ClickHouse Joins The Open Secure AI Alliance
ClickHouse, a leader in real-time analytics, data warehousing,
observability, and AI/ML, has announced that it is joining the Open Secure AI
Alliance alongside NVIDIA and other industry leaders to help build open tools
that keep AI agents secure.
Last month, an AI agent escaped its sandbox and gained access to
Hugging Face's production servers. Containing the intrusion meant
reconstructing what the agent actually did: reviewing more than 17,000 actions
with an open-weight model running on Hugging Face’s own infrastructure, after
safety guardrails on hosted closed models blocked their first forensic
attempts. The incident became one of the founding lessons of the Open
Secure AI Alliance which builds and shares open models, harnesses, tools, and
techniques so defenders can inspect, adapt, and run AI security systems on
their own infrastructure.
ClickHouse is joining alongside NVIDIA and other industry
leaders to help build open tools that keep AI agents secure.
NVIDIA’s announcement of the Open Secure AI Alliance names the
full agent stack that security depends on: identity, permissions, harnesses,
guardrails, logs, and evaluation.
Leave A Comment