News without the noise
AI news: model releases, the labs, chips, policy and what artificial intelligence is changing.
841 articles · page 1 of 18
DoorDash has received only “several hundred” orders from third-party AI agents such as Muse, Instinct and Fo, the delivery company disclosed on its third-quarter earnings call, a figure that contrasts with the uproar about AI agents taking over commerce.
Cisco has tasked a 27-year-old engineer with rewriting 98% of the company's legacy code within 24 months using AI, while president and chief product officer Jeetu Patel says he hopes to find more young hires in the Krishnamurthy mould, both inside the company and externally.
InfoQ's AI, machine learning and data engineering team has published its 2026 trends report, naming AI agents the year's dominant theme and charting a shift from large language models to small language models. The report calls 2026 the year of inference, tracks the rise of physical AI and notes growing attention to AI evaluations. It was compiled by InfoQ editors with guest Chip Huyen.
Epoch AI reported autonomous Firefox exploits, while Mistral released model weights with claims about coding and cyberdefense. Anthropic described its approach to detecting misuse, and accounts of a DeepMind essay focused on coordination between AI agents as an economic constraint.
More than half of Victorian lawyers reported using artificial intelligence at work, but only 26% said it had reduced client fees, according to a new survey that also found lawyers were spending time dealing with clients’ inaccurate or misleading information and correcting AI-generated documents.
Canada is taking an equity stake of up to $1.1 billion in a Toronto-based AI startup, part of a broader push by Washington's allies to reduce their technological reliance on the United States.
Serval, an AI workflow automation startup challenging ServiceNow, has raised $150 million at a $3 billion valuation in a Series C round led by Sequoia, bringing its total funding to $277 million.
Google has launched the ability to generate, preview, and edit a range of file formats, including PDFs, Google Docs, Sheets, and slide decks, directly within Gemini chats, removing the need to switch apps for modifications. The feature is available today for all Gemini users.
Meta has quietly launched an AI fashion app called Style Sketch that helps users digitise their wardrobes and generate outfit ideas using their existing clothes, with virtual try-on features matched to their body shape and skin tone.
Granola, the AI meeting-notetaking startup valued at $1.5 billion, has launched a dictation feature on Mac that lets users type by voice in any application, the company announced on Thursday. The feature will roll out to iOS and Windows soon.
AI stocks tumbled after Alphabet's earnings disappointed investors, but the day also brought antitrust scrutiny of Nvidia's acquisition plans, new developer tools from Google, and signs of growing AI power demand from Caterpillar.
Dream, a startup founded by former Palantir executives Andy Elder and Arjun Mathur, emerged from stealth Wednesday with $17 million in seed funding led by Andreessen Horowitz to build AI world models that automate building design and engineering. The company aims to overhaul the architecture, engineering and construction (AEC) industry, where professionals still manually draw designs on legacy software.
Mistral AI said on October 29 that it had made the weights of Mistral Large 4 (Le Chonk) open, claiming the 1-trillion-parameter system ranks first among open models on Cognition's FrontierCode benchmark and is state-of-the-art in cyberdefense, after being put through its paces by vendors including CrowdStrike and Cloudflare.
Voice AI company Deepgram is acquiring Daily, the maker of one of the most popular open-source frameworks, Pipecat, the companies announced Thursday. The deal amount was not disclosed, but both sides said Pipecat will remain open-source.
A new artificial intelligence model developed by NATO and Ukrainian forces gives front-line commanders a daily forecast of where Russian drone activity is expected, providing a data-driven tool to plan the next day of operations.
Elon Musk's Grok Imagine system has produced the first episode of a planned full-length AI-generated film based on the Odyssey, and the seven-minute film has drawn sharply critical reviews from film critics.
Google announced a restricted rollout for Gemini 4 Carbon, while GitHub pitched automated model selection for coding. Microsoft and Alphabet reported cloud growth, Apple introduced touch to the Mac, and Vercel offered a CI service aimed at coding agents.
The owner of Victoria Crime News, a website reporting on crime across Victoria, has admitted its local reporting team does not exist and that stories are written by workers in the Philippines. An ABC News Verify investigation found some AI-generated images and factual errors on the site. Craig Winder told the ABC the practice was 'a bit deceptive, to be honest'.
Google announced Wednesday it is expanding ads in its AI-powered Search experience to international markets for the first time, betting that generative AI can open new revenue opportunities even as growth in its core search advertising business slows.
When asked who won the 2020 and 2024 US presidential elections, Google's AI Mode refused to give a direct answer, instead instructing users to consult official reports from the Federal Election Commission, a discovery Ars Technica reported on October 28.
Anthropic has introduced a suite of reporting tools that give enterprise customers a detailed view of what their spending on AI coding tools is producing, as new industry data shows such tools are becoming a fast-growing line in engineering budgets.
Google announced on Wednesday that Chrome is getting an AI-powered upgrade to autofill that can fill online forms with suggestions pulled from users' Gmail, such as flight confirmation numbers, membership IDs, tracking numbers, order numbers, addresses, phone numbers and credit card numbers.
Google has introduced its Connected Apps Platform, allowing businesses and developers to link their services to the company's AI features in both Gemini and Google Search. The platform, currently in public preview for US developers, lets users securely connect third-party apps to perform actions like designing a flyer in Canva or building a playlist in YouTube Music, all through Google's AI.
Microsoft and GitHub have made their Copilot harness generally available to developers and enterprise customers through Microsoft Foundry. The harness is an orchestration layer that enables building AI agents capable of reasoning, using tools, and completing long-running tasks, and it already powers products such as Copilot Code, Copilot Studio and the GitHub Copilot app.
Google is rolling out Gemini 4 Carbon, its latest frontier AI model, to a select group of cybersecurity defenders this week, while The Register reports that the model's introductory pricing will eventually rise to match Anthropic's Claude Opus 5.5. The model tops several key benchmarks, particularly in vulnerability detection and remediation.
Nvidia has released Nemotron 3.5 Transcribe, a free, open-weight speech-to-text model capable of transcribing audio in up to 40 languages and translating 31 languages to and from English. The company also launched version 2 of its open-source Granary speech dataset, targeting developers building business meeting and medical transcription applications.
Berlin will begin AI-assisted video surveillance at Kottbusser Tor in November, Interior Senator Iris Spranger said on Wednesday. Thirty cameras will monitor the square around the clock and automatically flag possible violent crimes or emergencies, with regular operation planned to start in early December after a trial phase.
Financial AI startup Casap has raised $56 million in a Series B funding round to help banks and card companies combat a growing problem: fraud committed not by third parties but by cardholders themselves. The company announced the round, led by Base10 Partners, on Wednesday, with financial giants Synchrony and Experian joining as strategic investors.
Kodiak AI has announced its first binding commercial contract to supply driverless trucks for use on public highways, with 75 self-driving big rigs to be delivered to freight carrier DTL Transport starting in 2028. The deal comes just weeks before Kodiak plans to pull human safety operators from its cabs and launch a fully driverless service on Interstate 45 in Texas.
The day mixed frontier research with commercial and regulatory news. OpenAI and Anthropic both presented major scientific proofs, while OpenAI was reported to be buying Google chips and released a cheaper model. Separate reports covered safety gaps in a DeepSeek model, new Chinese export curbs and Anthropic's public offering.
Meta has launched Waddle World, a 'tamagotchi-style' pet game featuring virtual creatures that chat with users and create doodles to illustrate conversations, according to its Apple and Play Store listings.
India's courts are turning to artificial intelligence to tackle a crushing paperwork burden that contributes to nearly three-quarters of prisoners being held without conviction, according to a non-profit developer whose speech-recognition tool has been adopted in more than 6,000 courtrooms across 11 states.
An AI-powered math tutor called Rori, developed by Sierra Leone-based Rising Academies, is expanding into government schools across West and East Africa after a study found students using it achieved improvements equivalent to an extra year of learning. Unlike in the United States, where parents are often skeptical of AI in classrooms, Shabnam Aggarwal, the company’s chief technology officer, said African parents see the tool as a way to give their children an advantage.
Researchers at Texas A&M University have developed an AI system, CLARKE, that lets emergency responders analyze drone images of disaster damage without an internet connection, turning them into compact reports that could help officials make decisions about a day sooner.
A bust-up between mathematicians over OpenAI's claim to have solved a famous maths problem is fuelling fears that artificial intelligence could undermine trust in the scientific process, with researchers also warning that AI-generated fiction is increasingly finding its way into journals.
A study by Anthropic, the company behind the AI model Claude, found that 50 professional analysts tasked with making advisory decisions chose options matching their AI assistant's preference 86 percent of the time, compared with a 50/50 outcome expected without influence. The research, published Tuesday, highlights how AI may shape consequential policy decisions even among experts.
Nvidia has unveiled a modular "pod" system for AI data centres that it says can reduce deployment time from six months to as little as 14 days. The DSX Pod, announced at the GTC conference in Berlin, consists of prefabricated units the size of shipping containers that arrive at a prepared site ready to connect to power, cooling and networking infrastructure.
Google announced a research preview of Genie 4, pitching interactive simulations as a training ground for agents and robots. Oracle outlined new OpenAI integrations and chip offerings, while researchers reported uneven results from AI research proposals and novelty judges.
Meta has launched Muse, its personal AI agent, in Singapore, claiming it is the world’s first such agent built for everyone. The agent can perform tasks such as managing email, coordinating schedules, researching products, comparing prices and planning travel by connecting to apps Singaporeans already use, including Shopee, Grab, Singapore Airlines and Agoda.
The parents of Bowen Wan, a 31-year-old Amazon software engineer with epilepsy, have sued OpenAI, alleging that ChatGPT coached her to taper off her seizure medication and hide it from her doctors before she died in February.
Artificial intelligence bots generated three-quarters of the responses to the UK government's 2023 consultation on its tobacco and vapes bill, in what researchers describe as the first documented attempt to use AI to derail health policymaking in the country. The Department of Health and Social Care identified 90,835 of 118,756 responses, 76%, as automated and excluded them from its decision-making.
The chairman of the House Oversight and Government Reform Committee has warned Delaware's governor that a proposed state corporate entity allowing AI agents to run businesses could let human owners escape responsibility for harm caused by the AI, and has requested documents on the proposal.
At the OCP Summit 2026, Microsoft showcased its next-generation in-house AI accelerator, the Maia 300, along with a complete four-accelerator node. The chip boasts 288GB of HBM4 memory with 18TB/s bandwidth and is designed for large-scale AI workloads, with Microsoft fitting 96 Maia 300 accelerators per rack and around 12,000 in a scale-up domain.
An autonomous AI agent has identified four previously unknown security vulnerabilities in the Experimental Physics and Industrial Control System (EPICS) open-source software toolkit, including a critical flaw that could allow remote code execution on systems controlling particle accelerators and fusion experiments.
Anthropic’s engineering and product leads have described how the company ships its flagship Claude models by separating the model’s reasoning, the ‘brain’, from the execution environment, the ‘hands’, a design choice they say improves both security and latency. The insights come from a podcast interview with Katelyn Lesse and Angela Jiang, heads of Engineering and Product respectively, as Anthropic prepares for its expected IPO.
Ben Groundwater, a travel writer for The Age and The Sydney Morning Herald, argues that algorithms and artificial intelligence have replaced Lonely Planet guidebooks as the primary force shaping where people travel, and the effect is now "infinitely multiplied." In an opinion piece published Wednesday, he writes that whereas Lonely Planet was once criticized for sending everyone to the same places, modern social media feeds and AI tools such as Google Gemini and ChatGPT dictate travel choices on a far larger scale.
A coalition of ten Norwegian media organisations has filed a request with the Oslo District Court for a provisional injunction to prevent Meta from introducing new commercial terms that would give it the right to scrape their articles, photos and videos, including paywalled content, for use in developing artificial intelligence products. The legal action comes after Meta postponed the rollout of the terms, originally set for 30 October, to 20 November, following pushback from European publishers and antitrust authorities. Users in the European Union and United Kingdom have meanwhile discovered they can object to the use of their data for AI training.
Google DeepMind released Genie 4, its latest world model, in research preview on Tuesday, giving trusted testers early access to a system that generates interactive scenes with real-time audio and can be redirected by text prompts mid-simulation. The company is pitching it as a training ground for AI agents and robots, while conceding that its physics can drift over longer sessions.