2026​
- August 5 - What the 'rogue agent' headlines got wrong
- August 5 - OWASP published a security top 10 for MCP
- August 5 - The EU AI Act deadline moved. The obligations didn't.
- August 5 - Agent memory just became its own product category
- August 9 - NVIDIA shows KV caches can move between different-sized models
- August 12 - A new toolkit unifies three ways to catch a hallucination
- August 12 - OpenAI's cyber-hunting model just found a real Chrome vulnerability
- August 12 - Every major AI lab shipped an agent model this week
- August 16 - OWASP checked its LLM Top 10 against real incidents
- August 17 - Managed AI platforms trade one lock-in for another
- August 19 - Frontier models don't make safer agents
- August 23 - MCP just went stateless
- August 24 - MCP is done pretending agents are people
- August 31 - Stop calling LLM APIs directly
- September 4 - GPT-6 Astra: the harness is the product
- September 7 - AGENTS.md: one guide for your coding agents
- September 13 - Agent security starts before the first prompt