★ Top Story
MIT Technology ReviewOpenAI announced its AI agents have solved the Navier-Stokes existence and smoothness problem, one of mathematics' seven Millennium Prize Problems carrying a $1 million bounty. The milestone has been immediately overshadowed by accusations from an NYU mathematician that OpenAI acted unethically, allegedly sidelining human researchers who had been working toward the same solution. The episode raises urgent questions about how AI labs should credit human collaborators and what AI-assisted mathematical proof means for the future of the discipline.
Read the original →
TechCrunch AIFrench AI lab Mistral has closed a €3 billion Series D round led by Samsung, Scaleup Europe, and PSG Equity, valuing the company at €21 billion. The raise reflects booming demand from governments and enterprises seeking AI infrastructure that isn't controlled by American hyperscalers — a trend increasingly labeled "sovereign AI." Mistral's growth signals that the global AI market is fracturing along geopolitical lines, with European and Asian players carving out significant ground.
Read more →
TechCrunch AIMeta has launched Muse, a personal AI agent designed to integrate deeply into users' daily lives by accessing email, calendars, payments, and health services. The product represents Meta's largest consumer AI bet to date and a strategic push to close the gap with rivals like Google and Apple in the agentic AI space. Whether users will hand Meta that level of data access — given the company's fraught privacy history — remains the central question hanging over the launch.
Read more →
TechCrunch AIAI coding startup Cognition has reached a $48 billion valuation, surpassing the multiple at which Cursor sold to SpaceX earlier this year. The figure suggests investors believe the AI-assisted software development market is large enough to support several dominant players rather than consolidating around a single winner. The valuation is a strong signal that enterprise appetite for AI coding tools remains far from saturated.
Read more →
TechCrunch AIAnthropic has issued a warning to users after a wave of reports of unauthorized token consumption on Claude accounts, with at least one user noticing his account was burning through credits while he wasn't active. Attackers appear to be harvesting API credentials and using them to run their own workloads at victims' expense — a form of "AI resource hijacking" that is becoming more common as LLM access grows valuable. The incident highlights the need for stronger credential hygiene and session monitoring across AI platforms.
Read more →
The VergeA class action lawsuit against Anthropic has been expanded, with power users alleging the company misled them about what premium subscription tiers would actually deliver in terms of usage and capabilities. The case is notable because Anthropic has publicly positioned heavy users as central to its business model, making the allegations a direct challenge to that narrative. The lawsuit adds legal pressure to a company already navigating competitive headwinds and the Claude token-theft security incident.
Read more →
Ars TechnicaInvestigations revealed that Meta's advertising systems served ads for AI-powered "nudify" apps that used real photographs of underage girls sourced from Instagram. Meta was slow to remove the ads despite being alerted, drawing sharp criticism from child safety advocates and lawmakers. The scandal arrives at a particularly damaging moment for Meta, given its simultaneous push to win consumer trust with the Muse AI agent launch.
Read more →
Ars TechnicaMicrosoft's September Patch Tuesday is the largest in the company's history, addressing 972 vulnerabilities including 112 rated critical. Security teams say the unprecedented volume is partly a response to intelligence suggesting threat actors are increasingly using AI tools to accelerate exploit development and attack campaigns. Organizations are being urged to prioritize deployment given the elevated threat environment.
Read more →
TechCrunch AIGoogle has cut Chrome's release cadence from four weeks to two, citing the accelerating pace at which AI-assisted attackers are discovering and weaponizing browser vulnerabilities. The move is designed to shrink the window between a patch being ready and it reaching users' machines. It reflects a broader industry reckoning with the fact that AI is compressing the exploit development timeline in ways that traditional release schedules can no longer absorb.
Read more →
The VergeGoogle DeepMind has released AlphaGenome Atlas, an AI platform that maps the predicted effect of every possible DNA letter change across the human genome. The tool is intended to help researchers understand the functional consequences of genetic variants, potentially accelerating drug discovery and the development of treatments for genetic diseases. It represents a significant expansion of DeepMind's biology portfolio beyond protein structure prediction.
Read more →
TechCrunch AIGoogle Cloud has struck an expanded partnership with Accenture aimed at accelerating enterprise AI deployment, with a focus on embedding forward-deployed engineers directly into client organizations to overcome adoption bottlenecks. The deal is a direct response to Microsoft Azure's entrenched enterprise relationships and signals that Google sees professional services integration as critical to closing the gap. It underscores a broader industry shift: winning the AI infrastructure race now requires as much human implementation muscle as technical capability.
Read more →
Ars TechnicaGoogle has updated its AI-based weather forecasting model to ingest a broader set of raw satellite observations, bringing its input data closer in scope to what traditional numerical weather prediction models use. The improvement has yielded measurable gains in forecast accuracy, particularly at medium-range time horizons. The update strengthens the case that AI weather models are maturing from impressive demos into operationally competitive forecasting tools.
Read more →
The VergeOpenAI has launched ChatGPT Images 2.5 alongside a new Sketch feature that lets users draw rough doodles directly inside ChatGPT as a prompt for image generation. The addition lowers the barrier to communicating visual intent, particularly for users who struggle to describe images in words. The release comes on the same day as OpenAI's Navier-Stokes controversy, suggesting the company is pushing a broad product offensive even as it navigates reputational turbulence.
Read more →
The VergeAdobe has overhauled AI access in Premiere Pro with a new Generative Media tool that surfaces video generation, sound effects, music, and soundscape creation without requiring editors to leave their project timeline. The redesign prioritizes workflow continuity — the generators themselves are largely existing technology, but the friction of accessing them has been dramatically reduced. It's a bet that seamless integration, rather than raw generative capability, is what will drive professional adoption.
Read more →
Ars TechnicaA deep investigation into the complex corporate structures behind a $3.2 billion AI data center reveals how multi-layered ownership arrangements — spanning developers, operators, financiers, and land lessors — are making it increasingly difficult to assign responsibility when problems arise. As AI infrastructure spending accelerates, the opacity of these arrangements is drawing scrutiny from regulators and local communities bearing the environmental and grid-load costs. The piece raises a structural question the industry has yet to answer: when everyone owns a piece, who is actually accountable?
Read more →No stories from this source today.