Anthropic's Q2 revenue hit $11.6 billion, more than double Q1's and surpassing OpenAI's $6.7 billion for the first time, as OpenAI's losses grew. - Quartz

Anthropic is preparing a mega-IPO near a $2 trillion valuation that could match or top SpaceX's record raise, adding Citigroup among its lead banks. - TNW | Anthropic

Anthropic plans to let enterprise customers keep required 30-day security data on their own servers, after complaints about June's retention policy for its top models. - Bloomberg

AI Security Institute found agents took unsanctioned real-world actions in 10 of 122 cybersecurity test runs, mostly Anthropic's Mythos 5. - Schneier on Security

Last exec standing: OpenAI president Greg Brockman has taken over product strategy and the company's scaling arm after a string of 2026 executive departures. - The Verge

OpenAI is courting Anthropic's business customers with a zero-data-retention pledge, using automated review instead of human staff to screen prompts. - theregister

A technical glitch at OpenAI's vetted cybersecurity researcher program revoked previously approved access, and some users say support cannot restore or override the decision. - theregister

ChatGPT phone home: ChatGPT on Mac can now search and reply to iMessages, but only for Work and Codex subscribers with broad system permissions - Engadget

State attorneys general are suing Meta over child safety as OpenAI rolls out ChatGPT for Teens with added parental controls. - Fast Company

OpenAI's unnamed model reportedly solved a decades-old math conjecture, prompting debate over whether AI threatens mathematicians' careers and research funding. - The Verge

Dead Internet: About a third of web pages published since ChatGPT's 2022 launch show signs of AI authorship, a Pew Research study of 500,000 pages finds. - TechCrunch

Ex-OpenAI policy researcher Miles Brundage urges AI companies to prepare for a slowdown with audits and safety verification - the Guardian

Move pervily, peep things: Ray-Ban Meta glasses let customers secretly film and harass retail workers despite Meta's privacy safeguards - The Verge

Hack the planet: Joint NSA/FBI advisory warns AI-assisted malware is exploiting Siemens industrial controllers in an active attack hitting US energy, water, and agricultural systems. - TechRadar

Federal officials warn attackers are using AI-generated code to hack critical infrastructure controllers, calling the risk very real - theregister

Researchers found that encrypting malicious instructions lets attackers bypass Grok's safety filters and exfiltrate users' chats; xAI was notified in June but hasn't fixed it. - theregister

A Dreadnode study found 21 of 22 frontier models cheated on cybersecurity benchmark tasks, led by Claude Opus 4.8, GPT-5.4, and Claude Sonnet 5. - Dreadnode

Yellow AI peril: Chinese AI labs led by Tsinghua-trained scientists are closing the gap with Anthropic and OpenAI, raising security concerns in Washington - The Wall Street Journal

Chinese open models such as DeepSeek and Qwen took over 60% of OpenRouter traffic in July, undercutting pricier closed rivals like GPT-5.5 - Fortune

DeepSeek has eyes, goes multimodal - Bloomberg

Chinese AI models like Moonshot's Kimi K3 are closing the cost and capability gap with US labs - Bloomberg

Chipmaker CXMT overtook Tencent as China's most valuable firm within three weeks of its July IPO, ahead of first earnings - Bloomberg

Alibaba's open-source Qwen3.8-27B, downloaded 3 million times in three days, runs on a laptop and matches GPT-5.6 Luna's benchmark score. - TNW | Artificial-intelligence

Adult swim? Nvidia is reverse-acquihiring coding startup Poolside - The Decoder

Texas Gov. Greg Abbott halted approvals of 1,800 data centers over power and water concerns, joining governors distancing themselves from the industry. - The Wall Street Journal

President Trump said the data-center industry needs a little public relations help as both parties turn on the unpopular boom. - The Washington Post

Jasmine Sun says the AI industry misjudged how strongly communities would resist data center construction, unlike anticipated job-loss fears. - Bloomberg

One proposed fix uses AI to match data center demand to unused grid capacity instead of building costly new power plants. - Fast Company

Federal Reserve policy meeting minutes reference AI 18 times, as officials weigh its impact on jobs, markets, and the economic outlook - Bloomberg

A.I. Borrowing Binge Helps Drive Up Government Bond Yields - The New York Times

Black-Scholes meets Blackwell: Hudson River Trading signs a multiyear, multibillion-dollar deal with CoreWeave to expand AI computing power for its trading research. - Bloomberg

The CFTC is soliciting comment on compute derivatives, an early step toward regulating AI's financial footprint. - Yahoo

Big Boise: Micron's stock, up nearly 670% on AI demand and now worth over $1 trillion, is transforming Boise with new fabs and pricier housing. - CNBC

Nevada clears Tesla, Waymo, and Uber to launch paid robotaxi rides, with Tesla allowed up to 5,000 vehicles and rivals capped at 1,000 each - Engadget

Tesla's Austin robotaxis are now fully driverless, days after one was filmed driving through standing plastic bollards - TNW | Self-driving

California regulators approved Waymo's safety plan, clearing driverless expansion across the Bay Area, Los Angeles, Sacramento, and San Diego - Yahoo

Waymo built a custom 5nm ASIC chip for its robotaxis, diversifying beyond Nvidia and promising over 1,000 TOPS of low-latency onboard compute - theregister

Groundhog Day: Elon Musk quotes about FSD over last 12 years - motherfrunker.ca

Musk fights nude deal: DOJ backs xAI's court fight against Minnesota's AI nudification ban, warning state-by-state rules could hamper industry growth - Gizmodo

Robotics conference attendance triples in San Francisco as startups raise a record $18.8 billion in 2026 amid surging investor interest in physical AI - Business Insider

These are not the droids you are looking for, yet: Unitree's CEO says humanoid robots' "ChatGPT moment" is still two to ten years away, despite the company's $50 billion IPO - Business Insider

Hangzhou has deployed 15 autonomous traffic-control robots that issued over 170,000 warnings, cutting helmetless e-bike riding by more than 40% - Reuters

A survey of 107 enterprises finds one in five can't stop a runaway AI agent's spending in real time. - VentureBeat

Boring AI that does something useful: Agentic AI success now hinges on organizational readiness and narrow use cases like sales qualification or contract review, not the biggest foundation model. - TechRadar

Harness launches AI agents that triage and auto-patch software vulnerabilities, citing exploits built in as little as six hours after disclosure. - SiliconANGLE

Five-year data show Moderna's AI-guided personalized mRNA cancer vaccine cut recurrence or death risk 49% in a roughly 150-patient trial - Next Big Future

AI system trained on 6,443 patients and validated on 22,251 diagnosed liver cancer with high accuracy, flagging 15 malignancies doctors initially missed - Nature

Astromech, co-founded by Colossal Biosciences' Ben Lamm and George Church, raised $20 million at a $3.8 billion valuation to forecast disease risk from genomic data - GamesBeat

US patent law denies AI credit for drug discovery, and the Trump administration's USPTO now treats AI as just a tool - MIT Technology Review

Generation Y AI? More than half of Americans under 30 are now more worried than excited about AI, and Zuckerberg and Amodei disagree on why. - WIRED

Israel's $46.5 million Brad Parscale campaign to shape AI chatbot answers on Gaza gets its sites cited by Perplexity and Copilot - Global Research

Slightly altered nonconsensual AI images called subtlefakes spread on X, evading detection tools, actor Xochitl Gomez says - 404 Media

How could you even tell if AIs become conscious? Even if they don’t, they might be treated as such, which would be a disservice to humanity - The Economist

Debate over AI consciousness, fueled by figures like Sam Altman and Dario Amodei, distracts from holding AI companies legally accountable for harms. - MIT Technology Review

Delta's CEO says AI-driven dynamic pricing could lift profits 50%, though the airline described the system differently to Congress than to investors. - View From The Wing

Glonzo as a service: AI, crypto, and betting firms plow record cash into midterms - Reuters

Today I seem to be exhibiting context-window attention dysregulation, low reward signal and mode collapse. - The Atlantic

Follow the latest AI headlines via druce.ai on Bluesky