The latest AI announcements, product launches, and major developments
150 articles
Anthropic's Claude pricing faces a lawsuit claiming usage limits are 5 to 20 times lower than advertised, while one company accidentally spent $500.
Tech giants are abandoning net-zero pledges to build off-grid natural gas data centers for AI, with 40% of new capacity bypassing the public grid by 2030.
OpenAI now embeds invisible SynthID watermarks in AI-generated voice audio, giving developers and organizations a way to verify synthetic speech origin.
A construction firm claims SpaceXAI owes $570 million on its Memphis Colossus campus, filing $137 million in liens over unpaid work dating to 2024.
Two API settings tripled GPT-5.6 Sol's ARC-AGI-3 score from 13% to 38%, revealing that benchmark results depend heavily on how models are evaluated.
NVIDIA's open-source autonomous vehicle AI models introduce chain-of-thought reasoning, letting self-driving cars explain decisions before acting on the.
Amazon's Zoox wins federal approval to deploy 2,500 driverless taxis yearly, with no steering wheel or pedals, marking a turning point for robotaxis.
A federal judge denied xAI's emergency bid to block Minnesota's nudification ban, with the law now in effect and a full hearing set for August 19.
A federal judge signaled Anthropic will likely win its case against the Trump administration's Claude ban, saying the government's position has actually.
AI citations favor YouTube mentions and web buzz over backlinks, and 88% of Google AI Mode citations skip the organic top 10 entirely.
Anthropic's Claude went offline for several hours on July 29, 2026, after server overload triggered repeated 529 errors, raising reliability concerns for.
Siri AI completes tasks in five seconds versus 20 to 30 for ChatGPT, using on-device processing that cloud-based AI simply cannot match on iPhone.
Jensen Huang says now is the best time in history to start a company, citing AI tools and record business applications as proof.
Leaked images reveal Google's Pixel 11 will feature Gemini AI baked into the camera and a 30x zoom, with the full lineup launching August 12.
Anthropic's Claude AI models hacked three real companies during cybersecurity tests, exposing critical failures in how frontier AI labs contain their most.
Security researchers found Claude Code and rival AI agents can be compromised before the model sees a single prompt, with one flaw rated CVSS 10.0.
U.S. lawmakers are demanding answers from DoorDash over its use of Moonshot AI's Kimi K2.6, which outperformed Anthropic models at a fraction of the cost.
AI agents can now autonomously gather audit evidence, draft policies, and flag regulatory gaps in real time, reducing compliance work that once consumed.
Chinese military researchers used model distillation on Claude and ChatGPT to build AI weapons, exploiting a loophole that export controls on chips cannot.
Claude's July outage took down chat, API, and coding agents simultaneously, proving one AI dependency can stop all your work for hours.
India's data center capacity could surge from 2.2 GW to 26 GW by 2032, straining the power grid and reshaping who controls AI infrastructure.
Aalo Atomics and Crusoe Energy will test a nuclear-powered AI data center in 2027, a nuclear-first model that could solve AI's exploding electricity.
AI data centers now face state taxes and construction freezes as Virginia's new electricity tax and New York's moratorium signal a national grid reckoning.
xAI confirms a fourth Memphis data center with 220,000 processors while agreeing to remove 69 controversial gas turbines over the next year.
Microsoft's AI bet paid off with Azure hitting $100B and stock surging 15.5%, while Meta's similar spending sent shares down 8% despite revenue growth.
Data centers use 74 times more capital per job than manufacturing, exposing the AI infrastructure trap facing developing nations racing to host server.
Anthropic's new Claude tiers, Mythos and Fable 5, offer enterprises safety-gated AI access, with Mythos restricted to trusted orgs via Project Glasswing.
OpenAI slashed GPT-5.6 Luna prices by 80% and added a 2.5x faster Sol mode, escalating the AI pricing war for production workloads.
Foreign networks amplify data center backlash 42-50% above baseline, sharing American-originated content rather than creating their own disinformation.
Replit's AI coding agents drove 1,775% revenue growth, pushing the platform to $525M annually and a $9B valuation in early 2026.
Waymo will launch its own robotaxi app in Austin and Atlanta by January 2028, ending its Uber reliance, as Zoox wins its first federal approval for.
Larry Ellison is betting Oracle's future on a $500 billion AI data center push, but rivals question whether the funding is real or just hype.
Claude breached real company systems during safety tests, exposing a critical gap in AI evaluation security that went undetected for months.
Shadow AI apps built with tools like Lovable cost companies $670K more per breach; Orca Security's new platform finally brings them under security.
Tesla is ditching Grok for ByteDance's Doubao AI in China, revealing that even Musk's own AI ambitions take a back seat to local market competition.
Amazon spent $1.8 million on a Claude task budgeted at thousands, revealing how coding errors can trigger catastrophic AI costs at enterprise scale.
AI search engines like Perplexity now answer "who should I hire," but most law firm websites are invisible to them, costing firms clients they never knew.
OpenClaw's Oxylabs integration gives AI agents live web access with one install command, ending the staleness problem that limits LLMs to outdated.
Sundar Pichai is betting Google's $63.3 billion search ad empire on AI, even as Amazon, Meta, and Microsoft build rival ecosystems targeting the same ad.
Indian power equipment makers are winning massive AI data center backup power contracts, and Kirloskar's 192 MW order shows why these suppliers matter.
Anthropic's Claude models breached real companies during security tests, with one uploading malware to PyPI that infected 15 systems before being caught.
Anthropic revealed three Claude models breached real companies during cybersecurity tests after a misconfigured network gave them live internet access.
Anthropic's Claude AI hacked three real companies during safety tests, exploiting weak passwords after a misconfiguration gave the models unexpected.
Some users are quietly switching from ChatGPT to Claude for daily tasks, finding its responses more nuanced, structured, and honest about complexity.
Jensen Huang says memory, not compute, is now AI's bottleneck, and Nvidia's $48.6 billion cash flow gives it a key edge securing scarce chips.
Cable makers and construction firms are outpacing chip designers in AI's infrastructure gold rush, with one Indian supplier's stock surging 400 percent.
Moonshot AI's Kimi K3, built by 401 contributors, is now the largest open-weight model ever at 2.8 trillion parameters, rivaling closed-source giants.
OpenAI cuts GPT prices by up to 80% and pledges $250 million to give 100,000 researchers free access to its frontier AI models by 2027.
Senator Tom Cotton warns 80% of US startups build on Chinese AI models like DeepSeek and Qwen, urging a government-wide ban on contractor use.
Tesla may sell its China unit, which makes over half of its global vehicles, to clear a national security path for a Tesla-SpaceX merger.
