CAIS 2026 Opens: First Academic Conference Dedicated to Agentic AI Systems
The first academic conference dedicated entirely to agentic AI systems opens today in San Jose. The ACM Conference on AI and Agentic Systems (CAIS 2026) runs May 27-29 and is co-chaired by Two Sigma’s Heather Miller.
The timing is notable. OpenClaw’s explosive growth — 368K GitHub stars, 12M downloads, enterprise deployments at Fortune 500s — has outpaced the academic community’s ability to rigorously study its security surface. The Claw Chain CVE advisory this week, backed by an arXiv paper (2605.23330v1), was the first systematic academic analysis of OpenClaw’s attack surface. CAIS 2026 is expected to produce more.
The conference covers topics directly relevant to OpenClaw’s current challenges: sandbox isolation, privilege escalation paths, credential store security, and governance frameworks for autonomous agents. NVIDIA’s NemoClaw — which adds formal guardrails to OpenClaw deployments — was positioned at GTC in March as an enterprise answer to exactly these concerns, but the academic community is now examining the underlying problems with more rigor.
For OpenClaw’s development team, CAIS 2026 outputs are likely to inform the security and governance roadmap over the next release cycle. With the project under active attack (see: Claw Chain CVE advisory, this week’s patch), the timing of a concentrated academic response is either very good or very overdue, depending on how you look at it.