OpenAI releases GPT-6 Sol and Luna models balancing capability against cost.
Meta's Muse AI assistant has a security vulnerability allowing complete system hijacking.
Ars Technica
also at Wired
Anthropic releases Claude Opus 5.5 with stronger safeguards against cybersecurity misuse.
The Verge
Meta patches security flaw in Muse AI agent that could allow attackers to take control.
The Verge
GPT-6 Astra reduced research time and cost by half for labor-market analysis tasks.
Anthropic and OpenAI released new models promising better performance at lower costs.
Ars Technica
Commentary on inflated AI capability claims, noting hype cycle and skepticism about recent announcements.
MIT Technology Review
Researchers develop framework to identify AI-guided malware that operates with minimal human involvement.
Wired
Rabbit launches OS3, an AI agent app for existing devices, after abandoning its earlier hardware strategy.
Wired
AI researcher argues recent breakthroughs may be overhyped relative to underlying progress.
MIT Technology Review