AI News Archive: October 2025 - Monthly Highlights
100 articles published this month
CrowdStrike and NVIDIA employ Nemotron models to train agents on Falcon data
CrowdStrike and NVIDIA are trying to make a smarter security guard. Their method is to train AI agents using a new set of open-source models called...
Adobe's Frame Forward AI removes subject from first frame, fills background
Video editing is a stubborn, tedious craft. Adobe’s new AI research project, Frame Forward, seems designed to gut the most painful part of it:...
Celosphere 2025: Company lifts sales order automation to 86%, saves USD 24.5M
Enterprise software vendors love to sell efficiency. They rarely deliver it. But when a company actually makes its internal machinery work, the...
Evaluating Agentic AI: Transparency, Reliability and Ethics Needed
AI is getting clever in a worrying way. The tests we use to grade it haven’t caught up. Researchers aren't just concerned.
Perplexity launches AI tool that lets users search patents with natural language
Patent searches are a special kind of hell. You need the precision of a lawyer, the vocabulary of an engineer, and the patience of a saint.
IBM and AICTE launch AI Lab in New Delhi with SkillsBuild's 1,000 AI courses
In New Delhi this week, a nondescript training lab opened its doors. Its purpose, however, is anything but small.
Sora adds paid option for extra AI videos, trimming free allowance
Sora, OpenAI's video generator, is no longer a free ride. The company just cut its daily free allowance and launched a paid option for anyone who...
NVIDIA to deploy 260,000 GPUs in South Korea, says AI needs accelerated compute
South Korea wants to buy a quarter of a million rocket boosters for its AI program.
OpenAI safeguard models outpace GPT-5-thinking and OSS versions in tests
OpenAI’s safety tests usually feel like a box-ticking exercise before a product launch. The latest results are different.
Guillermo del Toro Says He’d Prefer Death Over AI Art Going Mainstream
Guillermo del Toro would rather be dead than see AI art normalized. He said so himself.
Canva launches AI-Powered Creative OS with Design Model for editable designs
Canva has always been the tool for people who need a poster fast. Now it wants to be the operating system for how you think about design.
AI Psychosis Highlights Mental Health Pros' Lack of ChatGPT Access
Your therapist likely hasn't logged into ChatGPT. Many haven't bothered. That simple fact is becoming a clinical problem, as patients arrive with...
Meta researchers find signatures in LLM traces signal reasoning correctness
There’s a basic problem with the machines we’re told will reason. We have no real idea if they are.
xAI launches Grokipedia, an AI-driven, real-time, bias-free encyclopedia
Elon Musk thinks Wikipedia is biased. So he built a new one that runs on code. His company xAI has launched Grokipedia, an encyclopedia generated and...
Anthropic says language model checks its own activation states before responding
Anthropic's latest research suggests its language models might be doing more than just predicting the next word.
Google AI Studio lets users trace inputs, outputs and API usage in logs
Building with AI often means blindly trusting a black box. You feed it a prompt, and you get a strange, inexplicable reply.
OpenAI unveils Aardvark, an agentic researcher that hunts bugs like a human
Software bugs have long been the Achilles' heel of complex coding projects. Now, OpenAI might have found a game-changing solution with Aardvark, an...
Pinterest launches AI shopping assistant Thursday, suggesting looks
Pinterest thinks you need a robot to tell you what to wear. On Thursday, the company will launch an AI shopping assistant that suggests outfits based...
Canva’s ‘imagination era’ lets users tag AI for copy, edits, and image variants
Canva is putting AI on tap. The design platform now lets users tag an AI assistant directly in the workflow to rewrite text, tweak images, or spin up...
Anthropic's deal backs Claude on Google TPUs, Amazon Trainium, Nvidia GPUs
Anthropic is covering its bets. In one sweeping deal, the AI startup just wired its Claude model into Google Cloud, Amazon Web Services, and Nvidia's...
Perplexity offers free Comet AI browser; AI browsers cybersecurity time bomb
Perplexity is giving away a browser, and that tells you everything about the state of AI. Its new tool, Comet, is free as of early October.
Universal partners AI startup Udio on next-year platform, revenue for UMG artists
Universal Music Group sued the AI music startup Udio last year. Now they’re going into business together.
Manus 1.5 lets user build a full-stack web app in four minutes
Four minutes. That’s the new claim from the AI platform Manus: build and deploy a full-stack web app in less time than it takes to boil an egg.
Cursor 2.0 adds own coding model, runs 8 AI agents, adds voice and team commands
The coding editor is vanishing. In its place, Cursor 2.0 is building an AI operations center. Their power move?
Google plans Gemini 3 launch in 2025 as AI drives USD 100B revenue
Google’s next big AI model, Gemini 3, is scheduled for 2025. This is not a speculative roadmap.
Adani Airport Holdings partners with AionOS for multilingual AI travel service
Airports are stress factories. Language barriers turn the screws tighter. Now, the Adani Group—which operates eight of India’s largest airports,...
Google offers Jio users free Gemini 2.5 Pro access worth Rs 35,100 via MyJio app
Google just wired a strategic bribe worth Rs 35,100 straight to Jio's 470 million subscribers. This isn't charity. It's a land grab.
Tata Communications and NiCE combine AI suites for agent-first contact centres
Tata Communications is linking its customer service tech with NiCE's. The press release frames it as a major upgrade for contact centers.
Shunyalabs launches Zero STT Med, beating Whisper and AWS in medical ASR accuracy
Doctors are drowning in paperwork. The speech recognition tools thrown to them as a lifeline?
Confluent launches Real-Time Context Engine to unify stale data for AI
Most corporate AI projects are stuck. They're not failing on the whiteboard or in the lab.
ServiceNow lifts 2025 subscription revenue outlook to USD 12.8 B, 20% growth
ServiceNow just added another fifty-five million dollars to its 2025 forecast, because why not.
Choreographer partners with Google Arts & Culture Lab on AI tool AISOMA
Every new AI tool is pitched as a revolution. This one actually might be interesting.
IndicWav2Vec, Trained on 40 Indian Languages, Leads ASR Diversity
In an Indian call center, a single agent might field pleas in Tamil, requests in Bengali, and questions in Hindi—all before lunch.
Meta, Google, Microsoft boost AI spending; capex hits USD 34.9B, up USD 5B
Meta, Google, and Microsoft just cut a check for $34.9 billion. That single-quarter capital expenditure—nearly five billion more than Wall Street...
OpenAI's new moderation model swaps static classifiers for Safety Reasoner
AI content filters are famously dumb. They catch the obvious slurs and miss everything else.
Zuckerberg excited: AI will add huge corpus of content to social feeds
Mark Zuckerberg thinks your social feeds are about to become synthetic. Not just a few AI jokes sprinkled in, but a flood.
Cursor launches Composer LLM, claims 4× coding speed and automated testing
Another week, another AI model promising to save developers from their own jobs. This time it's Cursor, a startup that sells an AI coding...
Paired acoustic and semantic tokenizers preserve fidelity, enable long TTS runs
For years, text-to-speech has sounded great for a minute or two. Then it wobbles.
AI workloads projected to use >50% of data-center power by 2028, paper warns
We keep talking about AI as software. A model, a chatbot, an assistant. We should talk about it as hardware.
Generative AI's Limits in SDLC: Programmers' Intuition Still Essential
Every new tool that promises to make software write itself hits the same wall. Generative AI is no different.
NotebookLM adds full 1 million-token Gemini context window, boosts processing
Google's AI research continues to push boundaries with a significant upgrade to NotebookLM.
Anthropic scientists test whether Claude can tell injected thoughts from text
Researchers at Anthropic just pulled off a strange kind of brain surgery. Their patient was Claude, their own AI. The procedure?
TikTok's Smart Split AI makes short, captioned vertical clips from chosen parts
TikTok’s newest feature is a machine that eats your long, boring videos and spits out ready-to-post clips. It’s called Smart Split AI.
Deloitte forecasts 60%+ of large firms will scale agentic AI by 2026
Corporate AI is finally getting a job description. The consultants at Deloitte now predict that within two years, over 60% of large companies will...
Accenture launches Physical AI Orchestrator software-defined factories at NVI
Walk any factory floor today. You'll hear plenty of talk about artificial intelligence.
Grammarly Rebrands as Superhuman, Retains Familiar Sidebar UI
Grammarly thinks its brand is a problem. The name screams grammar checks, red squiggles, school. It doesn't say artificial intelligence.
Confluent and Redpanda race to build agent-ready streaming data infrastructure
Two companies are fighting over who gets to be the plumbing for the AI agent boom. It's Confluent versus Redpanda.
Adobe's Corrective AI changes voice-over emotions, swaps music with Adobe Stock
Audio is the bane of video editors. Adobe's latest AI project is an attempt to fix that, promising to swap the music and tweak the emotion in a...
AI for Math Initiative finds structures showing problems harder for computers
Computers hate hard math. We knew that. But a Google research project just found a new way for them to hate it.
Geostar leads GEO shift as AI chatbots cut traditional SEO 25%, Gartner reports
You can stop paying for half your SEO team. Gartner says AI chatbots have already gutted traditional search traffic by 25%.
Microsoft Agent Lightning uses reinforcement learning to automate AI agent tuning
Training an AI agent is a slog. You tweak a parameter, run a test, and wait. The process eats weeks.
Mphasis launches quantum-safe AI tools for enterprise security and decisions
Mphasis has launched a pair of tools that aim to solve a problem which, for most companies, doesn't exist yet.
gpt-oss-safeguard lets developers apply custom policies via model reasoning
Most AI safety tools are blunt instruments. They check for slurs, maybe some violence, and call it a day.
Cerebras Wafer-Scale Engine 3: 5 nm, 4 trillion transistors, largest AI chip
Everyone is chasing Nvidia. Cerebras, a persistent underdog, just threw another giant silicon rock into the pond.
ChatGPT Enterprise boosts knowledge preservation in ICT R&D, says Yohei Ishida
Dai Nippon Printing had a memory problem. Its researchers kept walking out the door, taking years of expensive technical knowledge with them.
Microsoft's Copilot adds low-code app builder, fulfilling nine-year effort
Microsoft has spent nine years trying to get office workers to build software. That bet just landed inside Copilot.
Fortanix, NVIDIA unveil attestation-gated AI security for regulated industries
Security tech for AI often promises more than it delivers, especially for the businesses that need it most.
Adobe launches Project Moonlight, AI assistant for social media admin
Adobe is building another assistant. This one is for the social media manager who has six apps open and three deadlines missed.
PayPal becomes first digital wallet on ChatGPT, enabling payments next year
PayPal is now the first payment method inside ChatGPT. You will be able to use it next year.
Linear MCP Gives AI Premium Access to Manage Projects via Natural Language
Most project management tools are digital naggers. You click through menus forever.
11 AI startups from Brazil, Chile, Colombia, Mexico chosen to enhance safety
Latin American tech has a security problem, and its governments have decided the answer is more Latin American tech.
Gemini for Home voice assistant: 100 commands to play music, podcasts, TV
Google announced one hundred new voice commands for its Gemini home assistant. It is an attempt to make it useful.
Doppel AI defense halts automated attacks faster than human analysis
A human security analyst sees a potential threat, studies it, and maybe, hours later, raises an alarm.
GitHub launches Agent HQ at Universe 2025, centralizing AI coding agents
Every tech company now sells an AI coding assistant. Developers are drowning in them. GitHub, owned by Microsoft, wants to be the life raft.
Law and tax firm cuts drafting time to minutes with ChatGPT Business
Christoph Korts, a partner at a German law and tax firm, bills by the hour but gets paid for the result.
Microsoft retains 32.5% stake in OpenAI as partnership agreement continues
Microsoft still owns almost a third of OpenAI. That fact is more important than any of the boardroom drama that dominated the headlines last year.
NxtGen’s M for Coding becomes AI autopilot, funded by Rs 10,300 cr IndiaAI Mission
India just wrote a $1.2 billion check for its AI independence. The national IndiaAI Mission commits Rs 10,300 crore to domestic infrastructure.
OpenAI hires Solutions Architect in Bengaluru to support early startups
One job listing in Bengaluru. That’s it. For all the talk of a corporate invasion, OpenAI’s first move into India is a single opening for a Solutions...
Opera AI Browser: Mature Mainstream Choice, Needs Recent Updates
Opera's browser now comes with an AI assistant pre-installed. This is not a radical act. It is, however, a practical one, which makes it unusual.
OpenAI makes ChatGPT Go free for all Indian users for a year, Nick Turley says
OpenAI just made its move in India. ChatGPT Go is now free for a year. Sign up, use it, and don't think about a bill until 2026.
DOE and AMD pledge USD 1 billion for two AMD-chip AI supercomputers at Oak Ridge
The government just handed AMD a billion-dollar vote of confidence. The Department of Energy is putting up $1 billion with the chipmaker for two new...
OpenAI's ChatGPT Atlas Browser Mirrors Chrome, Faces Perplexity Comet
OpenAI now makes a browser. It looks a lot like other browsers. The company launched ChatGPT Atlas for macOS, a navigation tool that immediately...
SAP Concur warns AI-generated receipts spark surge in expense fraud
Corporate finance teams face a new enemy, and it's the same tool the marketing department loves.
MiniMax-M2 leads benchmarks in agentic tool calling and coding workflows
Everyone wants an AI model that can actually do things. Not just chat. Not just list ideas. But actually write, test, and execute code.
OpenAI says 0.07% of ChatGPT users show possible mania or psychosis signs weekly
Forget the flashy demos. OpenAI’s latest research report drills into a darker reality of human-AI interaction.
LexisNexis CEO: AI now core legal infrastructure, like email or word processor
Legal tech’s big AI shift is no longer a shift. It’s a done deal. The software lawyers use every day is now built on it, baked into the core like...
Google Cloud offers managed Slurm to rival CoreWeave and AWS in AI training
Google's cloud division just made a standard move for a company its size: it copied a competitor's idea.
Google's Jules AI tackles routine code upkeep while developers design
Software maintenance is a tax on every developer’s time, and Google thinks its new AI can finally automate the payment. The tool is called Jules.
Fitbit's health coach preview gives timely updates at wake-up and after a workout
Fitbit's AI health coach hit public preview this week. Don't call it a revolution. It's a wrist-based nudge system, optimized for timing.
Anthropic embeds Claude AI in Excel, eyes finance market against Copilot
Every AI company wants to sell to banks. Anthropic, tired of waiting, just parked its truck in Microsoft’s driveway.
Sora exposes failures in AI labeling, including OpenAI-overseen C2PA
A woolly mammoth trudges through snow, generated by OpenAI's Sora. That single video didn't just bypass Content Credentials. It shattered them.
Meta AI app sells vacation photos as escapism, says designer Laurent Del Rey
Your vacation photos are bad and you are sad. Meta has a product for that. An experimental app called Endless Summer, cooked up by designer Laurent...
AIjacking Threat Grows as Prompt Injection Tricks Agents in Customer Ops
Companies are handing customer service over to bots. The bots are gullible. A new security threat makes this a serious problem.
OpenAI shares 50 ChatGPT prompts in new ‘Chats for College Students in India’
ChatGPT isn't just for essays anymore. In India, it's getting plugged into the grind of college life.
MiniMax-M2 Beats GLM 4.6, Offers Compact, High-Efficiency Multi-Step Reasoning
Most open-source AI models are either huge and capable, or small and weak. The new MiniMax-M2 is an attempt to break that trade-off.
India’s AI push promises more creative work, but workers must reinvent
Every Indian office is slowly being injected with code that thinks. The result isn't a mass firing, at least not yet.
Axis Communications presents AI-powered IP cameras for business intelligence
Axis Communications sells security cameras. They don't just watch. Mounted in a store ceiling, the latest IP model can analyze foot traffic, quietly...
Students Aim to Translate Brain Science for Communities, Sparking Tech Interest
In Ann Arbor, Michigan, a neuroscience graduate student is diagramming synapses on a community center whiteboard.
Mark Solms says AI algorithm may tap the evolutionary source of consciousness
The search for consciousness has hit a wall. We have theories, scans, and philosophical debates, but no clear picture of how awareness actually...
Mphasis launches NeoIP, AI platform for continuous enterprise transformation
Enterprise technology vendors have settled on a new anxiety to sell against: the fear of standing still.
ChatGPT, Gemini, DeepSeek, and Grok Cite Sanctioned Russian Propaganda
Your chatbot might be a Kremlin asset. Or at least a useful idiot. A new report finds that when you ask AI about the war in Ukraine, it often...
AI Videos Fuel Influencer Drama, Raising Legal Threats Over Facial Likeness
Influencers are now fighting with video fakes of themselves. The legal threats are flying because no one knows who owns a face in the age of AI.
AI agents browsing exposes fragility of human-first web assumptions
The internet is being retooled for machines, and your browser is first on the chopping block.
Google Nest Cam 2K Review: AI-powered Gemini cuts spammy alerts
Smart home security cameras have always been liars. They panic over falling leaves, document the secret lives of squirrels, and treat every shift of...
AI 'Slop' Era Lets Agents Create Listings in Minutes, Says Alok Gupta
Scroll through any listing site now. The sameness is palpable, a cheaper feel permeating the digital curb appeal. This is no accident.
AI trained on two books mimics famous authors, beats human imitators
Software can now fake a famous author's voice. That’s the claim. But researchers at Stony Brook University and Columbia Law School put it to a hard...
AI browsers turn into security threats: Four ways they worsen risk
AI browsers are selling you a robot butler. What they're delivering is a master key for hackers. The pitch is all convenience.
Study: AI chatbots cite less-known sites, unlike Google search results
When you ask a chatbot a question, you aren't getting Google's results. You're getting something else, something pulled from a darker, stranger...
OpenAI markets ChatGPT as work data search tool, but citation reliability remains distant
OpenAI wants you to use ChatGPT to find your next company expense report or a lost client memo.
Southeast Asia boosts AI momentum with regional skill training
Talk of an AI workforce is cheap. Building one is a logistical nightmare. Now, Southeast Asia is attempting it.