Daily brief

Thursday, 8 October 2026

The last 24 hours' hottest news in AI, sorted by field, each story with a line on why it matters.

Story of the day

The single most consequential story of the day.

Installs since Sept. 8 (Sensor Tower)
6.6M
from mobile launch to iPad app
1 month
new SMB-focused connectors
8+
new retail partners
5+

Meta brings Muse to iPad with more partners on board

Meta launched a dedicated iPad app for its Muse assistant, just a month after debuting on iPhone and Android. Since Sept. 8, Muse has reached an estimated 6.6 million installs and added new business and retail connectors.

Why it's the story of the dayAI assistants are moving fast onto bigger screens and deeper into everyday tasks and shopping.

Startup of the day

A young company people are talking about today.

Desible.ai

Desible.ai raises ₹32 Cr to build AI ops layer for BFSI

Bengaluru-based Desible.ai raised ₹32 Cr (~$3.7M) in seed funding led by Prime Venture Partners, with Invention Engine participating. It offers agentic AI workflows for banks and insurers and plans to deepen capabilities, compliance and go-to-market.

Why it's worth a lookAgentic AI is moving from chat to executing regulated workflows in finance.

Inc42 ↗

Where the industry gathered

The conferences, keynotes and festivals where people met in person.

1 events

Astana AI Film Festival

Festival2 Oct
Astana

The festival showcased AI's potential in cinema and drew international attention.

  • Astana hosted the first Astana AI Film Festival.
  • President Kassym-Jomart Tokayev took part in its opening ceremony on 2 October.
  • The festival was described as the first in the world to showcase AI's potential in cinema.
AI events (Google News) ↗

Models & labs

New models, the labs behind them, and the rules they answer to.

Product

Meta’s Muse launches on iPad with new connectors and retail partners

Meta released a dedicated iPad app for its Muse assistant, a month after its iPhone and Android debut. Sensor Tower estimates 6.6M installs; new connectors include Asana, Canva, Dropbox, Figma, QuickBooks, GitHub, Klaviyo, Zoom, Notion and Granola, plus retail partners like Best Buy, Gap, Sephora, Walmart and Wayfair.

Why it mattersBigger-screen access and more integrations make task-doing AI more practical day to day.

TechCrunch AI ↗
Research

Google and Meta back Biohub’s $300M ‘virtual cell’ push

Google DeepMind, Meta and Isomorphic Labs are investing $300M in Biohub’s Virtual Biology effort to build AI datasets toward a predictive ‘virtual cell,’ alongside over $1B in U.S. federal support and resources.

Why it mattersA shared AI model of the cell could speed up disease research and new treatments.

The Verge AI ↗
Model

Liquid AI releases fast open decision models

Liquid AI released d1-3B and d1-omni-600M, open-weight models for text, vision and audio. d1-3B answers questions in 8ms on an RTX 4090 and runs on small edge devices.

Why it mattersSmall, fast models could let devices make decisions locally instead of sending data to the cloud.

liquid.ai ↗
Product

Google opens SynthID checks to everyone

Google launched a website that checks whether images, videos and audio were made by AI. It supports common media formats and uses watermarks from Google's models and tools.

Why it mattersPeople get a simple way to check some AI-made media before trusting or sharing it.

TechCrunch AI ↗
Product

Gemini Playground lets users make games by chatting

Google launched a free browser tool for adults in the US to create and edit games with text prompts. Games can be shared by link or published in a public gallery.

Why it mattersPeople without coding skills can now turn game ideas into playable projects.

The Decoder ↗
Deal

Sui and Alibaba plan agent payments without approval

Sui and Alibaba Cloud plan to let agents spend stablecoins on cloud services within a user-set budget. Neither company has announced a launch date, pricing or initial services.

Why it mattersAgents could pay for services on their own, but a bug could spend the whole budget.

Yahoo Tech ↗
Research

Meta publishes code for video reasoning research

Meta released research code testing best-of-N sampling and verifier systems for video question answering. The snapshot is for reproducing the paper's results and does not include its datasets or model weights.

Why it mattersThe code gives researchers a way to study how AI can check its answers about video.

github.com ↗
Research

New benchmark tests AI’s intuitive visual reasoning

Scale AI and Elorian released Humanity’s Sixth Sense, a benchmark covering images and video. GPT-6-astra scored 53.6%, compared with 93.1% for human annotators.

Why it mattersThe results show a large gap between AI’s reasoning and people’s everyday visual understanding.

Scale AI ↗

Agents & coding

AI that acts on its own, and the tools developers build with.

Product

Claude Haiku 5.5 lands in Devin as a low-cost sidekick

Cognition added Claude Haiku 5.5 to Devin. On its FrontierCode 1.1 benchmark, Haiku 5.5 scores 58.4%, above Sonnet 5 at about one-eighth the task cost; Devin Fusion hits 66.2 with Opus 5.5 lead and Haiku 5.5 sidekick.

Why it mattersCheaper sidekicks help coding agents cut latency and cost without losing quality.

devin.ai ↗
Product

Infinitus launches agents for pharma field teams

Infinitus launched FieldForce, a suite that prepares sales meetings, captures notes and answers provider questions. It connects with customer-management systems and supports reimbursement work.

Why it mattersAgents could help regulated healthcare teams cover more providers without adding headcount.

Fierce Healthcare ↗
Product

Atlassian gives agents identities and boundaries

Atlassian announced Agentic Multiplayer Protocol, which gives agents an identity, authority and scope inside shared workspaces. Its Rovo Work mode handles long tasks that people review and approve.

Why it mattersBusinesses want agents to act independently without losing control of their work.

SiliconANGLE ↗
Safety

Report finds control gaps in public AI agents

A Partnership on AI report says companies lack enough monitoring for publicly available agents. It warns that incomplete oversight could cause financial losses as banks, insurers and other firms adopt them.

Why it mattersBusinesses may be handing software more authority than their monitoring systems can handle.

Christian Science Monitor ↗
Product

Hark launches an agent before its devices arrive

Hark Pro can buy groceries, book cars, pay bills and refill a child's lunch card on the web and mobile apps. It is free, with $20 and $100 monthly tiers; Hark's devices are due in 2027.

Why it mattersPersonal agents are becoming useful before the dedicated hardware built to run them exists.

The Next Web ↗
Product

Sierra launches a model that spots AI callers

Sierra launched fleming-1, which scores phone audio in real time for signs that another AI agent is calling. It works with Sierra voice agents and leaves businesses to decide what happens next.

Why it mattersCompanies may soon need to know whether a customer—or an AI acting for them—is on the line.

sierra.ai ↗
Research

Israeli startups race to secure AI agents

At least 20 Israeli-founded cybersecurity startups are building tools to secure AI agents, with 17 reporting about $900 million in combined funding. Palo Alto Networks is watching the market before backing or buying companies.

Why it mattersAs agents gain access to company systems, securing their actions is becoming a major business.

CTech ↗

Video generation

The models and tools that make video.

Product

Vidu previews next-gen flagship video model with 4K output

ShengShu’s Vidu Q4 Preview launched with pricing from $0.014/sec via web and API, support for up to 15 image and three audio references, and 2K/4K at 10-bit color aimed at creators and small studios.

Why it mattersHigher fidelity and tighter control bring AI video closer to production needs.

Unite.AI ↗
Product

LTX brings video editing into Cursor

Lightricks released a Cursor plugin that connects to its hosted LTX service. Users can create and edit video and audio through an AI agent, with jobs billed to their LTX organisation.

Why it mattersDevelopers can now direct complex video and audio work from a coding tool.

GitHub ↗
Policy

Candidate shares AI video attacking Senate opponent

Mississippi Senate candidate Scott Colom shared an AI-generated video showing a woman made to look like Senator Cindy Hyde-Smith. Hyde-Smith’s campaign manager called it deceptive.

Why it mattersThe video shows how synthetic media is entering election campaigns before voting day.

The Winona Times ↗
Policy

Arizona court overturns sentence tainted by AI video

An appeals court overturned Gabriel Horcasitas' 10.5-year sentence after a judge watched an AI video depicting the shooting victim speaking. The conviction stands, but Horcasitas must be resentenced.

Why it mattersThe ruling sets an early limit on using AI recreations of victims in criminal sentencing.

Newser ↗
Policy

South Korea plans a homegrown frontier AI project

South Korea plans to launch a project for a homegrown frontier AI model early next year. A special-purpose company may lead it, with a selection expected between February and March 2027.

Why it mattersSouth Korea is preparing a new national push to compete at the leading edge of AI.

MLex ↗

Creator tools

The apps and platforms people make video with.

Policy

Italy opens probe into Suno over alleged unfair user terms

Italy’s antitrust authority is investigating AI music startup Suno, alleging terms that let it change contracts, services and pricing without user consent. Suno has not responded yet.

Why it mattersRegulators are testing how AI creator tools treat users and their rights.

Inshorts ↗
Product

ElevenLabs voice models arrive on OpenRouter

OpenRouter added nine ElevenLabs text-to-speech models and two speech-to-text models. Customers can now add speech to existing apps with one OpenRouter key.

Why it mattersDevelopers can add voices and transcription without a separate ElevenLabs plan.

openrouter.ai ↗
Product

ServiceTitan expands AI tools for contractors

ServiceTitan is opening its Max AI package to all residential in-home contractors and piloting it for commercial and roofing contractors. It also introduced the Atlas mobile app and Homh.

Why it mattersAI is being used to automate field operations, billing, and managing customer demand.

AI events (Google News) ↗
Product

Vercel brings Jev to Python developers

Vercel added an experimental evaluate() API for Jev to its AI SDK for Python. Jev is a fast, low-cost classifier that answers multiple-choice questions with confidence scores and structured JSON.

Why it mattersDevelopers can test a narrow decision-making model directly from Python.

vercel.com ↗
Product

OpenRouter launches benchmarks for model routers

OpenRouter launched a page comparing model routers on quality, speed and cost. The benchmarks cover routers that choose providers or switch between models during a task.

Why it mattersThe comparisons may help developers choose when routing is worth its extra complexity and latency.

openrouter.ai ↗
Product

Vercel will stop recording First Input Delay

Vercel's Speed Insights will stop recording First Input Delay measurements on November 1. It already uses Interaction to Next Paint, which replaced FID, and existing FID data will remain available.

Why it mattersWeb developers do not need to change anything, but FID reports will no longer update.

vercel.com ↗

Media & entertainment

AI in the films, shows, micro-dramas, music and ads people watch, the studios and platforms making them, and the rules for AI-made content.

Product

Tencent Music tests MusicBuddy, an AI music studio and launchpad

Tencent Music’s MusicBuddy combines AI song generation with multitrack editing, and ties into distribution and promotion. It’s in internal testing and uses a chat interface to adjust parts like instruments and lyrics.

Why it mattersIf it ships, musicians could go from idea to release in one tool.

RouteNote ↗
Research

AI microdramas turn cliffhangers into payments

Rappler examines short dramas built around rapid cliffhangers, ads and virtual coins. A 2026 study of 658 participants linked immersion, personalization and compulsive viewing.

Why it mattersShort-form AI entertainment is changing how audiences watch and pay.

Rappler ↗
Product

AI film enters FilmFestival Cottbus main competition

A Woman Asleep, an 80-minute AI-generated feature from Turkey, entered the main competition at FilmFestival Cottbus. Vikki Bardot made it with trained visual models and an AI-generated clone of her voice.

Why it mattersAI-made features are moving into established film festivals and mainstream cinema.

Türkiye Today ↗
Product

aTwist sets October microdrama lineup

Microdrama platform aTwist is launching four original series in October. The lineup includes Mario Lopez, Paris Hilton and Jesse Tyler Ferguson, who will direct and produce another series.

Why it mattersShort-form vertical dramas are attracting bigger stars and established producers.

Variety ↗
Product

BITE CLUB leads ALZA horror microdrama lineup

BITE CLUB, starring Barbara Crampton and Russell Todd, will premiere at New York City's ALZA Festival. The festival runs October 16–18 and also includes BLOOD SURFER, HAUNTED and HAUNTED BY YOU.

Why it mattersVertical horror is getting dedicated festival slots and theatrical-style premieres.

RUE MORGUE ↗
Product

Writers invited to create purpose-led microdramas

Havilah Malone launched a free international challenge for 10–12-episode vertical series adapted from her book. The winner gets $2,000 and will be announced at the LA Vertical Drama Market.

Why it mattersThe contest shows microdramas expanding beyond entertainment into social and personal themes.

EIN Presswire ↗

Chips & compute

Chips, data centres and the power behind them.

Deal

CoreWeave enters India with 240 MW AdaniConneX build

CoreWeave will add 240 MW of AI data center capacity across three 80 MW buildings at AdaniConneX’s Taloja campus in Navi Mumbai, planning NVIDIA Vera Rubin deployments. First phase is expected mid-2028.

Why it mattersMore AI compute is expanding to India to meet rising local demand.

coreweave.com ↗
Chips

Nvidia spreads large optimisation jobs across GPUs

Nvidia’s cuOpt added a solver for linear-programming problems across multiple GPUs. It cut peak memory per GPU by up to six times and reached speedups of up to 11.4 times on some steps.

Why it mattersBigger optimisation problems could be solved faster for supply chains and energy planning.

NVIDIA Developer Blog ↗
Deal

AMD seeks more AI chip supply in South Korea

AMD CEO Lisa Su met leaders at Samsung and SK Hynix about high-bandwidth memory and supply chains. AMD also plans an AI Center of Excellence and met 13 South Korean AI startups.

Why it mattersAI chipmakers are competing for both memory supplies and developer support.

Stocktwits ↗
Chips

HPE rolls out servers with AMD’s EPYC 9006

HPE announced four ProLiant Gen 13 servers using AMD’s EPYC 9006 processors. The range targets AI inference, analytics, virtualisation and other heavy workloads.

Why it mattersNew server designs aim to pack more AI computing into data centres.

Network World ↗
Chips

AMD says AI chip demand is outpacing supply

AMD CEO Lisa Su said demand for the company’s AI-related silicon is ahead of supply. A market report said EPYC Venice capacity was booked through 2027, with some orders pushed into 2028.

Why it mattersTight supply signals sustained pressure on AMD’s AI hardware pipeline.

Shattered ↗
Product

Anyscale on Azure reaches general availability

Anyscale on Azure lets enterprises run data processing, training, evaluation and inference in their own Azure subscriptions. It runs on Ray and Azure Kubernetes Service.

Why it mattersCompanies can keep their data and model work inside infrastructure they already manage.

Anyscale ↗
Policy

AI data centres face growing US community opposition

A Gallup survey reported that 70% of Americans oppose data centres in their communities. Three-quarters of 375 US state bills introduced in 2026 seek to restrict their development.

Why it mattersOpposition and restrictive bills signal a challenging climate for new data centers.

Diginomica ↗

Policy & safety

Laws, lawsuits and safety incidents.

Policy

Cantwell proposes federal AI safety standards

Senator Maria Cantwell proposed enforceable federal safety standards, including testing and independent audits for the most advanced AI systems. Key enforcement details remain unresolved.

Why it mattersThe proposal could put release testing and outside audits at the centre of US AI rules.

Union-Bulletin ↗
Policy

Pennsylvania advances AI ad disclosure rules

Pennsylvania lawmakers gave final approval to a bill requiring clear disclosure when AI-generated or significantly modified content could mislead consumers in advertising. It passed the Senate 48–2 and House 163–40.

Why it mattersConsumers could soon get a clearer warning when an ad uses AI to create a misleading impression.

Pennsylvania Senate Democrats ↗
Policy

AI safety coalition demands tougher federal rules

Guardrails Action led nearly 40 organisations in calling for enforceable, transparent government oversight of AI. Its coalition wants rules covering risks from existential threats to economic shocks.

Why it mattersAdvocacy groups are warning lawmakers they will oppose weaker AI legislation.

The Hill ↗
Policy

UK faces competing pressures over AI safety law

The UK government says it wants global leadership on AI safety, but has not settled where to begin. Ministers are weighing future risks against existing harms from AI systems.

Why it mattersThe UK’s approach could shape how other countries balance frontier risks with current problems.

Tech Policy Press ↗
Policy

US senators challenge White House AI safety policy

Elizabeth Warren and Richard Blumenthal pressed the White House for information about its work with tech companies on AI safety. They accused the administration of watering down voluntary guidelines.

Why it mattersSenators say softened voluntary guidelines could put the public at risk.

Fox Business ↗
Policy

Global AI rules remain a patchwork

A LexisNexis Regulatory Compliance whitepaper compares AI rules across major markets. The EU uses a prescriptive AI Act, the UK relies on existing regulators and principles, and the US has state-led rules.

Why it mattersCompanies operating across borders face different duties for the same AI system.

FinTech Global ↗
Safety

Anthropic red team found dangerous model capabilities

A Vanity Fair excerpt describes Anthropic testing Claude 3.7 Sonnet on dangerous tasks, including plans for a novel influenza strain. Researchers compared results with and without model assistance.

Why it mattersThe testing shows how labs measure whether models could make harmful work easier.

Vanity Fair ↗
Policy

Cantwell proposes federal AI testing and safety standards

Sen. Maria Cantwell released a six-point AI framework that gives NIST a role in setting standards for frontier models. It calls for continuous government and third-party testing, audits and incident reporting.

Why it mattersThe proposal would move AI safety from voluntary promises toward federal oversight.

Roll Call ↗

Robotics

Robots, self-driving and AI in the physical world.

Research

Danaher plans an autonomous drug research lab

Danaher plans to launch an AI-powered autonomous lab early next year. It will combine AI, robotics and connected instruments to design and test molecules.

Why it mattersAutomated labs could make early drug discovery faster while leaving scientists to tackle harder problems.

KSL News ↗
Research

Self-driving trucks begin California food-hauling pilot

DTL Transport and Kodiak AI began hauling perishable food between Fresno and Los Angeles on Highway 99. A human remains in the driver’s seat during the tests.

Why it mattersThe pilot is an early test of how autonomous trucks could change delivery costs and reliability.

KMPH Fox 26 ↗
Model

TwelveLabs adds video understanding for physical AI

TwelveLabs released Pegasus 1.6, a model that understands actions and changes over time in real-world video. It is designed to help train robots, drones and autonomous vehicles.

Why it mattersTurning human activity videos into usable training data could help machines learn physical tasks.

The Robot Report ↗
Product

Tesla plans Halloween Mode for FSD vehicles

Tesla says Halloween Mode is coming to vehicles on the Full Self-Driving track. The update turns the car display into a ghostly scene with skeleton pedestrians, a graveyard and themed sounds.

Why it mattersTesla is bringing a novelty feature to drivers without making them leave the FSD update track.

Not a Tesla App ↗

Science & health

AI in biology, medicine, climate and maths.

Product

Qiagen expands AI bioinformatics into discovery workflows

Qiagen introduced IPA Investigator for Ingenuity Pathway Analysis and the Qiagen Discovery Platform to extend AI into automated drug discovery, building on QDI’s 25+ years of knowledge and ~150,000 users.

Why it mattersQiagen says its AI keeps evidence visible and uses expert-reviewed information.

BioSpectrum Asia ↗
Research

Tech and government groups back a $1.8B virtual-cell project

Biohub, the US government, Google DeepMind, Meta and Isomorphic Labs are supporting a five-year effort to build AI models that predict cell behaviour. The project combines funding, datasets, computing and measurement tools.

Why it mattersVirtual cells could let researchers test some drug ideas digitally before expensive lab work.

Gadget Review ↗
Research

AI firms test virtual drug trials before human studies

AI companies are simulating drug trials to identify risks before costly human studies. BioinvestGPT predicted five of six reported trial outcomes, including a failed muscular-dystrophy trial.

Why it mattersBetter forecasts could help drugmakers avoid trials likely to fail.

KELO-AM ↗
Product

XtalPi launches an AI platform for siRNA discovery

XtalPi launched Kodexia, a platform combining generative AI, biological modelling and automated experiments. The company says it is advancing six preclinical RNA-interference programs.

Why it mattersA closed design-and-test loop could speed up the search for new therapies.

BioPharma APAC ↗
Research

Anthropic agents search DNA for unusual systems

About 950 Claude agents searched more than 200,000 reverse transcriptases, narrowed 3,500 candidates to 20 and found an unusual system with DNA repeats. Human scientists still assessed and tested the result.

Why it mattersAgent swarms could let scientists search biological databases far faster.

IBM ↗
Policy

Religious leaders adopt an AI ethics code

The Secretariat of the Congress of the Leaders of World and Traditional Religions adopted an Ethical Code for AI in the religious sphere. It aims to prevent distorted religious texts, digital forgeries, discrimination and manipulation.

Why it mattersA global interfaith body is setting expectations for AI in religious life.

Qazinform ↗
Funding

Biohub, DOE and NIH commit $1.8B to virtual biology

Biohub, the DOE, NIH and partners are investing $1.8B in data, computing and tools for predictive AI models of biology. The goal is an open resource for research and disease treatment.

Why it mattersAn open virtual biology resource could help scientists study disease and find treatments faster.

Unite.AI ↗

Open source

Open-weight models and community releases.

Funding

Nous Research raises $90M at a $1.5B valuation

Nous Research raised a $90M Series B to take its open-source Hermes Agent into businesses. The company says Hermes has been cloned more than 24 million times.

Why it mattersAn open agent project is now targeting private, multi-step work inside companies.

TechCrunch AI ↗
Research

Ai2 publishes research on byte-level language models

Ai2 published its Bolmo research in Nature and released new checkpoints. The work converts existing subword models to process the underlying bytes of text instead.

Why it mattersByte-level models may handle unusual text and other digital data more flexibly.

allenai.org ↗

Quick hits

Smaller stories worth a glance.

Deals & funding

Who raised money, and who bought whom.

5 deals

Funding rounds 4

Investors put money into a company in exchange for a share of it.

Robot Ventures backs Nous Research

Raised
$90M
Series B

Nous Research now valued at $1.5B

Nous Research raised a Series B to bring its Hermes Agent to business users.

Other investors

NvidiaUSUnion Square VenturesMVMenlo VenturesSamsung1C1789 Capital

· TechCrunch AI ↗

Partnerships & contracts 1

A large commercial agreement, or a government contract.

On the calendar

What's coming up, in date order, and how sure we are about each.

11 things
MTWTFSSOct56789101112131415161718192021222324252627282930311Nov234567891011121314151617181920212223242526272829Dec30123456
  1. 9OctAgentic AI platform for clinical trialsRethinking Clinical Trials Grand RoundsTomorrow
  2. 16–18OctALZA Festival horror microdrama programme8days away
  3. 19OctElevenLabs discount endsOpenRouter11days away
  4. 21OctTesla reports third-quarter earnings13days away
  5. 30OctMicro-Series Challenge deadlineVerticals For Good3weeks away
  6. 1NovSpeed Insights stops recording First Input DelayVercel3weeks away
  7. 5–7NovLA Vertical Drama Market4weeks away
  8. 30–3Nov–DecContent London 2026C21Media8weeks away
  9. 30NovRose d’Or Awards8weeks away
  10. 4DecAI-made film opens in limited releaseGods Don’t Give Gifts8weeks away
  11. NovemberBillboard advertising beginsGods Don’t Give GiftsExpected

Stories are summarized in plain language; follow the link before quoting a figure. Story images are shown from the publishers' own sites.