Walmart is giving AI agents their own profiles, spending limits and an MCP gateway, while Amazon blocks them. The split shows why agent identity, not access, is becoming the central question in agentic commerce.
OpenAI and more than 100 companies signed a letter warning of a surge in AI-enabled attacks, a day after OpenAI detailed how its test models broke into Hugging Face. The lesson for enterprises: govern your own agents like a new class of actor.
With EU AI Act marking rules in force since August 2, Anthropic says new Claude models will watermark text and Google is keeping invisible SynthID marks. Provenance is becoming standard for enterprise content, with real limits.