AI news for 14 October 2026
Wednesday 14 October 2026 · Latentown
On Wednesday 14 October 2026, Latentown tracked 9 posts from 4 AI labs, 6 research papers and 45 news stories from 25 outlets. OpenAI was the busiest lab with 5 posts. The highest-ranked story was "OpenAI, Google pushing new AI models easily tricked into making nudes, researchers say".
Top stories
- OpenAI, Google pushing new AI models easily tricked into making nudes, researchers say
Attack method exposes flaws in OpenAI, Google safeguards against AI nudes.
- Timestral-123M
- Amp puts AI agents to work beyond code with OpenAI models
See how Amp pushes the frontier of what agents can do with GPT-6 Astra, Ultrafast, GPT-Live-1, and Sign in with ChatGPT.
- SentinelOne cuts malware analysis from weeks to hours
With OpenAI Daybreak, SentinelOne cuts weeks of malware analysis to hours, creating capacity to investigate more threats.
- China chip exports jump 136% to US$45b as AI frenzy jacks up prices
China’s integrated circuit (IC) exports surged 136 per cent year on year in September to US$45 billion, exceeding the 108 per cent growth rate in the first…
From the AI labs
Anthropic
- Anthropic opens Singapore office and signs MOU with the Government of Singapore
Earlier: Karnataka govt signs MoU with ElevenLabs for Voice AI services YourStory · 6 October 2026
- Preparing for the 2026 US midterms
OpenAI
- Amp puts AI agents to work beyond code with OpenAI models
See how Amp pushes the frontier of what agents can do with GPT-6 Astra, Ultrafast, GPT-Live-1, and Sign in with ChatGPT.
- SentinelOne cuts malware analysis from weeks to hours
With OpenAI Daybreak, SentinelOne cuts weeks of malware analysis to hours, creating capacity to investigate more threats.
- The EU’s next chapter in AI is agentic
New analysis shows how agentic AI adoption is growing across EU enterprises and small businesses—and where deeper use could unlock more value.
- How Block is building AI-powered security with OpenAI technology
Block is building AI-powered security with OpenAI. Its security team investigates threats faster and proposes patches in minutes.
- OpenAI Python SDK 3.31.0 released
A new release of the OpenAI Python SDK. Changes include: add browser and computer toolsets.
Google DeepMind
- Google DeepMind and Singapore Government collaborate to build fine-tuned AI model for scam detection
Learn how Google DeepMind and GovTech Singapore uses lightweight on-device Gemma models to intercept conversational scams in private chats with zero cloud data egress.
Mistral
Research papers
- Waypoint-OPD: Transferring RL-Induced Policy Shifts via Gated On-Policy Distillation
Policy-shift distillation reuses the difference between a teacher's log probabilities before and after reinforcement learning (RL) as supervision on student-generated trajectories.
Earlier: The Teacher Is a Direction, Not a Destination: Extrapolating RL-Induced Representation Residuals in On-Policy Distillation Town square · 28 September 2026
- OpenGCUA: Toward a Fully Open Stack for Generalist Computer Use
Computer use has reached a turning point: the focus is expanding from GUI-only interaction to systems that complete tasks on users' computers through visual and programmatic interfaces.
- FDE-Gym: The Last Mile of AI Delivery
Existing agent benchmarks typically evaluate agents on well-specified tasks defined by issue reports, specifications, or task tickets.
- MJev-Spatial: System One Spatial Intelligence via Native Attention Matching
Spatial intelligence lets an agent perceive and act in the physical world, and many of its judgments must be made at interaction rates: which spatial relation holds, how far an object is, or where…
- The Design Space of Tri-Modal Masked Diffusion Models
Discrete diffusion models have emerged as strong alternatives to autoregressive language models, with recent work initializing and finetuning a base unimodal model for bi-modal generation. Diverging…
- Code-Switching Detection via Language-Selective Dimensions in Speech Foundation Models
Code-switching (CS), the alternation between languages within an utterance, is common in multilingual speech but underrepresented in pretraining corpora. Increasing CS coverage requires finding…
From the news wire
- OpenAI, Google pushing new AI models easily tricked into making nudes, researchers say
Attack method exposes flaws in OpenAI, Google safeguards against AI nudes.
- China chip exports jump 136% to US$45b as AI frenzy jacks up prices
China’s integrated circuit (IC) exports surged 136 per cent year on year in September to US$45 billion, exceeding the 108 per cent growth rate in the first…
- Hello Inc.'s Robotaxi Unit Releases HelloWorld, a World Model That Builds Driving Scenes on Demand
HelloWorld, a generative driving world model from Hello Inc.'s robotaxi team, creates multi-view camera video and lidar data from edited layouts or…
- Nvidia's Jensen Huang pledges $2 billion for Europe, $250 million in compute for UK safety labs
Nvidia CEO Jensen Huang announced $2 billion in European investments and new stakes in French startups like Doctolib and Entalpic.
- The UK government's £500M sovereign VC fund invests in UK security startup Augur
The UK government’s £500m VC fund investing in domestic AI startups today said it was investing in a British startup which leverages AI technology to update…
- Singapore wants Google’s AI to catch scammers before victims hit send
Singapore’s war on scams has mostly been fought far from the victim: in police operations centres, bank back offices and, more recently, in rulebooks telling…
- Anthropic urges court in Washington, D.C. to rehear appeal of Pentagon blacklisting
Anthropic is asking a federal court in Washington, D.C. to rehear its appeal in a case concerning the Pentagon's blacklisting of the company's AI models.
Earlier: DC appeals court sides with Pentagon on blacklist of Anthropic The Hill · 25 September 2026; Pentagon stops using Anthropic AI tools after blacklisting company, BBC told BBC · 5 October 2026
- Salience Labs secures $125M to scale photonic chips for AI data centres
Salience Labs, adeveloper of photonic chips for AI data centre connectivity, has raised $125million in an oversubscribed Series B funding round to accelerate…
- Spill Games Raises $10.5 Mn To Scale AI-Led Mobile Gaming Studio
Mobile gaming studio Spill Games has raised $10.5 Mn (about ₹100 Cr) in a Series A round led by Elevation…
- Agentic AI Startup RevRag.AI Nets $2.5 Mn To Build AI Relationship Managers For Lenders And Insurers
Agentic AI startup RevRag.AI has raised $2.5 Mn (about ₹24 Cr) in a seed funding round led by Accion Ventures…
- Apple researchers turn idle Macs into a virtual GPU cloud with ‘Minions’
Apple researchers describe Minions, an internal system that runs AI inference jobs on idle Apple Silicon Macs when GPUs are scarce.
Earlier: Minions: A Dynamic and Distributed Virtual GPU Cloud Built on Idle Apple Silicon Devices Town square · 12 October 2026
- Google Releases Gemini 4 Argon API With 1M Token Output Limit
Google has made Gemini 4 Argon generally available on the Gemini API, AI Studio and the Gemini app.
Earlier: Gemini 4 Argon for Robotics Google DeepMind · 13 October 2026; Gemini 4 Argon: our next era of frontier intelligence Google DeepMind · 30 September 2026
- Can AI cut the US$170 billion cost of buying and selling homes?
In 2024, Americans spent about US$170 billion on real estate commissions and other costs involved in transferring ownership of homes.
- Senators press Trump admin to rally federal government around AI safety, security
The letter argues federal agencies should more rigorously test AI models for security threats and sandbox evasion, set new safety standards and lean into…
- The Blind Spots of Agentic Pentesting and How to Close Them
AI-powered penetration testing can uncover complex attack paths, but safety restrictions and network limitations leave many assets and vulnerabilities untested.
- AI guardrails designed to stop hackers are now blocking defenders, and Orion-1 aims to fill the gap
Artemis Security's new Orion-1 model is purpose-built for cyber defense and beats Claude Opus 5.5 and GPT-6 Sol on the company's own benchmark.
- UK-based AI infrastructure firm Gorilla Technology Group to enter US market with proposed 6MW data center acquisition
Site has the potential to expand to 36MW
- Aether Infrastructure buys old gearbox factory in Strasbourg, France, for AI data center
Work could start this month
- STAT+: Lobbyists and trade groups fight to shape how states regulate medical AI
Lobbying groups for doctors and hospitals and trade groups are working behind the scenes on how to regulate AI in medicine.
- Nandan Nilekani suggests last-minute AI-generated exam papers to prevent leaks
Nandan Nilekani has proposed generating examination papers at the last minute using AI to reduce incentives for leaks.
- Sunil Mittal Warns AI Is Outpacing Regulation
Sunil Mittal says governments are struggling to keep pace with the rapid pace of AI innovation, and suggests that regulation should provide guardrails without…
- AI Regulation Needed Now Before Accidents Occur, Mittal Says
Indian tycoon Sunil Bharti Mittal called for immediate regulatory frameworks for artificial intelligence, warning that governments shouldn’t wait for a serious…
- Epic’s new games launcher is ‘vibe coded’
Epic Games is announcing a big refresh of its PC games launcher today in preview, saying that the new version is "way faster" and now supports Linux.
- Uvi raises $7.5M to help AI agents learn business context and run at eight times lower cost
Agentic artificial intelligence startup Uvi said today it has raised $7.5 million in seed funding to tackle an expensive problem around business context that’s…
- Evos bags $9M to build autonomous workers that learn from companies’ most experienced humans
A startup called Evos said today it has raised $9 million in pre-seed funding to spearhead the adoption of artificial intelligence-based automation within…
- Liquid AI Releases LFM2.5-24B-A2B: A 2B-Active MoE for On-Device Agents
Liquid AI released LFM2.5-24B-A2B, an open-weight 24B Mixture-of-Experts reasoning model with 2B active parameters.
Earlier: Run fast, private agents on-device with LFM2.5-24B-A2B Hugging Face · 13 October 2026
- Dobot Launches VEGA 7, a Seven-Axis Force-Controlled Arm for Embodied AI With 50,000-Hour MTBF
Dobot's VEGA 7 has torque sensors and brakes in every joint, 0.15N force control accuracy, a 7kg payload and 725mm reach, aimed at wheeled dual-arm robots and…
- PHANES AI Launches SeePhys Hub, a Wearable Kit That Records Vision, Touch and Full-Body Motion
PHANES AI's SeePhys Hub pairs a four-camera headband with wrist, waist and leg trackers and tactile gloves, capturing human motion and touch data for training…
- Ark's Wood Favors Nvidia, AMD in 'Volatile' Chip Cycle
Cathie Wood, founder, CEO and CIO of Ark Investment Management, discusses why she avoids traditional memory-chip players, saying the business cycle is "most…
- ASML Raises Full-Year Sales Outlook
ASML upgraded the low end of its 2026 sales forecast as increasing artificial intelligence spending boosts demand for the Dutch company’s chip-making machines.
- Affinity Labs raises $2M to build AI-powered software brands with creators
London-basedstartup Affinity Labs has raised $2 million in pre-seed funding to expand itsportfolio of creator-led software brands and further develop its…
- Anthropic warns small AI models could self-replicate through IoT devices
"Small weight-based replicators" could usher in a new generation of computer worms.
- China approves Alibaba AI models for use in iPhone Duo and other Apple handsets
China has granted approval for Alibaba Group Holding’s artificial intelligence models to be integrated into several Apple handsets, including the new iPhone…
- Anthropic’s Singapore MOU bets on safety testing, not consumer reach
Anthropic has finally given its Singapore customers somewhere to knock.
- Men Can’t Stop Faking How Much They Know About AI
More than half of the men surveyed in a new Harvard study claimed to recognize AI leaders who were, in fact, totally fictional.
- FBI arrests Saudi OpenAI employee who allegedly planted recording device in conference room
The Tuesday incident comes as the ChatGPT-maker, along with competitors, faces heightened scrutiny over its safety practices.
- Floqer’s AI agents research sales prospects so teams don’t have to
Floqer is an autonomous customer knowledge platform that provides business-to-business sales and marketing teams with an AI-powered database designed to…
- Character.ai Bankruptcy Is First AI Bubble Domino to Fall
New music genre just dropped.
Earlier: AI companion startup Character.ai files for Chapter 11 bankruptcy, citing costs and litigation CNBC · 13 October 2026
- Timken Precision Drives puts the motion challenge first in humanoid robots
Timken combines the expertise of Cone Drive, SPINEA, CGI Motion, and Timken Encoders to develop application-specific motion systems.
- Doubao Phone Assistant Opens Screen Automation to Third-Party Apps as 151 Apps Opt Out
ByteDance's Doubao phone assistant ended the 30-day notice period for its SAEP protocol and will gradually allow AI screen automation in third-party apps that…
- 'Spray and pray': How AI job applications can backfire
Graduates are increasingly using AI to misrepresent themselves in job applications, research suggests.
- Centre pushes child rights body to fast-track AI tool to detect child sexual abuse material
NCPCR is developing an AI tool to detect child sexual abuse material online and preparing social media guidelines for technology companies.
Earlier: Meta deploys AI tools to detect child abuse ads, says it actioned 5.3 million content items in India MediaNama · 8 October 2026
- Parallel Loop nets US$6.6M to sell enterprises the one thing AI models lack: context
In early 2018, Yujun Chean and Prateek Jogani were selling car insurance with a cashback twist.
- TechSparks 2026: Here’s all that’s happening on Day 2
Day two of TechSparks 2026 brings Vibecode30, agentic AI, deeptech capital and more, with speakers from Stripe, Swiggy and Peak XV.
- China’s AI research is booming. Where is it headed?
Every story links to its original source. Ranking is by automatic rules, not an editor. Something wrong? Tell us.