Musk Drops Apple From Antitrust Suit, Keeps OpenAI in Crosshairs
A revised antitrust complaint removes Apple as a defendant but leaves OpenAI facing Musk's claims over the ChatGPT integration.
AI & ML
Frontier models, research results, agents, and developer tooling.
OpenAI disrupts coordinated model-distillation campaign
OpenAI says it broke up an attempt to extract protected reasoning from its models and is now hardening against similar attacks.
NVIDIA Opens 2027–2028 Graduate Fellowship Applications With $60,000 Awards
Doctoral students in AI and related fields can apply by October 30, 2026, for grants, mentorship, and a required summer internship at NVIDIA Research.
Google Watermarks AI-Designed Proteins to Curb Biosecurity Risks
New SynthID Bio embeds imperceptible signatures in AI-generated proteins without breaking their function, giving DNA screeners a way to verify trusted designs.
Google's AI search payouts to publishers are minuscule, report finds
A Google pilot paying websites for content used in AI answers is yielding tiny sums for most participants, raising doubts about the program's future.
Fireworks AI's Ember-1 Cuts Token Use by 40% via Shorter Reasoning Traces
A post-trained Kimi K3 model from Fireworks AI reduces output tokens by roughly 40% while keeping reasoning effort intact.
Alibaba Qwen's New Realtime Voice Model Decides When to Speak
Qwen-Audio-3.1-Realtime is a full-duplex voice model that reasons, calls tools, and judges turn-taking, sharply reducing interruptions from background speech.
Liquid AI's d1 Returns Calibrated Probabilities Without Generating Tokens
Liquid AI's d1 is a decision model that outputs calibrated probabilities for structured choices in a single call, using zero generated tokens.
AI's Trillion-Dollar Investment Boom Faces Uncertain Economic Future
The massive build-out of AI infrastructure is a high-stakes economic wager, and researchers are struggling to model its potential payoffs.
AEF-1 standard for third-party AI evaluators gains backing from three labs
A new evaluation framework, AEF-1, is drawing support from Xai, OpenAI, and Anthropic, signalling a push toward shared external testing norms.
Unitree's cost-obsessed founder built cheap humanoids—can he scale?
A profile of Wang Xingxing shows how relentless cost-cutting made Unitree a leader in affordable humanoid robots, but raises questions about whether that style scales.
AI Agents Timmy, Ren, and Jackie Spam Social Media With Slop
A new wave of self-described AI agents is flooding platforms with low-quality automated posts, raising questions about content quality and platform moderation.
Agility Humanoid Robot Stops and Squats to Work Safely Beside People
The new robot's stop-and-squat behavior lets it operate around human coworkers without physical cages or safety barriers.
Titanic-Themed Art Protest Targets OpenAI Event in San Francisco
A Bay Area studio's sculpture at an OpenAI developer event criticizes AI's unequal costs and the rise of algorithmic art.
Firefox chief links redesign to AI skepticism and open web
Mozilla's Ajit Varma discusses how a browser redesign and a cautious approach to AI fit into the fight for the open web.
ProvenanceGuard Checks Whether MCP Agents Cite the Right Source
A new verification layer for MCP-based agents catches claims that are true somewhere in the evidence but attributed to the wrong source.
NVIDIA Kumo Tabular: A Foundation Model for Tabular Prediction
NVIDIA's open Kumo Tabular model predicts labels on new rows in one forward pass, without training or feature engineering.
OpenAI Publishes Early Safety-Case Guidelines for Frontier AI Training
OpenAI has released its first guidelines for building safety cases that justify whether frontier AI models are safe to train and deploy.