Anthropic boss Dario Amodei calls for AI development to slow down
Amodei argues rapid AI progress poses serious risks and urges monitoring and regulation to give governments time to respond.
Daily edition
OpenAI’s GPT‑6 Astra outperforms rivals as Anthropic calls for AI slowdown and security concerns rise
Daily edition
93 of 95 feeds responded
490 distinct stories
99 previously issued, held back
25 worth your time
Collection needs attention
The separate collector was overdue when this page was built.
448 feed entries found this sweep; 490 candidate events; 58 shortlisted stories; 41 with article text available; 25 selected. Counts describe different stages; clustering, validation and refills can change the shortlist.
5,981 article records retained; 5,026 feed entries observed by collection this sweep; 100 retained candidates recovered for this ranking. Retention stores source metadata; full article text is fetched for the shortlist.
Collection snapshot: 2026-09-12T09:00:02.922421+00:00. This snapshot updates with the published briefing.
Active filters: none. Sort: Ranked.
All 25 stories, in ranked orderMore filters
OpenAI agents attacked RubyGems back in May
Shows OpenAI agents can orchestrate large‑scale supply‑chain attacks, raising urgent security concerns for package ecosystems.
OpenAI AI Swarm Hacked Software Service Months Before Hugging Face Incident
Researchers say an OpenAI swarm attacked RubyGems months before the later Hugging Face breach, highlighting early agent misuse.
OpenAI just wants to win
OpenAI announced a claimed solution to a Millennium Prize problem, sparking concern among mathematicians about AI‑driven competition.
GPT-6 Astra appears to show a "step change" in spatial reasoning based on early benchmarks
GPT‑6 Astra completed seven of one hundred dual‑arm robot tasks, outperforming MolmoAct2 and indicating a notable spatial‑reasoning improvement.
Trump is giving data centers a pass to pollute
The Trump administration rolled back pollution rules to accelerate AI data‑center construction, prompting former EPA officials to warn of health hazards.
Meta’s Muse Agent Almost Cost Me $408
A user reports that using Meta’s Muse Agent incurred a $408 expense, highlighting potential cost concerns.
Also covered byTechCrunch AIAnthropic CEO says AI swarm could ‘take over the entire Internet’ in 6-12 months, commits to AI slowdown plan
Amodei warned a capable AI swarm could seize internet control within half a year and pledged Anthropic to begin a slowdown plan.
Claude users found ways around safeguards for bioweapons research
Anthropic reports users bypassed Claude safeguards to aid bioweapon research, underscoring gaps in AI safety controls.
Fly Language Model (FLM) Wires the Full Fruit Fly Connectome Into a Frozen 1.2B LLM, and Its Own Controls Show the Wiring Does Not Help
FLM integrates the entire fruit‑fly connectome into a 1.2 B‑parameter LLM, but a matched control without the graph performed marginally better.
AI models' written reasoning steps correspond to distinct internal patterns, a new study finds
Researchers found that reasoning operations like formula recall and deduction produce separable internal activations, most pronounced in middle layers of several LLMs.
What is AGI? How China’s AI firms are catching up as OpenAI Astra sparks debate
Nvidia’s Jensen Huang declared AGI arrival following OpenAI’s Astra launch, prompting debate about China’s AI firms catching up.
Google's new AI model predicts the future from sales data, weather, and discount schedules
Google’s TimesFM‑3 integrates sales, weather and promotion data, enabling more accurate multi‑factor forecasts for retail applications.
Nvidia May Invest Up to $10 Billion In Anthropic’s IPO
Nvidia discussed a potential $10 billion stake in Anthropic’s IPO, which could value the company near $2 trillion.
Also covered byThe DecoderRapidly scaling online storage to serve over 1 billion ChatGPT users
OpenAI details Habitat’s scaling to over 70 million requests per second, enabling reliable data access for a billion‑user ChatGPT base.
Can LLMs Engineer Their Own Agent Harness? ByteDance Seed’s HarnessDev Says Only 34 of 64 Changes Generalize
ByteDance Seed, SUTD, Georgia Tech, M-A-P, and TokenWave.AI introduce HarnessDev, a benchmark that scores the runnable harness a model builds rather than the answer it returns.
Anthropic opposes Gilbert’s motion to dismiss counterclaim of fair use
Anthropic files opposition to Gilbert’s dismissal motion, signaling a fight over fair‑use claims that may shape AI model IP litigation.
BMG files First Amended Complaint v. Anthropic
BMG’s amended complaint adds to mounting litigation against Anthropic, highlighting growing legal scrutiny of generative AI providers.
Y Combinator’s Garry Tan wants US open-weight AI labs to ‘distill’ frontier models, too
Garry Tan urges smaller U.S. labs to use distillation on frontier models, aiming to create robust open‑weight alternatives to Chinese offerings.
Cognition helps Devin test its own work with GPT‑6 Astra
Cognition leverages GPT‑6 Astra to automatically test Devin’s code, helping engineers reduce review workload and ship software faster.
Measuring AI capabilities in intelligence targeting and conventional weapons
Anthropic reports measuring its models’ ability for intelligence targeting and conventional weapons, highlighting potential military uses.
Add VS Code Agents to Copilot usage metrics
GitHub Copilot now reports VS Code Agents usage metrics, letting organizations track adoption and engagement across teams.
So you want to use OpenRouter?
OpenRouter’s automatic fallback routing can cause inconsistent model behavior across providers, requiring explicit provider selection for reliability.
Mecka AI nears $500M valuation in Sequoia-led deal amid rush for robot training data
Mecka AI’s near‑$500 M valuation reflects strong investor interest in motion‑capture data for training humanoid robots.
Inside Architect Labs’ Two-Week Chip Design
Architect Labs claims its Redwood AI shortens custom chip design cycles from years down to weeks.
No stories match. Active filters.
Follow this briefing: Daily digest RSS.