Building trust infrastructure for agentic AI: the governance gaps exposed by OpenClaw

Chris Riley, executive director of the Data Transfer Initiative, argues for building a trust infrastructure for agentic AI, saying the OpenClaw-style open agent ecosystem sits atop a failed governance system. After researchers examined OpenClaw's skill repository ClawHub and broader agent repositories, they found 12%-20% of skills to be malicious, rather than merely unsafe. Cyera says the risk comes not from a single vulnerability but from the collapse of data governance boundaries across the entire AI agent lifecycle, with a Cisco security researcher calling it a "security nightmare."