SAIS '27 — USENIX Conference on Secure Agentic-AI Systems

SAIS '27, the USENIX Conference on Secure Agentic-AI Systems, runs 7-8 June 2027 at the Hyatt Regency Santa Clara in Santa Clara, California. It is USENIX's new venue for the security and privacy of AI agents — systems in which models reason, plan and act on a user's behalf rather than simply answering questions. The programme is built from peer-reviewed papers on designing, verifying and deploying secure, privacy-preserving agents into real-world systems, with paper submissions due 4 February 2027, author notification on 16 March and final papers on 20 April. Its steering committee draws on DeepMind, Google, NVIDIA, ETH Zurich and university research groups, and the conference sits in the same USENIX calendar as the Security Symposium and PEPR — USENIX PEPR '27 runs on the same two days in the same city.

• Paper submission deadlineFeb 4, 2027
• Notification to authorsMar 16, 2027
• Final (camera-ready) papers dueApr 20, 2027

Date

June 7–8, 2027

Location

Hyatt Regency Santa Clara, California

Format

In-Person

Event Type

Conference

Status

Upcoming

Pricing

Paid

Organizer

USENIX Association

Added

Sep 16, 2026

Updated

Sep 16, 2026

• ✓Security researchers
• ✓CISOs
• ✓AI platform engineers
• ✓Machine learning engineers
• ✓Privacy engineers
• ✓Enterprise security architects

Agent deployments are where enterprise AI risk is concentrating: prompt injection, tool abuse, over-broad credentials, and actions taken with no human in the loop. SAIS '27 is the first academic venue devoted entirely to that problem, so a CTO, CISO or head of AI platform leaves with peer-reviewed threat models, evaluation methods and defences they can take straight into their own agent architecture review — rather than a vendor deck. Two focused days, and direct access to the researchers from DeepMind, Google, NVIDIA and ETH Zurich who are setting the field's agenda before it hardens into product requirements.

• ✓Peer-reviewed research track on designing, verifying and deploying secure AI agents in real systems
• ✓Scope covering agents that reason, plan and act — not chat-only model safety
• ✓Privacy-preserving agent deployment as a first-class topic area
• ✓Two-day single-venue programme at the Hyatt Regency Santa Clara
• ✓Programme shaped by a steering committee spanning DeepMind, Google, NVIDIA and ETH Zurich
• ✓Runs alongside USENIX PEPR '27 (7-8 June 2027, Santa Clara)

• ★USENIX's first conference dedicated solely to securing agentic AI, rather than an AI track bolted onto a general security programme
• ★Peer-reviewed and archival — a full submission, notification and camera-ready cycle, not a sponsored-talk agenda
• ★Steering committee drawn from DeepMind, Google, NVIDIA and ETH Zurich alongside academic research groups
• ★Run by USENIX, whose Security Symposium and PEPR set the tone for practitioner-facing security research
• ★Scheduled on the same two days and in the same city as USENIX PEPR '27, so one trip covers agent security and privacy engineering

USENIX had not published SAIS '27 registration rates as of September 2026 — the conference page currently carries the call for papers and its deadlines only. Registration normally opens closer to the event on the same conference page.

When is SAIS '27?

SAIS '27 takes place June 7–8, 2027 in Santa Clara, California.

Where is SAIS '27 held?

SAIS '27 is held in Santa Clara, California. The event format is in-person.

What is the SAIS '27 submission deadline?

The next SAIS '27 deadline is paper submission deadline on Feb 4, 2027. Remaining dates: Notification to authors — Mar 16, 2027; Final (camera-ready) papers due — Apr 20, 2027.

How much does SAIS '27 cost?

SAIS '27 is paid. USENIX had not published SAIS '27 registration rates as of September 2026 — the conference page currently carries the call for papers and its deadlines only. Registration normally opens closer to the event on the same conference page.

Who should attend SAIS '27?

SAIS '27 is aimed at Security researchers, CISOs, AI platform engineers, Machine learning engineers, Privacy engineers, Enterprise security architects.

What topics does SAIS '27 cover?

SAIS '27 covers Agentic AI security, AI agents, Prompt injection and agent abuse, LLM security, Privacy-preserving machine learning, Verification of AI systems, Trustworthy autonomy, Secure agent deployment.

This page was written from 3 sources, 1 on domains other than usenix.org.

1. 1.usenix.org — sais27vendor
2. 2.usenix.org — conferencesvendor
3. 3.sec-deadlines.github.io — sec-deadlines.github.io