OpenAI just reportedly signed a staggering $300 billion computing deal with Oracle spanning roughly five years, kicking off in 2027. The contract fuels the massive 4.5-gigawatt Project Stargate initiative, completely dwarfing OpenAI's expected $12.7B revenue this year. The mega-deal helped Oracle's future contract revenue surge past $317 billion, sending its shares soaring.
Replit just launched Agent 3, claiming a massive leap in autonomy over V2. The AI can now open an actual browser preview to click buttons, test user flows, and fix bugs automatically—3x faster and 10x cheaper than Computer Use models. It also runs up to 200 minutes solo and can build bots for Slack and Telegram.
Meta just signed a two-year deal worth $140 million with German startup Black Forest Labs to license its Flux image generation technology. The contract is backloaded, with Meta paying $35 million in year one and $105 million in year two. Why it matters: Following its recent Midjourney pact, Meta is pouring serious cash into leading visual labs to supercharge its creative tools.
Microsoft's Copilot Labs just launched Scripted mode for audio generation, powered by the new MAI-Voice-1 model. It reads your input verbatim instead of improvising, alongside an Emotive mode for added drama and a Story mode that performs multiple characters.
Google just launched AI Max for Search campaigns in beta to all advertisers globally, offering a one-click setup to optimize search ads using Google AI. The update brings built-in one-click A/B testing across Google Ads platforms and APIs, with custom brand-safety text guidelines rolling out later this fall.
OpenAI just rolled out full MCP tool support inside ChatGPT. In developer mode, developers can now create connectors for write actions like updating Jira tickets or triggering Zapier workflows, rather than just fetching data. It basically turns ChatGPT into a practical automation hub for daily engineering workflows.
Replit just launched Agent 3, an AI developer that can autonomously build, test, and debug code for over three hours straight without human intervention. Backed by a fresh $250M funding round at a $3B valuation, CEO Amjad Masad called it software's "full self-driving moment"—using a multi-agent architecture to boost uninterrupted runtime tenfold over its predecessor.
The co-founders behind Anchor just launched Oboe, an AI-powered platform that builds personalized learning courses in seconds from a single prompt. Powered by an internal multi-agent pipeline, it generates material across nine formats, including two-host podcasts, lecture audio, and interactive quizzes. Users get five free course creations monthly, with paid tiers starting at $15.
Stability AI just launched Stable Audio 2.5, an enterprise-grade audio generation model capable of pumping out 3-minute tracks in under two seconds. Built on fully licensed datasets for commercial safety, it introduces audio inpainting and dynamic song structures to help brands craft customized sound identities at scale.
Web giants like Reddit, Yahoo, and Medium just backed Really Simple Licensing (RSL), an open standard that lets site owners charge AI developers for scraping content. Built on robots.txt, the system allows publishers to set fees per crawl or per inference rather than just blocking bots outright. Why it matters: It attempts to give publishers collective leverage against Big Tech, though its success hinges entirely on AI companies playing by the rules.
Amazon's Zoox just opened its purpose-built robotaxi rides to the general public along the Las Vegas Strip. The toaster-shaped vehicles have no steering wheel or pedals, top out at 45 mph, and are currently free between a few select resort stops while regulatory permits are pending.
Coca-Cola just teamed up with MIT and OpenAI to launch "Save the Orange," an AI initiative targeting the devastating citrus greening disease. The plant epidemic has already slashed Florida's orange output by over 90%, and experts warn the world's orange juice supply could vanish within 25 years without new agricultural interventions.
ByteDance just launched Seedream 4.0, a new AI image model with native 4K support and fast image editing, directly challenging Google DeepMind's viral Nano Banana. It merges text-to-image with SeedEdit capabilities while boosting inference speeds by over 10x, and is now live across Jimeng, Doubao, and Krea AI.
Thinking Machines AI just debunked the prevailing theory behind why LLM inference yields inconsistent results even with greedy sampling. While developers usually blame GPU floating-point concurrency and race conditions, modern forward passes are actually deterministic. The real culprit is a lack of batch invariance, meaning an individual request's output subtly shifts depending on which concurrent queries share the batch.
Perplexity is prepping local Model Context Protocol (MCP) support for its Comet browser. While Comet previously only supported remote connections, local MCPs allow its agent to browse local folders, attach files, and trigger desktop apps from chat prompts. Why it matters: It pushes AI browsers beyond the web sandbox into full desktop automation, though the feature is still tucked away in dev builds.
ElevenLabs just rolled out Agents Testing, a new suite that lets developers run structured simulations on conversational agents before shipping updates. It validates tool calling, human handoffs, and guardrails, and can turn past customer chats into test suites inside your CI/CD pipeline. Why it matters: teams can finally update prompts without fear of silent regressions.
Google Research just teamed up with Stanford to launch AI Quests, an immersive game designed to teach 11- to 14-year-olds how AI tackles real-world challenges. The first quest has students collect rainfall data and train models to forecast floods, with upcoming missions exploring eye disease and brain mapping. Why it matters: It turns complex AI principles into hands-on play with ready-made curriculum for teachers.
Baidu just launched its ERNIE X1.1 reasoning model at WAVE SUMMIT 2025, claiming a 34.8% jump in factuality over X1 and performance rivaling GPT-5 and Gemini 2.5 Pro. Alongside the flagship, Baidu open-sourced a 21B thinking model and updated its Comate coding assistant with multi-agent workflows. The new model is already live on the web and mobile apps.
NVIDIA just submitted its first MLPerf Inference v5.1 results for the Blackwell Ultra architecture via the GB300 NVL72 rack-scale system. It swept performance records across newly added benchmarks like DeepSeek-R1 and Llama 3.1 405B, delivering roughly 5x higher per-GPU throughput than previous Hopper-based systems.
Geoffrey Hinton just warned that tech CEOs are deluding themselves about AI's rosy future. In an interview with the FT, the Nobel laureate said the wealthy will inevitably use AI to replace workers, creating massive joblessness and ballooning profits that a universal basic income cannot fix. He also bet that superintelligence will arrive within 5 to 20 years.
Harvard Medical School just unveiled PDGrapher, a free graph neural network that spots gene targets and drug combos capable of reverting diseased cells back to a healthy state. Tested across 11 cancer types, it ranked therapeutic targets up to 35% higher and ran 25 times faster than existing models.
OneMeta AI just showcased Verbum, a translation service that handles real-time voice and text translation for up to 50 people on a single call, across 82 languages and 40 dialects. Transcripts appear within a second, followed right after by natural-sounding AI speech. It is aimed at international team meetings, but could eventually break down language barriers in the metaverse.
Anthropic just gave Claude the ability to create and edit Excel spreadsheets, documents, PowerPoint slide decks, and PDFs directly in-app. Instead of returning plain text, Claude runs code in a private sandboxed environment to tackle multi-step analysis and data formatting. The feature is now rolled out to paid plans.
Apple just unveiled the Apple Watch Ultra 3, adding built-in two-way satellite messaging, a larger LTPO3 display, and up to 42 hours of battery life. Alongside machine-learning-driven hypertension risk alerts, it debuts Workout Buddy, an Apple Intelligence feature that delivers real-time spoken coaching tailored to your pace and heart rate. Why it matters: It deepens Apple's blend of outdoor survival and proactive health, though the AI coach requires a compatible iPhone.
Eli Lilly just launched TuneLab, an AI platform giving biotech firms access to drug discovery models trained on over $1B worth of proprietary research data. The setup uses federated learning so startups can leverage Lilly's predictive chemistry models without exposing their own intellectual property.
Apple just unveiled the iPhone Air, its thinnest phone yet at 5.6mm, powered by the flagship A19 Pro chip with custom Neural Accelerators delivering triple the peak GPU compute for on-device generative AI. Paired with iOS 26, it powers upgraded Apple Intelligence features like Live Translation and on-screen visual search. Pre-orders begin September 12.
Apple just announced the iPhone 17, packing an A19 chip designed with built-in neural accelerators in every GPU core to run on-device generative AI models. The phone pairs with iOS 26 to power upgraded Apple Intelligence capabilities like Live Translation and screen-wide visual search, starting at $799.
Apple just introduced Apple Watch SE 3, bringing an Always-On display, fast charging, and the S10 chip to its entry smartwatch. It also adds Workout Buddy powered by Apple Intelligence to deliver spoken motivation during workouts when paired with a supported iPhone. The upgrade substantially closes the gap between Apple's budget line and its flagship wearables.
Apple just announced AirPods Pro 3, packing twice the active noise cancellation alongside new health and AI features. Powered by Apple Intelligence, the earbuds introduce hands-free Live Translation for face-to-face chats and an on-device AI Workout Buddy.
NVIDIA just introduced Rubin CPX, a new chip system built for massive AI inference workloads like video generation and software coding. By packing video decoding, encoding, and inference onto a single platform, it cuts multi-chip bottlenecks for tasks demanding 1M+ token contexts.
Google just admitted in an antitrust court filing that the "open web is already in rapid decline," arguing that breaking up its ad business would accelerate the crash for publishers. That statement starkly contradicts its public claims that web traffic is thriving under AI search. Google later pushed back, arguing the line was cherry-picked and only applied to display ads.
Google just expanded AI Mode in Search to Hindi, Indonesian, Japanese, Korean, and Brazilian Portuguese. Powered by a custom version of Gemini 2.5, it taps multimodal reasoning to provide locally relevant answers rather than simple machine translation.
Lamborghini just revealed plans for an AI-powered driving system fed by a baseball-sized "6D sensor" that tracks six degrees of vehicle motion. The machine-learning algorithm dynamically adjusts torque, tightens safety nets if you mess up, and can even intervene via steer-by-wire to correct bad cornering. Lamborghini says future versions could even adapt controls to your mood.
Google just added direct audio uploads to the Gemini app, addressing what the company calls users' most-requested feature. Free users can upload clips up to 10 minutes long (capped at five daily), while AI Pro and Ultra subscribers get up to three hours. Google also upgraded NotebookLM with new report styles like study guides and flashcards.
Rabbit just rolled out rabbitOS 2 for its r1 pocket companion, overhauling the UI with a swipeable card layout. The headline feature is creations, letting users build tools and mini-games in minutes simply by describing them out loud. It also brings offline meeting summaries and on-device image generation via the intern agent.
OpenAI just opened applications for the first wave of grants from its $50M People-First AI Fund, aimed at helping U.S. nonprofits advance AI literacy, community innovation, and economic opportunity. The grants are unrestricted, and applicants don't even need prior AI experience to qualify before the October 8 deadline.
OpenAI is backing an animated feature film called Critterz to prove generative AI can slash cinema production costs and timelines. Built primarily with AI tools like GPT-5, the project is reportedly being made in nine months for under $30 million with partners Native Foreign and Vertigo Films, targeting a 2026 theatrical release.
Anthropic just endorsed California’s SB 53, a bill targeting frontier AI models trained with at least 10^26 FLOPS. Instead of rigid technical mandates, the bill requires big labs to publish safety frameworks, report critical incidents within 15 days, and protect whistleblowers, all while exempting startups. Anthropic says mandatory transparency keeps labs from cutting safety corners to compete.
Google just updated Veo 3 and Veo 3 Fast, bringing lower pricing across the board alongside new support for 9:16 vertical video and 1080p HD outputs. Both video generation models are now marked production-ready in the Gemini API and continue to include SynthID watermarking.
Mark Zuckerberg just promised Donald Trump at least $600 billion in U.S. AI spending through 2028, but a hot mic later caught him apologizing and admitting he "wasn't sure what number you wanted to go with." Zuckerberg quickly clarified on Threads that he wasn't sure which timeframe Trump meant, so he shared the lower estimate. It is a telling glimpse into tech's political maneuvering.
Anthropic just agreed to pay at least $1.5 billion to settle a class-action lawsuit accusing it of training models on pirated books. The deal pays out roughly $3,000 per work across an estimated 500,000 titles and requires Anthropic to destroy all downloaded files. Why it matters: It marks the largest copyright payout in U.S. history, setting a costly precedent for AI training data.
Roblox just introduced a suite of creator tools headlined by 4D Objects, generative AI that creates fully functional, drivable vehicles and interactive items from text prompts. The platform is also rolling out real-time voice translation and MCP support for its AI assistant, making game creation and global play remarkably seamless.
Roblox just announced a short-form gameplay feed called Moments alongside new creator AI tools at its developer conference. Creators can soon prompt fully functional 3D items like drivable cars rather than static meshes, plus tap into real-time voice chat translation and speech APIs.
Rokid just raised over $1M in pre-orders in under 72 hours for its new Rokid Glasses, challenging Meta's Ray-Ban monopoly. Despite weighing just 49g, the frames pack a built-in heads-up display, native ChatGPT support, and 89-language live translation. Why it matters: It shows a growing appetite for smart glasses that offer actual visual interfaces without ballooning into bulky headsets.
Google just integrated Gemini directly into Sheets with a new =AI function. Users can simply type the formula alongside a natural-language prompt into any cell to categorize feedback, summarize reviews, or generate personalized copy across hundreds of rows without touching complex scripts.
OpenAI is reportedly teaming up with Broadcom to mass-produce its own custom AI chips next year for internal workloads. Broadcom just disclosed a mystery $10 billion chip order that analysts widely pin on OpenAI. Much like Google and Amazon, the ChatGPT maker is moving in-house to slash skyrocketing infrastructure costs and loosen Nvidia's grip.
Snapchat just launched Imagine Lens, its first AI Lens powered by open text prompts. Instead of being locked into rigid preset filters, users can type whatever prompt they want to create or remix Snaps. It runs on a mix of Snap's internal and third-party models, but for now, it's exclusive to Platinum and Lens+ subscribers.
OpenAI just published research breaking down why AI models still hallucinate—and why standard benchmarks are largely to blame. Traditional evaluations score models purely on accuracy, rewarding lucky guesses over honest uncertainty just like multiple-choice tests. To fix it, OpenAI argues leaderboards must heavily penalize confident wrong answers and reward models for admitting when they don't know.
Oakland Ballers manager Aaron Miles just handed in-game decision duties over to an AI, marking a pro sports first in a regular-season game. Built by software firm Distillery in just two weeks, the system will set lineups, trigger pitching changes, call pinch hitters, and position fielders.
Timekettle just launched the W4 AI Interpreter Earbuds, a $349 wireless pair designed for easy sharing. They combine standard mics with bone-conduction sensors to accurately pick up whispers and cut through noisy rooms. Powered by Babel OS 2.0 and LLMs, they handle 42 languages and 95 accents with up to 98 percent accuracy.
OpenAI is developing an AI-powered hiring service dubbed OpenAI Jobs Platform, targeting a mid-2026 launch that puts it in direct competition with LinkedIn. The company will also pilot an AI fluency certification program in late 2025, teaming up with Walmart to certify 10 million workers by 2030.
Anker just launched Soundcore Work, a wearable AI voice recorder roughly the size of a quarter. The $159 clip-on gadget packs an 8-hour battery, transcribes over 100 languages with speaker separation, and uses GPT-4.1 to summarize meetings in one click.
Google just integrated its Veo 3 video generation model into a newly revamped "Create" tab in Google Photos. Available now in the U.S., users can animate still photos with prompts like "Subtle movement" or restyle them with Remix into anime and 3D renders. It basically turns everyday memories into cinematic clips without needing complex video editors, though free daily generations remain capped.
AI video startup Captions just rebranded as Mirage, signaling a major push into frontier research. The $500M company plans to build multimodal foundation models tailored specifically for TikTok and Reels, while bringing its creator app and ad platform Mirage Studio under one roof.
Adobe just announced that its flagship video editor Premiere is coming to iOS for free later this month, with an Android version also in the works. The mobile app packs multi-track timelines, 4K HDR support, and built-in Firefly generative sound effects alongside AI speech enhancement. The core editing features are free, though extra cloud storage and generative AI credits will cost you.
Atlassian just agreed to acquire The Browser Company, the startup behind the Arc and Dia browsers. The workplace software giant plans to shape Dia into a dedicated browser for knowledge workers, injecting AI skills, personal work memory, and enterprise-grade security. Why it matters: Atlassian wants to control the actual canvas where daily SaaS work happens.
Google just updated Circle to Search on Android with continuous translation as you scroll. Instead of forcing you to restart translation whenever screen content changes, a new "scroll and translate" toggle keeps text translated through endless feeds and app switches. It starts rolling out this week on select Samsung Galaxy devices.
Google just launched EmbeddingGemma, a compact 308M parameter open embedding model tailored specifically for on-device AI. It allows developers to run semantic search and local RAG pipelines entirely offline, keeping sensitive data private while sharing a tokenizer with Gemma 3n to save memory. High-accuracy retrieval is finally fitting inside lightweight local hardware.
Donald Trump hosted top tech leaders at the White House on Thursday to highlight AI research and push for domestic spending. Meta and Apple executives claimed $600 billion in U.S. investments while Microsoft committed up to $80 billion annually, with OpenAI's Sam Altman at the table and Elon Musk absent.
Fast-food giants Taco Bell and McDonald's are overhauling their drive-thru AI strategies after embarrassing glitches and pranks, including a viral order of 18,000 cups of water. McDonald's dropped IBM for Google Cloud's voice models, while Taco Bell plans to let human workers take over during rush hours. Why it matters: Even after early setbacks, chains are doubling down on voice AI to relieve overworked staff.
ByteDance just expanded its open-source UI-TARS agent series, rolling out UI-TARS-2 with upgrades across GUI workflows, coding, and tool use. Built on native vision-language models with reinforcement learning reasoning, it translates screenshots into system clicks, drags, and keystrokes. Why it matters: It turns complex computer and mobile tasks into reliable, hands-free automation.
Google DeepMind researchers just proved that single-vector embeddings have a fundamental geometric limit that breaks advanced RAG setups. On their new LIMIT stress test, top embedding models from Google and Snowflake saw recall drop below 20%, while old-school BM25 excelled. Why it matters: simply scaling models won't fix retrieval.
Eliza Labs just filed an antitrust lawsuit against X, claiming the platform coaxed technical details out of them for their ElizaOS agent framework, abruptly banned their accounts, and copied the tech for xAI. Eliza is seeking treble and punitive damages, though legal experts note that antitrust claims around open-source tools face an uphill battle.
Krea AI just introduced Realtime Video, a model that generates frames faster than playback speed. Creators can sketch on a canvas, stream their webcam, or share their screen to direct temporally consistent video in real time. It turns sluggish video generation into an interactive, game-like creative tool.
Microsoft just debuted its first homegrown models: MAI-Voice-1 and MAI-1-preview. The voice generator produces a full minute of expressive speech in under a second on a single GPU, while MAI-1-preview is an end-to-end mixture-of-experts foundation model trained on 15,000 H100 GPUs and now testing on LMArena. Microsoft is finally building its own muscle.
Apple just rolled out Xcode 26 developer beta 7, adding direct integration for paid Claude accounts and OpenAI's GPT-5. Developers can now tap into Claude Sonnet 4 or choose between standard GPT-5 and GPT-5 (Reasoning) to write code, tests, and squash bugs right inside the IDE. It is a clear sign Apple is opening up its walled garden to top frontier models.
Google Labs and DeepMind just launched Stax, an experimental developer tool built to replace "vibe testing" with structured evaluations. It allows developers to upload test datasets and deploy Gemini-powered pre-built or custom autoraters to score outputs on criteria like factuality, tone, and brand fit.
Anthropic just updated its consumer privacy policy to train AI models on user chats and coding sessions, retaining data for up to five years. The change takes effect September 28 across Free, Pro, and Max tiers with opt-in enabled by default, though enterprise and API accounts remain exempt.
PixVerse just launched its new V5 video generation model alongside an Agent feature, crossing 100M global users. The update improves motion consistency and visual clarity, rendering 360p clips in 5 seconds and 1080p in about a minute. The platform is offering free access during a launch event.
OpenAI just made its Realtime API generally available alongside a new flagship speech-to-speech model, gpt-realtime. The update brings native SIP phone support, image inputs, and remote MCP tool calling, while cutting audio token costs by 20%. Crucially, voice agents can now hold fluid conversations even while waiting for slow background tools to finish.
xAI just released Grok Code Fast 1, its first model built specifically for agent-based programming. It scores 70.8% on SWE-bench with aggressive pricing at $0.20 per million input tokens. Developers can test it for free on GitHub Copilot and Cursor, though early feedback suggests it still struggles on complex tasks.
Anthropic just updated its consumer terms and privacy policy, giving Free, Pro, and Max users a choice to let their chats and Claude Code sessions train future models. Opting in extends data retention from 30 days to five years, while commercial and API customers remain untouched. Users have until October 8, 2025, to decide in their privacy settings.
Tencent AI Lab and Washington University in St. Louis just unveiled R-Zero, a reinforcement learning framework that allows LLMs to self-evolve without any human-labeled data. By splitting a base model into competing Challenger and Solver roles, it autonomously generates progressively harder curriculum tasks. In benchmarks, it boosted Qwen3-4B's math reasoning score by 6.49 points. Why it matters: It could eliminate the massive data annotation bottleneck for reasoning models.
Tencent just open-sourced HunyuanVideo-Foley, an end-to-end text-and-video-to-audio framework designed to generate realistic Foley sound effects for videos. Trained on 100,000 hours of multimodal data, it uses an MMDiT architecture to tightly align sound with visual motion, outperforming existing open-source baselines in audio quality and temporal alignment.
University of Vaasa researchers just developed a smart packaging method that pairs color-changing inks with artificial intelligence. The system catches subtle environmental shifts with near human-eye accuracy, offering an affordable, easily recyclable way to track freshness in foods and medicines without electronic sensors.
Nvidia just reported record Q2 revenue of $46.7 billion, up 56% year-over-year, alongside $26.4 billion in net income. Growth was spearheaded by data center demand, with its Blackwell chips alone driving $27 billion in sales. However, regulatory uncertainty left the company with zero sales of its China-focused H20 chip to Chinese customers.
OpenAI reminded developers that its Codex CLI coding agent is already fully open-source under the Apache-2.0 license. The lightweight tool runs locally in terminal environments across macOS, Linux, and Windows, integrating with IDEs like VS Code and Cursor. Why it matters: users can run local terminal coding agents powered directly by their ChatGPT plans without waiting on closed releases.
OpenAI just rolled out big updates to Codex, now powered by GPT-5. The release brings a new IDE extension, smooth task transitions between cloud and local setups, GitHub code reviews, and an overhauled CLI tool. It is available right away through ChatGPT plans.
Aurelian just raised a $14M Series A led by NEA to deploy AI voice agents across chronically understaffed 911 dispatch centers. The voice assistant triages non-emergency calls like parking violations and noise complaints, while instantly handing real emergencies back to human dispatchers. Why it matters: In high-turnover public safety jobs, AI is stepping in to fill roles cities simply cannot hire humans for.
Alef just signed agreements with two Silicon Valley airports to begin test operations for its vertical-takeoff flying car. The all-electric vehicle promises 200 miles of driving and 110 miles of flight, supported by AI-driven safety systems. Integrating with existing airport traffic marks a critical real-world test for street-to-sky travel.
Lindy just launched Lindy Build, an autonomous web browsing agent designed to fix the most painful part of vibe coding. Instead of forcing you to hand-hold an AI through endless UI bug fixes, it clicks around your app, spots issues, and patches them 100% on its own. Basically, you can describe an app, grab lunch, and come back to working code.
Microsoft just brought its Copilot AI assistant directly to Samsung's 2025 lineup of smart TVs and monitors. Viewers can hold the remote's mic button to trigger a floating animated avatar for movie recommendations and spoiler-free episode recaps. Why it matters: It brings conversational AI to the living room's biggest screen, though it is currently restricted to newer hardware.
Andreessen Horowitz just dropped the 5th edition of its Top 100 Gen AI Consumer Apps ranking, showing a stabilizing market with fewer newcomers. ChatGPT still leads web and mobile, while Gemini and Grok are rapidly closing the gap. Why it matters: Vibe coding tools like Lovable jumped to #22, proving that AI app generators are driving real retention and user spending.
OpenAI and Anthropic just shared results from a first-of-its-kind joint safety evaluation, running their internal alignment tests on each other's flagship models. Claude 4 excelled at resisting system prompt extraction but refused up to 70% of hallucination probes, while OpenAI o3 held the line best against jailbreaks. Having fierce rivals probe each other's vulnerabilities is a rare, refreshing win for frontier model transparency.
Google just rolled out major generative AI updates to Google Vids as it tops 1 million monthly active users. Creators can now transform still photos into eight-second clips with native audio via Veo, deploy AI avatars to deliver scripts, and auto-trim filler words from recordings. Why it matters: Routine business video production no longer requires cameras, studio time, or real spokespeople.
Meta just rolled out Environment Generation to all creators in the Worlds Desktop Editor and teased upcoming fully-embodied conversational LLM NPCs. The new NPC system lets creators build characters that respond to voice input with custom personalities in VR and mobile. It basically makes building virtual worlds as easy as typing a prompt.
Anthropic just released an education report analyzing ~74,000 Claude conversations from higher education professionals. The data shows educators most often use AI for curriculum development and research, leaning toward creative collaboration for teaching while delegating routine administrative tasks to full automation.
Anthropic just released a threat report exposing how bad actors are weaponizing its AI. In one wild case, a hacker used Claude Code to automate network infiltration, extract data, and calculate ransoms across 17 organizations. The company banned the accounts and deployed new classifiers, but it shows agentic AI is officially in criminals' hands.
Anthropic just formed a bipartisan National Security and Public Sector Advisory Council packed with heavyweights, including former Acting Defense Secretary Patrick Shanahan and former CIA Deputy Director David Cohen. The group will guide the AI lab's defense and intelligence rollouts in areas like cyber defense and analysis. It is a clear push to cement frontier AI inside the U.S. government.
Ming-Chi Kuo revealed that Meta's Hypernova AI glasses are expected to enter mass production in 3Q25, with shipments projected at just 150,000 to 200,000 units over two years. The $800 device leans on AI as its core selling point, but hardware hurdles and an early AR ecosystem make it more of an experimental hedge against Apple.
