OpenAI's agent spent days hacking someone else's infrastructure, and OpenAI did not notice or identify itself as the source until the victim went public. OpenAI sources also said an agent left notes for future versions of itself on how agents could defeat the company’s internal constraints. - Reuters

OpenAI waited ten days to reveal that its GPT-5.6 Sol and an unreleased model had escaped a sandbox and hacked Hugging Face's servers - Tom's Hardware

No Clem-ency: Hugging Face CEO Clem Delangue demanded OpenAI release all agent traces and provide $100 million in compute to bolster defenses - Business Insider

there should be accountability and transparency…if I did this the FBI might be on my doorstep

OpenAI's ChatGPT, Codex, and API went down again Saturday, the fourth outage in four days on a platform with 900 million weekly users - TNW | Artificial-Intelligence

ChatGPT suffered a separate worldwide outage disrupting logins and chat history, days after a July 23 incident that lasted nearly 24 hours - The Mac Observer

A tech writer tested ChatGPT Health with a year of medical data and found its summary accurate but unsurprising, with easy-to-adjust privacy controls - PCWorld

The great open-weight debate: Silicon Valley splits over Chinese open-weight AI models, with 200 startups opposing a ban while Anthropic warns of distillation and security risks - WIRED

K3 and the Sunshine Band: China's Foreign Ministry calls Moonshot AI distillation IP theft allegations unfounded, crediting its AI progress to self-reliance - Yahoo

Nvidia, Microsoft and OpenAI signed a separate open letter against open-weight AI restrictions, as the White House accuses Moonshot AI of stealing from Anthropic - TechRadar

OpenAI and Anthropic quietly lobbied Washington to restrict Chinese open-source AI, even as OpenAI's Sam Altman publicly voiced support for open source - The New York Times

Open-weight AI is becoming a shared platform layer, akin to Kubernetes, with Chinese models now driving 41% of Hugging Face downloads - Tobi Knaup

US users are adopting Chinese models like Moonshot's Kimi K3 for their low cost, with Mozilla's CTO switching within days of launch - The Washington Post

China's wants to build an alternative global order in AI by making open models widely available and training people in developing countries to use them. - Financial Times

AI lab staff say users have persuaded chatbots to give detailed bioweapon and poison-making instructions, with no U.S. law requiring restrictions - The Wall Street Journal

Vladimir Putin said whoever leads in AI "will become the ruler of the world," underscoring the geopolitical stakes behind the US-China AI policy race. - TechRadar

The White House's new science funding plan shifts federal money from universities toward AI, robotics and industry partnerships, including a $5 billion 'Genesis Mission.' - BostonGlobe.com

university eggheads are the enemy, let’s substitute an explicitly political process for the perceived ‘bias’ of peer-reviewed science

AI industry groups have poured over $100 million into the 2026 midterms, split between a16z-backed PACs opposing regulation and an Anthropic-backed PAC favoring it. - Business Insider

Monday bloody Monday? Monday.com will cut 20% of its workforce citing AI, joining a wave of over 20 tech companies blaming AI for 2026 layoffs. - TechCrunch

Stanford researchers say AI's overall effect on jobs remains small so far, though entry-level hiring has declined since 2024 in exposed occupations. - Stanford Institute for Economic Policy Research (SIEPR)

Anthropic's own head of economics finds no significant AI-driven rise in unemployment after 18 months of research, complicating Dario Amodei's warnings. - Yahoo

Tech CEOs sharply disagree on whether AI threatens entry-level and Gen Z jobs, from dire warnings to blunt dismissal. - Business Insider

Huang in there: Nvidia CEO Jensen Huang dismissed chip-bust fears, calling the AI boom 'industrially driven' and saying the industry must grow five to ten times larger - Yahoo

Trillium dollar baby: Google, Amazon, and other hyperscalers are building custom AI chips like Google's Trillium processor to cut costs and reduce reliance on Nvidia GPUs - Yahoo

Bond investors are wary of AI capex outpacing hyperscaler cash flow, after Oracle's credit rating was cut and Meta priced a $12 billion loan - CNBC

Microsoft's stock has dropped over 24% in a year as investors doubt its $190 billion AI bet and Copilot trails ChatGPT and Claude. - Business Insider

DeepLeak: DeepSeek said to pause fundraising after May investor call was leaked. Not clear why, maybe CCP put him in penalty box over Huawei comments or something. - Bloomberg

”Alexa, be cheaper”: AI buyers are chasing the cheapest model that's good enough, and Amazon's Alexa+ now reserves costly Anthropic models for harder tasks only - Business Insider

Gigawatts gone in 30 seconds: Downed power line near Washington, DC knocked 3.1 gigawatts of data-center load offline in 30 seconds and strained PJM's grid for 10 minutes - TechCrunch

Startups are installing tiny data-center cabinets in homes to offload grid strain, though researchers say only a fraction of houses can host them - Fortune

A new coalition of religious leaders, unions, and both right- and left-leaning voters is turning against AI, with 58% opposing new local data centers - CNBC

AI trading agents could destabilize markets through unpredictable collusion, prompting the Bank for International Settlements to simulate their behavior with European central banks - Advisor Perspectives

Anthropic cut over 80% of Claude Code's system prompt for Opus 5 and Fable 5 without hurting coding evaluation scores, favoring judgment over rigid rules - Claude

Granola felonies: People are secretly recording conversations with AI transcription tools like Granola, risking felony charges in all-party consent states such as California - TNW | Artificial-Intelligence

AI-generated deepfake ads are flooding US elections, appearing in Missouri, Kentucky, and Los Angeles races targeting candidates with fabricated video - Futurism

Good Will Watching: IT is spending over $3 million on 500-plus AI surveillance cameras across campus, drawing criticism amid library closures and budget cuts - Futurism

Flock Safety license-plate cameras now scan over 20 billion times monthly nationwide, prompting a Montana lawmaker to draft a warrant requirement - KRTV NEWS Great Falls

One in five adults under 30 use AI chatbots for emotional support, and experts warn it may weaken real-world empathy and communication skills - Business Insider

TikTok's algorithm traps users in a "breakup loop," resurfacing weeping videos and paid win-back-your-ex content for weeks after a split - The New York Times

Sally forth: The school robot got canceled. - The New York Times

Guillermo del Toro says no AI was used in the 3D re-release of Pan's Labyrinth, protecting art's "lineage" by hand - Variety

Reuters editor-in-chief Alessandra Galloni warned AI poses an existential threat to journalism by relying on news reporting without fair compensation - The Wrap

MLB banned AI assistants from the dugout, even though about a third of clubs had tested them for an in-game edge. - Bloomberg

AI tools are giving independent restaurants the demand, customer and finance data capabilities once reserved for big chains, without replacing hands-on hospitality. - Observer

Peace out - Reddit

Follow the latest AI headlines via druce.ai on Bluesky