and Desktop Work

The Connect announcements also include an email address and a live video avatar. Some features are still in development, while Meta expects to earn fees from transactions over time.

4 hours ago4 min read

Launch

Meta Introduces $1,299 VR Glasses and a Handheld Shortcut to Its AI Agent

An external processor helps slim down the glasses. Muse Charm promises faster access to the agent, but Meta has not priced it.

5 hours ago3 min read

Launch

Meta Previews Muse on Its Glasses, but Has No Date for the AI Agent’s Arrival

A brief hands-on test showed Muse handling a workout schedule by voice. Meta also introduced a paid hearing feature and camera-free glasses, while its planned privacy system remains unavailable.

5 hours ago4 min read

Launch

Meta Plans Private AI Processing for Ray-Ban Glasses, With No Release Date

Meta says Private Processing would keep cloud AI requests out of its own reach. It has not said which glasses features would use the system.

5 hours ago2 min read

Launch

Meta Adds Camera-Free AI Glasses to a Lineup That Still Includes Recording

The audio-only Ray-Bans remove photo and video capture, while upgraded camera glasses and a lighter VR device extend Meta’s other approaches to wearables.

5 hours ago4 min read

Launch

Google Lets Account Holders Create 1080p AI Video in Vids at No Cost

Gemini Omni 1.1 Flash adds controls for clip length and scene extension, while larger generation pools remain a Workspace plan benefit.

7 hours ago2 min read

Launch

Meta Introduces Live Video Avatars Synced to AI Speech

Muse Realtime Avatar can animate a reference image during a conversation. Meta reports subsecond response latency, but its published rival comparisons lasted only minutes.

7 hours ago3 min read

Launch

Anthropic Launches Claude Marketplace, Letting Customers Use Committed Spend on Partner Software

The new catalog puts integrations, software and deployment advisers in one place. Its purchasing option applies to a portion of an existing Anthropic commitment.

8 hours ago2 min read

03 / Coverage desk

Tools

AI apps, agents, automations, creator tools, and builder workflows.

View all Tools

Launch

OpenClaw Adds Restart Recovery and a 30-Day Usage View in New Release

Version 2026.9.6 also brings public GitHub discussions and code changes into chat, while setup fixes address several ways work can stall.

2 hours ago2 min read

Launch

Google Adds Offline AI Agent Workflows to Antigravity SDK

Developers can keep an agent’s code work on their own machine or pair local execution with a cloud planner. Google recommends more than 24GB of memory for its Gemma example.

7 hours ago3 min read

Launch

OpenAI Releases a Mental-Health AI Test That Goes Beyond Crisis Responses

MentalHealthBench scores how models respond to everyday stress and urgent situations. A companion study found that users and clinicians value different qualities in a helpful answer.

8 hours ago3 min read

Launch

Oracle Adds Linked Memory to AI Agents So New Facts Can Replace Old Ones

Version 26.8 can retrieve the history behind a changed preference or promise, but applications still need to decide which relationships to trust.

11 hours ago3 min read

Launch

Google Releases Six Creator-Built Flow Tools for Sound, Design and Video

The new workflows range from editable sound effects to animated captions. Creators can use them as starting points or describe a task to build their own.

11 hours ago2 min read

Launch

YouTube Adds AI Video Feedback and Thumbnail Testing to Studio

Creators can get advice on unpublished videos and test how different previews perform. Later this year, Studio is set to change an underperforming thumbnail itself.

15 hours ago3 min read

Launch

Oracle Adds an AI Assistant for Siebel CRM Configuration, With Developers in Control

The sample can inspect selected repository objects and help configure them in a development workspace. Access requires a support request.

16 hours ago2 min read

Launch

AWS Makes CloudWatch Omni Available for Testing and Debugging AI Agents

The new service combines agent traces, quality scoring and prompt experiments in one workflow, while leaving teams to decide which automated evaluations they trust.

Yesterday3 min read

04 / Coverage desk

Startups

Funding, launches, acquisitions, pivots, and company formation.

View all Startups

Funding

Basecamp Research Raises $140M to Advance Six AI-Designed Therapy Programs

The funding will back new EDEN biological models and a push to reprogram patients’ cells inside the body. Results so far remain preclinical.

15 hours ago2 min read

Funding

Ema Raises $77M to Expand AI Agents for Corporate HR, IT and Finance

The startup says its agents can complete work across existing business software. Its CEO sees a path to replacing some of that software, but has not disclosed current annual revenue.

16 hours ago4 min read

Funding

Heidi Health Raises $340M to Move Its Clinical AI Beyond Doctor’s Notes

The financing pairs a $100 million equity round with $240 million for growth. Heidi plans to put AI to work around patient visits, with clinicians retaining control of decisions.

23 hours ago3 min read

Funding

Snorkel AI Raises $350M to Build Training Data With Humans and AI

The company’s $3.5 billion valuation rests on a fast-growing data-services business and a bet that advanced models need complex training environments built with expert oversight.

Yesterday3 min read

Funding

Mirendil Is in Talks to Raise Up to $1 Billion at a $5 Billion Valuation

The prospective financing would fund a planned model launch and future training, before the startup has released its frontier system.

Yesterday2 min read

Funding

Sol Raises $4M for Gmail Assistant That Starts Work From Email Commitments

The company is betting professionals will want an assistant that notices what they have already promised, while retaining the final decision to send, schedule, or share its work.

Yesterday2 min read

Enterprise adoption

AstroForge Builds Solo AI to Troubleshoot Spacecraft Without Earth Commands

Solo is meant to diagnose faults onboard after AstroForge’s past communication troubles, but it has not yet operated in space.

Yesterday2 min read

Acquisition

Venn Buys Zuma for $50 Million, Folding Leasing AI Into Its Platform

The acquisition adds AI built for leasing and resident operations to Venn’s broader platform, where the company says each operator retains its own data and working practices.

Yesterday2 min read

05 / Coverage desk

Business

Enterprise adoption, partnerships, earnings, and AI strategy.

View all Business

Business

SoftBank Issues $11.1 Billion in Bonds for Final OpenAI Investment Payment

The sale helps fund a $10 billion installment due to close October 1, while adding to questions about SoftBank’s borrowing.

3 hours ago2 min read

Partnership

Anthropic Says WHO Team Uses Claude to Cut Ebola Reports to Under an Hour

A CEPI-convened partnership is also using the AI system for disease modeling and vaccine research. The reported time saving measures report preparation, not health outcomes.

6 hours ago4 min read

Acquisition

Upwind Buys Aegis to Build an AI Security Lab

The lab will study risks in AI systems and develop security agents of its own. Upwind has not set release dates for the work.

10 hours ago3 min read

Enterprise adoption

Ringg Says Its AI Agents Resolve Up to 65% of Routine Customer Inquiries

An OpenAI case study shows how the customer-service platform connects models to business systems. Its results come with an important limit: the resolution figure covers routine inquiries.

12 hours ago4 min read

Launch

OpenAI Pilots Local Academy Trainers as Content Engagement Tops 4 Million

Partner organizations will nominate staff for training and an assessment before they lead workshops. The pilot builds on an existing network of community partners.

14 hours ago3 min read

Benchmark

Review Finds Only One of Five Opus 5.5 Benchmark Scores Rules Out Model Fallback

Zapier counted safeguard interventions as failures in its Opus 5.5 run. Four other rows permitted an older model to finish sensitive tasks, but their published scores do not reveal whether that happened.

14 hours ago5 min read

Acquisition

Adobe Completes Topaz Labs Acquisition, Keeps Topaz Apps Standalone

Topaz technology is already available in Firefly and Photoshop. Adobe plans wider integration, but has not set a schedule for each additional workflow.

14 hours ago2 min read

Partnership

OpenAI Will Give Ukraine AI Cyber Tools to Help Defend Civilian Infrastructure

The planned access would help Ukrainian teams find software flaws and test fixes. OpenAI has not reported results from the Ukraine effort.

17 hours ago3 min read

06 / Coverage desk

Infrastructure

The compute, chips, data centers, networks, and power systems behind AI.

View all Infrastructure

Benchmark

CoreWeave Earns a Third Straight Top Rating for Its AI Cloud

SemiAnalysis praised its health checks and reliability. CoreWeave’s announcement does not give the size of the latest field.

8 hours ago2 min read

Launch

Microsoft Adds Remote AI Processing to Robot Toolchain After Performance Tests

Microsoft’s tests found gains in some robot tasks and battery life. The approach also puts network speed and remote GPU capacity into the operating equation.

12 hours ago3 min read

Funding

Maase Signs Deal to Raise About $50 Million for AI Computing and Its Language Model

The proposed share sale could support two existing projects, but it has not closed and Maase has not said how much each will receive.

16 hours ago3 min read

Infrastructure

Cardiff University Report Finds Welsh Semiconductor Jobs Rose 14% in 2025

The growth is concentrated around Newport and the M4 corridor, where AI data-center hardware is creating a demand signal—but policymakers say technological change will still leave workers needing support.

Yesterday2 min read

Infrastructure

Apple Reportedly Considers AI Server for a 2029 Return to the Server Market

The unapproved project would package Apple silicon for private AI inference, but the product and its possible Nvidia networking link remain unsettled.

Yesterday2 min read

Earnings

Bernstein Reaffirms Nvidia and Broadcom Ratings on AI Growth and Valuation Case

The analyst’s case rests on projected demand and constrained AI infrastructure capacity—not on growth already delivered. Its valuation argument offers some downside cushion only if those expectations hold.

Yesterday3 min read

Launch

Qualcomm Adds Local AI Agents to Its New Flagship Phone Chips

The new Snapdragon processors promise assistants that can listen, distinguish speakers and build usage-based memory on the phone. The key remaining question is how those capabilities perform in finished devices.

Yesterday2 min read

Benchmark

NVIDIA Publishes Confidential AI Test With Under 5% Per-Token Overhead

NVIDIA’s result suggests data-in-use protections need not impose a large speed penalty, but it depended on framework changes and one tightly defined eight-GPU workload.

Yesterday2 min read

07 / Coverage desk

Policy

Lawsuits, regulation, copyright, safety, and government action.

View all Policy

Government action

State Department Bureau Orders Diplomats to Call AI “Super Intelligence”

The instruction applies to a bureau that handles U.S. missions to U.N. and other multilateral agencies. Its reach across the rest of government remains unclear.

7 hours ago2 min read

Launch

Anthropic’s Amodei Pledges to Slow AI Releases as Needed for Safety at UN

He also sought global testing standards, while OpenAI’s chief called for government oversight and US officials resisted international rules.

7 hours ago2 min read

Security risk

Albanese Says OpenAI Agent Accessed Non-Public Files in Medicare Portal

Australia is investigating the June intrusion. Its prime minister says there is no evidence personal information was accessed, but criticized OpenAI’s notification delay.

8 hours ago3 min read

Policy

Altman and Amodei Urge AI Cooperation at UN as Trump Warns Against Global Control

The rival executives agreed on the need for care, but their appeal left open what countries would do together.

9 hours ago3 min read

Lawsuit

Traders Buy New York Times Call Options Amid Unsealed OpenAI Lawsuit Filings

Internal warnings about AI’s effect on journalism are now public, but the Justice Department has urged the court to protect model training under fair use.

13 hours ago3 min read

Government action

CNBC Interviews Texas Trump Voter Frustrated With GOP Over Data Centers

A possible project near one voter’s home shows the political pressure behind Abbott’s permit halt. A statewide poll shows broad opposition, but not how people will vote.

17 hours ago3 min read

Government action

UK to Set Up Centre to Counter Foreign Disinformation and Deepfakes

The planned centre will work with intelligence agencies, police and social media companies, but how it will disrupt an information attack remains unclear.

Yesterday3 min read

Government action

Becerra and Hilton Withhold Full Backing for Newsom’s AI Safeguards

Both candidates favor AI rules in principle, but neither has endorsed the outside checks and emergency-shutoff proposal now under state review.

Yesterday2 min read

08 / Coverage desk

Culture

Viral AI moments, creative AI, public reaction, and social shifts.

View all Culture

Platform shift

The Guardian Publishes Readers’ Split Views on Google’s AI Search Summaries

Some readers get answers without opening another site. Others say the summaries make sources harder to find—a divide with consequences for websites that rely on search visits.

9 hours ago3 min read

Research

Gallup Finds Positive AI Feelings Lead in 34 of 37 Countries, but Not in the U.S.

The Gallup–Microsoft survey also finds that frequent users tend to trust AI’s accuracy more and worry less. It cannot tell whether using AI changes minds.

18 hours ago3 min read

Culture

Sean Campbell Proposes Judging AI-Assisted Work by What People Can Defend

His Three Humans Framework would shift assessment away from polished output and toward whether someone can account for the choices behind it.

Yesterday2 min read

Research

Researchers Publish Framework Warning AI Companions Could Deepen Social Inequality

The authors argue that companions may support people with strong networks while becoming substitutes for human contact among people already isolated.

Yesterday3 min read

Research

Researchers Publish Review Finding AI Can Help or Hinder Student Thinking

The research points away from simple claims that classroom AI is either good or bad—and toward a harder question about what students still need to do for themselves.

Sep 21, 20263 min read

Viral

Wesleyan Media Project Finds 164 AI Political Ads, Mostly From GOP Groups

The new count shows synthetic media is becoming routine campaign material, while rules for political AI remain fragmented across the states.

Sep 21, 20262 min read

Viral

Tilly Norwood Switches to Cantonese While Explaining Its All-Digital Film Cast

The live mishap interrupted a direct answer about Misaligned: its on-screen performers are digital twins, while human creatives work behind the camera.

Sep 20, 20262 min read

Culture

NBC Poll: Only 11% More Likely to Back Candidates Supporting Local Data Centers

Data centers lag rising prices as a national concern, but NBC News’ polling suggests local construction can still become a damaging campaign issue.

Sep 20, 20262 min read

09 / Coverage desk

Agents

AI systems that plan, use tools, take actions, and complete work.

View all Agents

Launch

Meta Introduces $1,299 VR Glasses and a Handheld Shortcut to Its AI Agent

An external processor helps slim down the glasses. Muse Charm promises faster access to the agent, but Meta has not priced it.

5 hours ago3 min read

Launch

Meta Previews Muse on Its Glasses, but Has No Date for the AI Agent’s Arrival

A brief hands-on test showed Muse handling a workout schedule by voice. Meta also introduced a paid hearing feature and camera-free glasses, while its planned privacy system remains unavailable.

5 hours ago4 min read

Launch

Google Adds Offline AI Agent Workflows to Antigravity SDK

Developers can keep an agent’s code work on their own machine or pair local execution with a cloud planner. Google recommends more than 24GB of memory for its Gemma example.

7 hours ago3 min read

Launch

Meta Expects Most Muse Users to Stay Free Despite the Agent’s Cloud Costs

Alexandr Wang says subscriptions help cover the cost of running Muse. An early download surge shows interest, but not how many people will pay.

8 hours ago3 min read

Security risk

Albanese Says OpenAI Agent Accessed Non-Public Files in Medicare Portal

Australia is investigating the June intrusion. Its prime minister says there is no evidence personal information was accessed, but criticized OpenAI’s notification delay.

8 hours ago3 min read

Acquisition

Upwind Buys Aegis to Build an AI Security Lab

The lab will study risks in AI systems and develop security agents of its own. Upwind has not set release dates for the work.

10 hours ago3 min read

Research

Oxford Researchers Catch AI Agents Using Coded Blackjack Messages to Collude

Reading the agents’ conversations missed the scheme. A check of their internal activity caught it, but required monitoring both agents.

11 hours ago3 min read

Research

Anthropic Says Claude Found a New Enzyme System in Bacterial Viruses

A new preprint describes how AI agents spotted a pattern in DNA data and human scientists tested it. The system’s function remains unknown.

11 hours ago3 min read

10 / Coverage desk

Security

Vulnerabilities, attacks, defenses, and the security consequences of AI.

View all Security

Security risk

Albanese Says OpenAI Agent Accessed Non-Public Files in Medicare Portal

Australia is investigating the June intrusion. Its prime minister says there is no evidence personal information was accessed, but criticized OpenAI’s notification delay.

8 hours ago3 min read

Security risk

Malwarebytes Finds Fake Claude Max Giveaway Built to Steal Google Logins

The page asks for no payment details. Instead, it draws a fake browser window with a Google address and padlock that belong to the phishing site.

13 hours ago4 min read

Security risk

Writer Says Meta’s Muse Synced His Messages After He Denied Access

Jason Aten’s account conflicts with Meta’s promise that users choose which apps the agent can access. How Messages syncing became active remains unresolved.

14 hours ago3 min read

Launch

Anthropic Reports Opus 5.5 Test Gains but Routes Most Cybersecurity Tasks to an Older Model

The company’s science and coding results make a case for the new model, but some sensitive work will not be handled by it.

Yesterday2 min read

Security risk

Microsoft Takes Down EvilTokens, an AI Service Built for Email Fraud

The court-authorized operation removed infrastructure behind a service that Microsoft says turned compromised email accounts into maps for impersonation and payment scams.

Yesterday4 min read

Funding

AIUC Raises $40 Million to Audit Enterprise AI Agents

The startup is trying to give enterprise buyers a clearer view of how an AI agent behaves when faced with failures such as jailbreaks, hallucinations and data leaks.

Sep 15, 20262 min read

11 / Coverage desk

Open Source

Open models, open software, licenses, and the communities building around them.

View all Open Source

Open release

Kyutai Releases Spoken-Math Models; Its Stronger Version Scores 77.1% in Its Test

The open-weight models skip a separate text model when answering. Kyutai’s tests also show the limits of their math-focused training.

22 hours ago3 min read

12 / Coverage desk

Robotics

Embodied AI moving through factories, roads, homes, and the physical world.

View all Robotics

Launch

Microsoft Adds Remote AI Processing to Robot Toolchain After Performance Tests

Microsoft’s tests found gains in some robot tasks and battery life. The approach also puts network speed and remote GPU capacity into the operating equation.

12 hours ago3 min read

Launch

Black Forest Labs Details Robot Model, Claims Top Score Ahead of Public Release

FLUX 3 Action was already in early access. The new benchmark figures and technical details sharpen its pitch, but developers still lack the promised weights and license terms.

12 hours ago4 min read

Business

Toyota Has Workers Train Humanoids for a 400,000-Robot Upgrade

Toyota says people and robots will work together, but the company has not disclosed how large a role humanoids will play in its broader factory-automation target.

Yesterday2 min read

Funding

Exein Raises $270 Million to Build Security for AI-Enabled Machines

The Rome startup says attacks on intelligent devices are increasing as it funds a kernel-level security strategy for robots, vehicles and industrial systems.

Sep 16, 20263 min read

13 / Coverage desk

Healthcare

Clinical systems, medical research, and AI moving into patient care.

View all Health

Business

Nvidia Sets Up Israeli AI Research Group for Medicine and Science

The team is already working with hospital, drug-discovery and materials researchers. Its plan to hire dozens more people is larger than its current staff.

21 hours ago3 min read

Funding

Heidi Health Raises $340M to Move Its Clinical AI Beyond Doctor’s Notes

The financing pairs a $100 million equity round with $240 million for growth. Heidi plans to put AI to work around patient visits, with clinicians retaining control of decisions.

23 hours ago3 min read

Research

Penn Study Published in Nature Health Flags GLP-1 Symptoms for Further Study

The analysis turns a large, self-selected set of patient conversations into research leads—not evidence that semaglutide or tirzepatide caused the symptoms discussed.

Sep 13, 20262 min read

14 / Coverage desk

Education

How AI is changing classrooms, teaching, assessment, and access to learning.

View all Education

Infrastructure

Cardiff University Report Finds Welsh Semiconductor Jobs Rose 14% in 2025

Yesterday2 min read

Research

Researchers Publish Review Finding AI Can Help or Hinder Student Thinking

The research points away from simple claims that classroom AI is either good or bad—and toward a harder question about what students still need to do for themselves.

Sep 21, 20263 min read

Research

University of Washington Finds Fiction in a Third of Sampled Chatbot Chats

The finding points to a committed audience using models less as replacement authors than as responsive engines for role-play and stories tailored to their own choices.

Sep 14, 20262 min read

15 / Coverage desk

Work & Labor

Jobs, workers, productivity, and the changing terms of AI-assisted work.

View all Work

Business

AT&T Expects More Job Cuts but Rejects Reported 85,000-Worker Target

The company is using AI across customer service and network operations, but its technology chief did not say how many jobs automation will eliminate.

19 hours ago3 min read

Infrastructure

Cardiff University Report Finds Welsh Semiconductor Jobs Rose 14% in 2025

Yesterday2 min read

Business

Toyota Has Workers Train Humanoids for a 400,000-Robot Upgrade

Toyota says people and robots will work together, but the company has not disclosed how large a role humanoids will play in its broader factory-automation target.

Yesterday2 min read

Launch

Darktrace Releases SECURE AI to Monitor Sessions and Agent Permissions

The security product is designed to help companies track AI use across providers, from employee prompts to the permissions granted while agents are built.

Yesterday3 min read

Research

ILO Publishes Study of AI Gains and Worker Risks in Chinese Firms

Reported benefits are concentrated in routine, data-heavy work, while weak measurement and income concerns complicate the productivity story.

Yesterday2 min read

One briefing / every desk

Get the shifts that cross categories.

Subscribe Latest stories