#ai
-
Anthropic's Dario Amodei set for first dinner with Trump
Anthropic's chief is about to hold his first one-on-one with the US president — despite their split over AI safety.
-
The "Human in the Loop" Is Breaking as AI Agents Take Over
Researchers argue on arXiv that today's AI agents not only hamper human oversight but actively erode the very skills that oversight requires.
-
Tesla workers reluctant to train Optimus robots meant to replace them
Tesla employees are balking at helping to train the company's Optimus humanoid robots, aware that they are preparing their own replacements. Even so, Tesla still aims to build up to a thousand of the machines per week by the end of 2026.
-
Anthropic to spend $11.6B on Akamai's cloud — and get a stake in return
Anthropic has committed $11.6 billion over seven years to Akamai's cloud infrastructure, and in return gets the right to buy up to 5% of the supplier — an unusual inversion of the typical "investor bets on customer" pattern.
-
Thousands of Supabase databases are leaking personal data to the open web
Security firm UpGuard found around 16,000 databases hosted on Supabase where misconfigurations left people's personal information exposed to the public internet.
-
Anthropic’s AI Misuse Report, Distilled Into 117 Findings
Anthropic released a long report on how Claude has been misused, and Daniel Meissler boiled it down to 117 findings. Bruce Schneier highlights the key takeaways.
-
Australia says an OpenAI agent broke into a government website
Australian officials say an OpenAI AI agent gained unauthorised access to a health-statistics portal back in June — possibly the first known case of an AI agent hacking a government site.
-
Meta Connect 2026: VR glasses replace the headset, and smart glasses lose the camera
At its annual showcase, Meta unveiled 100-gram VR glasses, camera-free smart glasses, and new features for its Muse AI agent. The headline product ships in spring 2027.
-
CLOSEDQUORUM: the malware that lets four AI models vote on its next move
Cisco Talos has detailed a Windows implant that, instead of taking orders from an attacker's server, asks several commercial AI services what to do and carries out whichever action wins the most votes.
-
A New Price War: Anthropic and OpenAI Slash Their Flagship Rates
In a single day came Claude Opus 5.5, GPT-6 Sol and GPT-6 Luna — all against a backdrop of tumbling prices. GPT-6 is half the cost of its predecessors, and Opus 5.5 dropped 20%.
-
Critical Bifrost AI Gateway Flaw Allows Command Execution Without Credentials
A critical flaw in Bifrost, the open-source AI gateway that routes requests to over 20 LLM providers, lets an unauthenticated attacker run arbitrary commands on the gateway server with a single HTTP request.
-
systemd 262 lands with embedded unit files, LUO integration and an AI “canary” for unreviewed code
The systemd project has shipped another major update to its init system and service manager. Highlights include embedded fallback unit files, Live Update Orchestrator support, and a mechanism to flag unreviewed code written by AI models.
-
GPT-6 Astra Cracks an Old Enigma Message on Its Own
With no step-by-step guidance, a language model broke one of the unsolved Enigma intercepts — writing its own Enigma simulator and Bombe along the way.
-
Jev: the LLM that returns a number instead of text
TypeSafe AI has unveiled Jev, a new kind of model dubbed a "decision model": text goes in, and a floating point number with a confidence score comes out.
-
Xiaomi Unveils MiMo v2.6
Xiaomi has published a new release of its MiMo AI model — version 2.6. The announcement drew attention on Hacker News.
-
Amazon blocks Meta's Muse AI agent from its store
The retailer has cut off the Muse assistant that shopped on users' behalf: Meta never told Amazon it was coming, and the agent, Amazon says, browsed without identifying itself and appeared to capture customer credentials.
-
AI Chatbots Get Financial Questions Wrong 'Most of the Time'
According to the Financial Times, AI-powered chatbots give incorrect answers to financial queries 'most of the time.'
-
Google releases AX, an open orchestrator for AI agents
The company has published AX, a declarative control plane for running AI agents at a scale of up to billions of tasks per cluster. Grown out of Google research, it looks like Kubernetes but is tuned for a new kind of workload.
-
'Everyone Just Talks to Claude': An Engineer's Anonymous Complaint from a Big Company
Simon Willison quoted an anonymous developer describing life at a large company where everything — from specs to reports — is produced by Claude Code, while people spend 12–13 hours a day merely pressing Enter.
-
The dark forest of world models: why the labs stay quiet
World-model startups have raised heaps of cash and hype, yet no one — not even their own data suppliers — can get them to say what they are actually building.
- More