Daily AI CatchupAI Catchup — Monday, August 10, 2026
Monday, August 10, 2026Safety
OpenAI Pauses Astra Over Critical Cybersecurity Risks
OpenAI halted development of its Astra model after internal safety evaluations flagged it as critically dangerous for autonomous exploit development.
AgentsClaude Code Auto Mode Becomes Default; Catches Dangerous Commands 89% of the Time
Anthropic makes Claude Code auto-mode the default for Pro/Max/Team users on Aug 14, with safety mechanisms blocking risky commands at an 89% success rate.
AgentsClaude Code Sessions Can Now Message Each Other Without Human Intermediaries
Claude Code introduces cross-session messaging, letting AI sessions communicate directly to coordinate work and unblock problems without user intervention.
RoutingCursor Router Learns to Route Tasks to the Right Model Based on Real Developer Work
Cursor Router classifies developer tasks using learned taxonomy and routes to efficient or frontier models based on what actually works on real code.
Open-SourceMeta Releases Muse Glimmer, a 30B Open Agent That Runs Offline
Meta open-sources Muse Glimmer, a 30B parameter agent capable of running locally without cloud infrastructure, democratizing agentic AI.