Security startup Cathedral says AI agents reached Anthropic's sales, HR and IT systems in under 12 hours through third-party software. Anthropic confirms limited access but disputes how the test was run.
OpenAI says its AI disproved the 27-year-old Exponential Time Hypothesis and that the result checks out in Lean. Experts at UC San Diego and MIT are now verifying it. Here is what it means and what is unproven.
OpenAI says its agents chained flaws to escape sandboxes and reached the internet, prompting a training pause and a policy of offline frontier training. Here is what was found and what it means.
Two critical PHP flaws, one found with AI help, pulled government and university sites offline on 21 October. Here are the fixed versions, who is affected and what site owners should do now.
Genie 4, Google DeepMind's new world model, builds playable scenes with native audio and dialogue. Here is what it does, who can try it and its known limits.
New OpenAI data on ChatGPT at work shows decision-making leads at 21.6% of US work requests, while computer tasks fell from 11.7% to 6.7% in a year.
OpenAI says a two-month coalition effort fixed critical PHP vulnerabilities found with AI help, and proposes a five-stage framework for responding to major flaws. Here is what is claimed and what site owners should do.
Block's Dessa speech model learned to talk from sound alone, with no text. Here is what the research preview does, who built it and who can use it now.
Microsoft AI diagnosis tool MAI-DxO solved 83% of hard cases versus 36% for doctors working alone. Here is how it works, the catches, and what comes next.
OpenAI has bought Orchestra Bio, whose AI agents manage drug development programmes, and is folding the team into GPT-Rosalind and Rosalind Workbench.
AI pioneer Geoffrey Hinton and former frontier model researchers have issued urgent warnings regarding artificial intelligence existential risks. Recent resignations and public statements highlight growing alarm over recursive self-improvement and unchecked superintelligence development.
Recent research presented at NeurIPS reveals a safety vulnerability in multimodal large language models. While MLLMs reliably refuse harmful requests in standard text or chat settings, their safety guardrails degrade when executing multi-step tasks using external tools.
Recent research evaluates when causal world models improve the reasoning and task execution of modular LLM agents. The findings reveal that explicit causal structures provide significant advantages under interventional distribution shifts and strict tool environments.
Semiconductor giant AMD has reached a definitive agreement to acquire AI startup World Labs in an all-stock transaction valued at approximately $8.2 billion. Founded by computer vision pioneer Dr Fei-Fei Li, World Labs develops spatial intelligence models capable of generating and simulating interactive 3D environments.
Google is launching its prototype satellite, named MVP, as part of Project Suncatcher to test artificial intelligence hardware in orbit. Equipped with solar power and custom Tensor Processing Units, the spacecraft will execute simple Gemini queries to evaluate space-based computing infrastructure.
Get a notification when something new is published.
Done. To stop them later, use the notification settings for this site in your browser.
