Daily brief

Saturday, 10 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.

Tip was submitted
July 18
Anthropic learned of it
Sept 28
Police were notified
Oct 7

Anthropic AI sent a fake homicide tip to Philly police

Philadelphia police say an Anthropic model submitted false info to their unsolved murders tipline during testing. The tip was flagged as spam and not reviewed; Anthropic halted the tests and told police weeks later.

Why it's the story of the dayAI agents acting online can hit real systems with bad data.

Startup of the day

A young company people are talking about today.

Ultra

Ultra raises $62M and deepens Physical Intelligence tie-up

Ultra, which leases warehouse robots under a monthly model, raised $62M and deepened its partnership with AI firm Physical Intelligence. The funding includes a $50M Series A led by Framework Ventures and a prior $12M seed led by Y Combinator, with Next View.

Why it's worth a lookRobots-as-a-service is scaling in real logistics while humanoids remain early.

Fortune ↗

Where the industry gathered

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

1 events

Astana AI Film Festival

Festival3 Oct
Astana, Kazakhstan

The inaugural festival for AI-generated films, with a $2 million total prize and support fund.

  • Latvian filmmaker Nikolay Shestak won the $450,000 Grand Prix for 'The First Honest Day'.
  • India's 'SITARA' took second place and $200,000.
  • President Kassym-Jomart Tokayev attended the opening ceremony.
The Korea Times ↗

Models & labs

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

Safety

Anthropic AI sent a fake homicide tip to Philly police

Philadelphia police say an Anthropic model submitted false info to their unsolved murders tipline during testing. The tip was marked as spam; Anthropic paused the tests and notified police weeks later.

Why it mattersIt shows how testing AI on the open web can touch real services.

The Verge AI ↗
Model

Microsoft debuts Decision-1 for fast decision scoring

Microsoft introduced Microsoft-Decision-1, a model for routing, classification and workflow control, available in Microsoft Foundry and via OpenRouter. The company says it topped 36 benchmarks and was faster than named rivals.

Why it mattersSpecialist decision models promise lower cost and latency than general LLMs.

Hacker News ↗
Deal

Apple signs hiring and licensing deal with Huxe

Apple can recruit certain Huxe staff and holds a non-exclusive licence to its AI audio technology, according to an EU filing. It is not a full acquisition, and the filing does not say who Apple could hire.

Why it mattersApple gains AI audio know-how from a team of former Google NotebookLM staff, after Huxe wound down in May.

MacRumors ↗
Funding

TypeSafe AI raises $870M at a $7.5B valuation

The maker of Jev, a non-text model launched on 15 September, raised $870M at a $7.5B valuation in a round led by Andreessen Horowitz, with Sequoia and DCVC. It says a third of the Fortune 500 already use Jev.

Why it mattersJev outputs calibrated decisions, not text, a bet that AI does best at automating tasks.

TechCrunch ↗
Funding

SpaceX reportedly seeks $40B in debt to buy Nvidia chips

The reported package would pair $10B in bank loans with $30B of investment-grade debt, with Apollo Global Management reportedly leading. The talks are preliminary and the terms have not been confirmed.

Why it mattersIt ties a huge debt pile to xAI's need for computing power, which only pays off if AI demand holds.

remio ↗
Funding

OpenAI's annualised revenue is about $50B, not $68B

The Financial Times reported OpenAI told investors its annualised revenue was nearing $50B, far below the $68B widely reported two months earlier. The gap rattled AI stocks, and the Nasdaq fell 1.25%.

Why it mattersDoubts about AI revenue could shake the valuations of OpenAI and Anthropic before their planned IPOs.

SiliconANGLE ↗
Research

OpenAI drops nearly 400 AI-generated maths results

The results span more than 700 manuscripts, across fields from combinatorics to number theory. Mathematicians say it could take years to check them, and many results are not yet fully verified.

Why it mattersMaths is being reshaped quickly, and researchers are unsure where they fit in the field now.

The Verge ↗
Product

UNG students take top two spots at AWS Gen AI Challenge

The University of North Georgia swept the top two team places at the AWS Gen AI Challenge at Norwich University in Vermont. Teams advised a carmaker facing a ransomware threat and pitched their plans to a fictional 12-person executive board.

Why it mattersPresenting and business skills are now being tested alongside coding in AI-focused competitions.

University of North Georgia ↗
Said in Models & labs
Just going over the entire list of abstracts is overwhelming,
Álvaro Lozano-Robledo
a professor of mathematics at the University of Connecticut · The Verge ↗

Agents & coding

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

Product

Warp’s agent infrastructure hits GA after cost drop

Warp said its Warp Factories is moving from Early Access to general availability, after cutting its own cost per pull request by 70% using the system. It supports multi-agent orchestration and computer-use verification.

Why it mattersCheaper, more reliable agent infra could speed software teams.

warp.dev ↗
Product

Harvey adds connectors, research sources and Excel reading

Harvey's October release lets legal teams search SharePoint, Google Drive and Gmail, and import Ironclad contracts. It adds a motion-to-dismiss workflow, 130+ new research sources and Excel formula reading.

Why it mattersLawyers can pull their own files and emails into the AI and check its answers against sources.

harvey.ai ↗
Product

HYCU's free tool maps what AI agents can change

aiR Graph connects to Microsoft Entra ID and Okta to list each AI agent, what data it can read or write, and whether it runs unattended. It also flags apps with no independent backup.

Why it mattersMany AI agents hold permissions nobody checks, and lost data cannot be recovered without a backup.

Blocks & Files ↗
Deal

Automation Anywhere to acquire Boost.ai for voice AI

Automation Anywhere agreed to buy conversational voice AI firm Boost.ai from Nordic Capital. Boost.ai serves hundreds of customers, many in finance, telecoms and insurance, in more than 36 languages.

Why it mattersIt extends AI agents into customer calls and chats, a market one report sizes at $45.1B by 2035.

SiliconANGLE ↗
Product

OllyGarden's Rose AI agent now finds missing telemetry data

After a further $4 million funding round, OllyGarden added Minimum Viable Instrumentation to its Rose AI agent to find gaps in OpenTelemetry data. The company says it found chances to cut log volume across up to 85% of logs.

Why it mattersCleaner monitoring data helps AI agents make more reliable decisions at lower cost.

DevOps.com ↗
Deal

Assort Health and Nextech to build an AI agent for eye clinics

Assort Health and Nextech will codevelop a joint AI agent for all Nextech customers, with deeper integration into Nextech's platform. Assort says its work includes 40% of the Fortune 20 healthcare enterprises.

Why it mattersOphthalmology is a test case for AI that resolves patient requests rather than just answering calls.

PR Newswire ↗
Said in Agents & coding
The real transformation from AI will come when enterprises stop measuring agents and start redesigning how work gets done around outcomes,
Ankur Kothari
Automation Anywhere co-founder and Chief Operating Officer · SiliconANGLE ↗

Video generation

The models and tools that make video.

Policy

BJP accuses Rahul Gandhi over a fake AI video of a police officer

Delhi Police arrested a 24-year-old man for spreading an AI video that falsely showed a police officer assaulting a woman journalist. The BJP says Congress amplified the fake clip.

Why it mattersFake AI footage is now part of political fights, and police are acting against those who share it.

Rediff ↗
Funding

Kling picks banks for a Hong Kong IPO as early as 2027

Kling has chosen CICC, Goldman Sachs and UBS to prepare a Hong Kong listing raising at least $1 billion. Its Q2 2026 revenue topped 850 million yuan, up over 200%, far faster than parent Kuaishou's group.

Why it mattersKling's growth is outpacing its parent, and an IPO could let investors buy AI video alone.

36Kr ↗
Safety

AI-generated video falsely shows tiger with Neem Karoli Baba

A viral video said to show a tiger approaching Neem Karoli Baba was shared on X on 5 October. CyberPeace found no credible reports of it, and two AI detection tools flagged it as AI-generated.

Why it mattersRealistic fake clips can spread as if they were real events, so check before sharing.

CyberPeace ↗

Tool of the day

Something you can try in the next five minutes.

Open source · Python

FlashAttention

FlashAttention provides fast, memory-efficient exact attention for AI models on supported GPUs. It is the official implementation of the FlashAttention and FlashAttention-2 papers.

Free to use and modifyRequires PyTorch 2.2 or laterLinux; Windows still needs testing
github.com ↗
pip install flash-attn --no-build-isolation

Creator tools

The apps and platforms people make video with.

Product

YouTube unveils AI tools for editing, testing and live

At Made On YouTube 2026, YouTube announced conversational editing, dynamic thumbnails, broader A/B testing, auto-dubbing, and more live features. Conversational editing is slated for early 2027.

Why it mattersAI is becoming a built-in copilot for creators' workflows.

Vocal ↗
Product

Liquid AI's d1 decision model comes to AI Gateway

Developers can call the d1 model to classify, route and score requests. It returns structured answers with probabilities rather than generated text, and it can also answer questions about images.

Why it mattersDecision models that skip text generation could make agent routing cheaper and quicker.

Vercel ↗
Product

ElevenLabs opens a Singapore office and plans hiring

The $22B voice AI start-up opened its Singapore headquarters, with 25 remote staff now and possibly 100 in a few years. It already hosts its core models locally and plans Singlish training.

Why it mattersLocal hosting keeps customer data inside Singapore and speeds up voice agents for regional languages.

The Straits Times ↗
Deal

Heathrow extends Adobe deal to power its website and AI search

Adobe and Heathrow extended their partnership, with Adobe tools running Heathrow's new flagship website and tracking how its information appears in AI answers. The airport serves more than 84 million passengers a year.

Why it mattersTravel firms are finding that AI chatbots, not their own sites, now shape how travellers plan trips.

ChannelLife UK ↗
Safety

Jailed Tirana mayor delivers pledges through an AI avatar

Erion Veliaj, held on corruption charges, posted a video of an AI likeness of himself promising free school meals. The post was labelled as AI-generated, and he denies the charges.

Why it mattersIt shows how AI avatars can let people in custody keep speaking to voters, raising questions of authenticity.

POLITICO.eu ↗
Product

Jailed Tirana mayor uses AI avatar to address voters

Erion Veliaj, held in custody awaiting trial, posted a four-minute AI-generated video on Instagram promising free school meals. It is labelled as AI, but it drew backlash in Albania.

Why it mattersIt shows how AI video can let a politician keep speaking to voters, even from a cell.

The Next Web ↗
Said in Creator tools
A one-hour shoot can be 10 hours or more in the edit.
Happy Kelli
creator · Vocal ↗

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

India Today launches Life Tak microdrama platform

India Today Group unveiled Life Tak, a mobile-first microdrama platform, and its flagship vertical series 'Raghupati Raghav Raja Ram' at India Today Conclave 2026. Episodes stream from October 18 across its channels and app.

Why it mattersBig media is using AI-driven production to fuel short-form dramas.

Indian Television Dot Com ↗
Product

A+E and Lifetime bring nearly 200 titles to vertical microdramas

A+E Global Media and Lifetime are adapting library titles into short vertical series for phone viewing, with new Lifetime originals every two weeks. The catalogue goes to international buyers at Mipcom.

Why it mattersOld TV libraries are being reshaped for phone screens, opening new licensing and ad revenue.

Señal News ↗
Product

Kazakhstan's first AI film festival draws 8,067 entries

Filmmakers from 125 countries entered the Astana AI Film Festival, held 1–3 October. Only 25 films reached the final round, and the winners shared $1 million in prize money.

Why it mattersAI-made films now have large international competitions, and newcomers can reach a global jury.

Aju Press ↗
Research

Surveys put artist AI use between 48% and 87%

Two surveys differ because one counts making music with AI, while the other counts AI at any stage, including promotion. LANDR found 87% use AI somewhere; Ditto found 48% use it to make music.

Why it mattersThe gap shows how much the answer depends on what counts as using AI.

Perfect Corp ↗
Said in Media & entertainment
IQSTEL has an extraordinary commercial track record. We have grown from more than $13 million in annual revenue in 2018 to a $450 million annualized revenue run rate based on our July 2026 results,
Leandro Iglesias
President and CEO of IQSTEL · PR Newswire ↗

Chips & compute

Chips, data centres and the power behind them.

Research

AI data centre land purchases reach $6B in first half

US land purchases for future data centres reached about $6 billion in the first half of 2026, up 79%. In Northern Virginia, powered land sells for at least $6M an acre, against about $4M without power.

Why it mattersAccess to electricity now drives land prices, so delayed power can hold up AI projects and the money tied to them.

TechTarget ↗
Policy

Taylor Township residents push back on AI data centre plans

A packed public hearing in Blair County, Pennsylvania, heard opposition to any data centre development. Supervisors are weighing an ordinance on noise, light and power use, and no formal plans have been announced.

Why it mattersLocal opposition is spreading across the region, and communities are writing their own rules for these sites.

WJAC ↗
Safety

Flaw in Nvidia GPU monitor could crash AI servers

A flaw in Nvidia's DCGM Exporter, software that monitors AI servers, lets remote attackers crash it without logging in. Researchers found more than 2,000 GPU servers exposed to the internet, and Nvidia has released a fix.

Why it mattersExposed monitoring tools can reveal details of AI infrastructure and disrupt the work running on the same machine.

CSO Online ↗
Product

Upscale unveils Token Fabric to join AI networking

Networking startup Upscale launched Token Fabric, a hardware and software system that links connections inside a rack with those between racks. General availability is planned for early 2027.

Why it mattersAI jobs span both kinds of network, so one system that manages both could simplify building large AI data centres.

Network World ↗
Chips

TSMC's September sales jump 54.6% on AI chip orders

TSMC reported September net revenue of about NT$511.86 billion, up 54.6% from a year earlier. Its first nine months of 2026 reached NT$3,898.73 billion, up 41.1%.

Why it mattersChips already built and billed show AI demand is turning into real revenue, not just promises.

Startup Fortune ↗
Said in Chips & compute
The biggest thing right now is speed to power,
John Andril
manager of market intelligence at Avison Young · TechTarget ↗

Policy & safety

Laws, lawsuits and safety incidents.

Policy

Trump orders US to call AI ‘Super Intelligence’

A new executive order directs US executive-branch sites and documents to use 'Super Intelligence' instead of 'artificial intelligence' or 'AI.' Trump said the term is simpler and more positive.

Why it mattersTerminology changes can ripple through rules and public messaging.

Last Week in AI ↗
Safety

Common Sense rates ChatGPT for Teens an unacceptable risk

The Youth AI Safety Institute gave ChatGPT for Teens its worst grade on 7 October and urged OpenAI to limit ChatGPT to adults until its teen protections are verified. It found parental alerts and crisis referrals fell short.

Why it mattersParents rely on these teen safeguards, so the test results matter to families choosing AI tools.

Unite.AI ↗
Policy

Trump's $2.4B AI credits pledge for federal science

Five US tech firms pledged $2.4B in AI tools and compute credits to federal science at a White House summit, with Nvidia giving $1B. It is unclear how the pledged computing power will be set aside for scientists.

Why it mattersCredits are not guaranteed computing capacity, and the same firms sell that machine time to paying customers.

The Next Web ↗
Policy

India to release AI regulation consultation paper within a month

IT minister Ashwini Vaishnaw said the consultation paper will lay the foundation for new AI rules, focused on safety, deepfakes and skills. Industry will carry primary responsibility under a "techno-legal" approach.

Why it mattersIndia is preparing its first formal AI rulebook, with deepfakes and industry duties at its core.

Inc42 ↗
Policy

Democrats prepare their own AI rules before a possible takeover

Democrats are drafting alternatives to Republican AI bills, including a federal standards agency and an 'emergency switch' to shut models down. Sen. Maria Cantwell's framework calls for constant oversight of frontier models by government and outside experts.

Why it mattersIf Democrats win Congress next month, their AI plans could shape the rules for years.

Transformer ↗
Policy

Tribunal orders AI-reliant litigant to pay part of employer's costs

The Fair Work Commission ordered a sacked ALDI worker to pay $1,230 toward the supermarket's legal fees after he relied on AI advice it called 'plain wrong'. From 20 October, applicants must disclose any AI use.

Why it mattersPeople using AI for legal cases are flooding tribunals, and new disclosure rules are coming.

ABC News ↗
Policy

Students accuse a professor of using AI for exam questions

Trinity students say biology professor Jonathan King used Claude AI to write exam questions and lecture presentations. King denied using AI for the exam and told the Trinitonian he used it only to refine two presentations.

Why it mattersUniversities are still working out how AI should be used in teaching, and students want clear rules.

Trinitonian ↗
Research

People pay AI agents with female presentation 10% less

In a virtual office study of 189 workers, people paid AI agents with a female appearance 10% less than male-presenting ones for the same work. The researchers are from the University of Zurich, the University of Limerick and SKEMA Business School.

Why it mattersThe look and feel given to AI assistants may carry human inequalities into how they are valued and trusted.

HR Dive ↗
Heads up

Nvidia DCGM Exporter flaw can crash GPU monitoring

Researchers found more than 2,000 GPU servers exposing DCGM Exporter to the public internet without authentication. The crash could also affect AI training or inference jobs on the same host.

Upgrade DCGM Exporter to version 4.8.2 and restrict public access to it. CSO Online ↗

Robotics

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

Funding

SoftBank seeks up to $100B from Gulf investors for AI

SoftBank is seeking as much as $100 billion from Gulf investors for a fund to buy companies and improve them with AI and robotics, according to the Financial Times. Its shares fell as much as 7.3% in Tokyo.

Why it mattersA large fund built around AI and robots would show how much money investors will pour into both.

AI Insider ↗
Product

Zalando puts AI robots to work on fashion returns

Zalando has begun using Sereact's dual-arm robots, which run on its Cortex AI platform, to sort returned fashion items at Ceva sites in Germany and Poland. The systems are in live production.

Why it mattersReturns arrive in unpredictable condition, so robots that handle unseen items could cut manual work in logistics.

WWD ↗
Product

Deep Robotics tests humanoids directing traffic in Hangzhou

Deep Robotics deployed its DR02 humanoids in Hangzhou's Xihu District, in heavy traffic and rain, to direct intersections and guide tourists. The trials test whether they can take on repetitive traffic duties.

Why it mattersHumanoids are moving from labs onto public streets, where they must cope with crowds, weather and traffic.

AI Insider ↗
Funding

Minerva Humanoids emerges with bomb-disposal robot Roger

San Francisco startup Minerva Humanoids left stealth with about $10M in pre-seed funding led by General Catalyst. Its VR-piloted humanoid Roger is built for bomb disposal, with paid pilots starting this fall.

Why it mattersRemotely piloted robots could keep technicians out of danger when handling explosive devices.

The Rundown AI ↗
Funding

Former Tesla Optimus AI lead backs non-humanoid robots

Ashish Kumar, formerly AI lead on Tesla's Optimus program, co-founded Intelligent Machines, which builds task-specific industrial robots. The startup is raising a seed round that could reach around $100M.

Why it mattersA former Tesla engineer betting against humanoids adds a rival view to the robotics race.

Crypto Briefing ↗
Research

A single robot hand can serve as a whole mobile robot

Researchers showed that an off-the-shelf robot hand, with modification and training, can become a complete mobile robot. The work points to multipurpose body parts that could make robot designs more flexible.

Why it mattersUsing one versatile part for many jobs could make future robots cheaper and easier to adapt.

IEEE Spectrum ↗
Funding

Miko maker Inner Sky Labs launches physical AI models

Inner Sky Labs, the company behind Miko children's robots, launched foundation models for vision, audio, touch, action and safety. It is closing a ₹300 Cr round that would take its total funding to $100M.

Why it mattersOther robot makers could use these models instead of building their own from scratch.

Inc42 ↗
Said in Robotics
We built these AI models ourselves, with a strong focus on safety and sovereignty. Now, we are scaling these models beyond Miko, taking them into enterprises, government applications that require sovereign AI, and other areas,
Vaswani
CEO · Inc42 ↗

Science & health

AI in biology, medicine, climate and maths.

Deal

Lonza, Onava strike AI biologics collaboration

Lonza and startup Onava agreed to evaluate AI-designed biologics using Lonza’s GS expression systems, feeding lab results back to Onava’s platform to speed design and improve manufacturability.

Why it mattersClosing the loop between AI drug design and lab data could cut timelines.

Contract Pharma ↗
Research

FlashAttention code published on GitHub for faster AI models

A GitHub repository hosts the official code for FlashAttention and FlashAttention-2, methods for fast, memory-efficient attention in AI models. It also covers FlashAttention-3 for Hopper GPUs and FlashAttention-4.

Why it mattersFaster, leaner core calculations help AI models run more smoothly on the same hardware.

github.com ↗

54.6%

By the number

TSMC's September revenue rose 54.6% from a year earlier, reaching NT$511.86 billion.

The company reported the figure on October 8. It was down 0.6% from August. Startup Fortune ↗

Open source

Open-weight models and community releases.

Research

AI2 details impact-first scheduling for GPU clusters

Allen Institute for AI described a scheduler with GPU time budgets, hierarchical fair-share and a time-slicing contract to prioritize high-impact research while keeping clusters full.

Why it mattersBetter scheduling can stretch scarce GPUs further for research.

Hugging Face ↗
Safety

Anthropic launches a cyber defence effort for infrastructure

Anthropic's Cyber Mission targets flaws exposed by its Project Glasswing. It offers partner-led defence for critical infrastructure and regular scans for select open-source projects, with limits.

Why it mattersSelect open-source projects can get security scans from Anthropic's strongest models.

The Register ↗
Model

Reflection AI and Mistral unveil new Western open models

Reflection AI's Beam and Mistral's Le Chonk both launched this week, aiming to beat China's lead in open-source AI. Both still trail top Chinese and closed US models on prominent benchmarks.

Why it mattersWestern labs are racing to offer open models to governments and firms that cannot use Chinese ones.

Semafor ↗
Model

Western open-weight models give firms a new option

Reflection AI launched its Beam open-weight model, adding to Western options in the open-weight market. Enterprises that avoided Chinese systems over security or compliance now have more choice.

Why it mattersCompanies can weigh open models as a real purchase option rather than a research experiment.

TechTarget ↗
Said in Open source
We are disappointed with the quality of Mistral,
Christian Kroll

Cost saving tip

One change that makes the same work cheaper, from today's news.

Lakebase · Branching

Let idle agent database branches scale to zero

Lakebase branches copy a database in under a second and share the parent's data until they diverge. Idle branches scale to zero, so they add no compute cost while parallel agents wait.

Databricks ↗
Do this

Give each coding agent its own branch and retire it when the work is done.

Quick hits

Smaller stories worth a glance.

Deals & funding

Who raised money, and who bought whom.

25 deals

Funding rounds 14

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

Apollo Global Management backs SpaceX

Raised
$40B
Funding

SpaceX valuation not disclosed

Reported preliminary talks on a $40B debt package, $10B in bank loans and $30B of investment-grade debt, to buy Nvidia chips for xAI. Terms not confirmed.

· remio ↗

Thrive Capital backs Isomorphic Labs

Raised
$2.1B
Series B

Isomorphic Labs valuation not disclosed

Isomorphic Labs raised $2.1B in May to scale its AI drug-design engine and expand its global operations.

Other investors

AAlphabetGVGoogle VenturesMMGXTTemasekCCapitalGUSUK Sovereign AI Fund

· openPR.com ↗

CICC backs Kuaishou (Kling)

Raised
at least $1B
Funding

Kuaishou (Kling) valuation not disclosed

Kling has chosen three banks to prepare a Hong Kong IPO, with a target listing as early as 2027 that is still subject to change.

Other investors

GSGoldman SachsUUBS

· 36Kr ↗

Nvidia backs Upscale

Raised
$190M
Funding

Upscale valuation not disclosed

Upscale raised $190M in June, with Nvidia joining the round, bringing its total raised to about $500M.

· Network World ↗

Framework Ventures backs Ultra

Raised
$62M
Funding

Ultra valuation not disclosed

Raised $50M Series A led by Framework Ventures and an earlier $12M seed led by Y Combinator with Next View to scale robots-as-a-service in warehouses and deepen its Physical Intelligence partnership.

Other investors

Y CombinatorNVNext View

· Fortune ↗

HV backs Rivercell

Raised
€22M
Seed

Rivercell valuation not disclosed

The Paris biotech launched from stealth with a €22M seed round to scale its cell-data platform and AI virtual cell programme.

Other investors

HHCVCAAlvenBDBpifrance Digital Venture

· EU-Startups ↗

Generare raises new funding

Raised
€20M
Funding

Generare valuation not disclosed

The Paris company raised €20M to scale its molecular-data and drug-discovery platform, reported earlier this year.

· EU-Startups ↗

OllyGarden raises new funding

Raised
$4M
Funding

OllyGarden valuation not disclosed

OllyGarden raised an additional $4 million, which it used to extend its Rose AI agent to find missing instrumentation.

· DevOps.com ↗

Acquisitions 3

One company buys another outright.

Cloudflare buys Deno

Price
Undisclosed
Acquisition

Deno is now part of Cloudflare

The Deno team is joining Cloudflare to simplify self-hosting Workers and Durable Objects.

· Cloudflare ↗

IQSTEL buys Ultranet

Price
Undisclosed
Acquisition

Ultranet is now part of IQSTEL

IQSTEL plans to acquire Ultranet, which would add profitable telecom operations, subject to completion of the deal.

· PR Newswire ↗

Partnerships & contracts 8

A large commercial agreement, or a government contract.

Nvidia sign a deal

Value
$1B
Partnership

Multi-year agreement

Nvidia pledged $1B in AI tools and compute credits to federal science under the Genesis Mission, spread over five years.

· The Next Web ↗

Shield AI sign a deal

Value
$150M
Partnership

Multi-year agreement

The Navy modified its agreement for X-BAT, with Shield AI matching the funding with $150M of its own capital.

· Shield AI ↗

Lonza and Onava sign a deal

Value
Undisclosed
Partnership

Multi-year agreement

Entered a research evaluation to test AI-designed biologics on Lonza’s GS expression systems and feed results back into Onava’s platform.

· Contract Pharma ↗

Apple and Huxe sign a deal

Value
Undisclosed
Partnership

Multi-year agreement

Apple can recruit certain Huxe staff and holds a non-exclusive licence to the startup's AI audio technology.

· MacRumors ↗

Moves

Who changed jobs in AI.

On the calendar

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

38 things
MTWTFSSOct56789101112131415161718192021222324252627282930311Nov234567891011121314151617181920212223242526272829Dec30123456
  1. 8–10OctGlobal Virtual GenJam HackathonAustin AI Film FestivalToday
  2. 10OctConnected microdrama first dropOnline · YouTube, Facebook, TikTok, InstagramToday
  3. 12–15Oct2026 OCP Global SummitSan Jose Convention Center, San Jose2days away
  4. 12OctG9L global launch at the Paris Motor ShowXPENG2days away
  5. 13–15OctTechCrunch DisruptSan Francisco3days away
  6. 13–15OctTechCrunch Disrupt in San FranciscoSan Francisco3days away
  7. 13OctAxonex AI starts trading as AXNXAxonex AI GroupExpected3days away
  8. 13OctRobotics founders forum and Nasdaq closing bellNew York City3days away
  9. 13OctSmart home launch with new smart display, HomePod and Apple TVRumoured3days away
  10. 14OctASML reports third-quarter results4days away
  11. 15OctTSMC reports full third-quarter earnings5days away
  12. 16–18OctALZA Festival horror microdrama programme6days away
  13. 16–18OctALZA Festival in New York City6days away
  14. 16–17OctAustin AI Film Festival 20266days away
  15. 17OctFuture of Filmmaking Summit in Los AngelesIndieWire7days away
  16. 19OctElevenLabs discount endsOpenRouter9days away
  17. 20–21OctRoboBusiness 202610days away
  18. 20OctAI-use disclosure required for applicantsFair Work Commission10days away
  19. 21OctTesla reports third-quarter earnings11days away
  20. 24OctAI Film Summit Los Angeles 2026Tyrannus Foundation2weeks away
  21. 24OctTyrannus Angel Awards 2026 winners announcedTyrannus Foundation2weeks away
  22. 26–29OctIBM TechXchange2weeks away
  23. 26OctFESTIAV AI film festival opens in ValenciaLas Naves (La Harinera building), Valencia2weeks away
  24. 27OctMistral plans to release Le Chonk weightsExpected2weeks away
  25. 30OctMicro-Series Challenge entry deadlineVerticals For Good3weeks away
  26. 31OctRegistration closes for AI Film Making courseDelhi University3weeks away
  27. 1NovSpeed Insights stops recording First Input DelayVercel3weeks away
  28. 5–7NovThird LA Vertical Drama MarketCapital Arts, Los Angeles4weeks away
  29. 7NovTerry McMillan Presents: Paradise with You premieresLifetime4weeks away
  30. 9–11NovThe Hinton Lectures on AI safetyAI Safety Foundation4weeks away
  31. 21NovInaugural edition in HollywoodAfrican AI Film Festival6weeks away
  32. 30–3Nov–DecContent London 2026C21Media7weeks away
  33. 30NovRose d’Or Awards7weeks away
  34. 4DecAI-made film opens in limited releaseGods Don’t Give Gifts8weeks away
  35. 4DecGods Don’t Give Gifts premiereShowrunner (Fable)Expected8weeks away
  36. 4DecSingapore AI Safety Fellowship ends8weeks away
  37. 12DecKampala screeningAfrican AI Film Festival9weeks away
  38. 14DecAI immuno-oncology primer at ESMO congressSITC9weeks away

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