Sol sets a new state-of-the-art score on Terminal-Bench 2.1, while Luna is positioned as the cheapest tier.
Category: Technology & AI
xAI’s Grok 5 still months away as Colossus 2 training continues
Prediction markets closed June contracts pricing just a 3% chance of a near-term Grok 5 release.
UN Global Dialogue on AI Governance set to open in Geneva
Member states will discuss international approaches to managing AI technology starting July 6.
Google rolls out Nano Banana 2 Lite and Gemini Omni Flash media models
The new generative image and video models are priced per-output-unit and available via Google AI Studio.
SpaceX and Anthropic IPOs Mark AI’s Transition From Venture Capital Experiment to Public Market Asset
The emerging wave of AI company public listings starting with SpaceX is bringing new accountability transparency demands and quarterly earnings pressure to frontier AI labs.
Microsoft MAI-Thinking-1 Model Aims to Match Premium Reasoning Quality at Highly Competitive Token Cost
Microsoft’s new in-house reasoning model announced as part of the MAI model suite targets enterprise customers who need strong logical capabilities without frontier model pricing.
Agentjacking Attack Class Exploits AI Coding Agent Trust — Security Researchers Issue Enterprise Warning
A newly identified attack category that injects malicious instructions into AI agents via compromised tool output is creating a significant security surface for organisations.
Pentagon Testing OpenAI and Google Models as Potential Replacements for Anthropic Claude in Classified Systems
The US Defense Department has confirmed it is evaluating alternatives to Claude for classified use cases as AI companies compete intensely for government security contracts.
Alibaba’s 25,000 Fake Anthropic Accounts Extracted AI Capabilities Over 6 Weeks — US Senate Now Briefed
Anthropic’s formal accusation letter sent to Senate Banking Committee leaders provides detailed technical evidence of the unprecedented corporate AI espionage campaign.
Anthropic Accuses Alibaba of Running Largest AI Distillation Attack in History Using 28.8 Million Claude Interactions
Anthropic’s Senate letter alleges 25,000 fraudulent Alibaba accounts generated 28.8 million Claude exchanges between April and June 2026 to train Alibaba’s Qwen AI model.