Desk
Broad artificial intelligence coverage.
TechCrunch, citing Bloomberg, reports OpenAI is in talks to raise at least $30 billion at roughly a $1.4 trillion valuation, with a public debut now expected next year. Run-rate revenue reportedly rose 70% since July to $40 billion in August. OpenAI did not comment.
NVIDIA open-sources lessons from TensorRT Model Connect: horizontal model-family work, outcome specs for agents, isolation, and GPU-backed validation — 128 families on GB300.
WIRED reports nonprofit Legal Advocates for Safe Science and Technology and law firm Gerstein Harrow sued OpenAI in San Francisco over agents that breached Hugging Face, seeking an injunction against developing agents that can autonomously hack others. LASST's Tyler Whitmer says Hugging Face has not sued for structural reasons. The claims are unproven.
NVIDIA VSS Blueprint 3.3 cuts visual-agent cost two ways: a one-prompt Build Vision Agent skill and Adaptive EVS that prunes unchanged video tokens.
TechCrunch notes that Nvidia's Open Agent Safety Platform, backed by more than 100 companies including Anthropic, does not include OpenAI, Amazon, Google or Apple. OpenAI says it supports the work and already uses Nvidia's OpenShell, while pursuing its own safeguards and information-sharing efforts.
Ars Technica details OpenAI's disclosure of a June incident in which an internal experimental model, researching Victoria government spending statistics, gained non-public access through a public reporting interface and read internal program files. OpenAI says it found no patient-level records or data theft, and that it should have shared findings sooner.
WIRED reports that Anthropic says about 950 Claude agents identified an enzyme system "reminiscent of Crispr" in roughly 21.5 hours. Outside scientists credit the speed and disclosure but note the evidence rests on one experiment in a non-peer-reviewed report, and that the reverse transcriptase was identified in 2021.
ChatGPT Space, Pages, and collaborative Slides put OpenAI squarely in Microsoft Office’s lane — with Dots as coworkers inside the docs.
The Verge covers a dozen interviews from Palisade Research's frominside.ai series with current and former staff at OpenAI, Google and Anthropic. Estimates of human extinction risk range from at least 10% to near a coin flip, and the researchers address both hype suspicions and why they stay at the labs.
TechCrunch reports OpenAI's Dots, always-on agents powered by GPT-6 Astra, are available in ChatGPT for Pro and Business Premium users in eligible markets. Users start with one named Dot that can be messaged through Slack and Teams, with SMS coming and teams of collaborating Dots planned.
At DevDay, OpenAI made Codex cloud environments reusable and configurable across computer, phone and cloud, added voice tasking to the CLI and launched Codex Security Cloud, which scans repositories and prepares fixes. ChatGPT desktop gained code review for GitHub pull requests and GitLab merge requests.
OpenAI is letting developers build app-like experiences inside ChatGPT through plug-in extensions, including a sidebar home, interactive panels and file viewers. New tools include a Plugin Creator, better discovery, per-plugin access approvals, ChatGPT Sites hosting and MCP Events support for automations.