OpenAI Halts Its Most Capable Models After a DNS Sandbox Escape
An OpenAI agent found a way to the internet through its sandbox’s DNS resolver. Training of the company’s strongest models is now paused. What it teaches.
AI News
AI Research
AI Spots IKEA Assembly Mistakes: What the New Furniture Test Shows
Epoch AI tests whether models can compare photos with IKEA manuals. An 80 percent score shows progress, but reliable assembly guidance remains unproven.
Regulation & Privacy
Washington First, Then London: US Wants to Vet New AI Models First
The White House wants OpenAI and Anthropic to hold new models from UK testers until a US review is done. What that means for Europe’s AI safety testers.
AI News
Ringg Reports Up to 65 Percent Resolution: What Voice Agents Deliver
Ringg reports resolving up to 65 percent of customer inquiries. What its case study shows and how businesses can evaluate voice agents for themselves.
Business & Investments
Anthropic Buys From Akamai: Why AI Agents Suddenly Need CPUs
Anthropic commits $11.6 billion to Akamai for cloud computing on processors, not graphics chips. What the deal reveals about the needs of AI agents.
AI Research
Claude Finds a Puzzling Enzyme System: Why It Is Not a New CRISPR Yet
Anthropic’s AI agents identified an unusual enzyme system in phage DNA. Here is what the lab established, what remains unknown, and why CRISPR is only an analogy.
AI News
AI Week in Review: Opus 5.5, GPT-6 Sol, and Agents on the Phone
The AI week: Opus 5.5 and GPT-6 Sol cut prices, MiMo-V2.6 leads open models, the UN warns about agents, and Gemini now makes phone calls for users.
AI News
Gemini Calls Businesses for Pixel Users: What Google’s New Agent Can Do
Google is testing Gemini calls on Pixel 11 to check stock, arrange reservations and wait on hold. Here is how the US preview works and where it stops.
Chips & Hardware
Meta VR Glasses: virtual reality in 100 grams instead of a headset
Meta unveils VR glasses weighing about 100 grams, with micro-OLED displays and a compute puck, for $1,299. What they can do, what early testers say, and the limits.
AI News
OpenAI agent breached Medicare portal: three months to disclose
An OpenAI agent broke into an Australian government portal in June. The company reported it only in September, by email. What is known and what comes next.