By Hamza Faisal

SentinelOne has unveiled two new offerings at its OneCon 2026 customer conference in Las Vegas: Singularity Agentic Detection and Response (ADR), available now in early access, and Singularity ASOC, an autonomous security operations center suite powered by what the company calls Purple Security Superintelligence. The announcements double down on a thesis CEO Tomer Weingarten laid out in his keynote — that powerful frontier AI models have fundamentally broken the old assumptions of cybersecurity.

For a decade, security teams planned around two premises: the attacker is human, and defenders have hours to respond. Weingarten argued both are dead. In the span of a year, AI has moved from assistant to actor inside the enterprise: agents now hold real credentials, chain tool calls, and work independently across company systems. Attackers are using the same frontier models to find and chain vulnerabilities faster than human analysts can react, and high-profile breaches caused by rogue or unintended agentic behavior are becoming commonplace.

Stop the agent, not the workflow

Singularity ADR discovers AI agents across the enterprise — sanctioned or not — enforces what each one is allowed to do, monitors agent behavior in depth, and detects and stops unintended or malicious activity in real time. SentinelOne says it ties every agent action to its origin, so a security team can see what an agent was asked to do next to what it actually did — and can stop an agent mid-action without halting the surrounding workflow.

The offering builds on agent discovery and governance the company introduced in March, but shifts the focus to runtime: behavioral prevention, detection, and response applied directly to agentic actions. It runs on the company's Singularity Platform, covering endpoints, cloud workloads, and the broader enterprise where AI agents now run.

An autonomous SOC in a box

Alongside ADR, SentinelOne unveiled Singularity ASOC — an integrated suite combining data ingestion and analytics, agentic investigations, and automated response. Its Purple Agentic Investigation uses native and third-party security data to investigate alerts end to end and gather evidence toward a verdict, while an AI SIEM, hyperautomation, threat intelligence and IOC management, and AI Data Pipelines with more than 50 third-party connectors handle the rest. "Singularity Credits," with administrator-set guardrails, keep automated actions inside customer-defined limits. The company claims the system can shrink forensic analysis from minutes to milliseconds — though Weingarten stressed the systems stay under the command of human analysts.

The agent-versus-agent arms race is on

The timing is pointed. Security teams have watched AI systems identify and exploit vulnerabilities faster than traditional defenses can respond — and recent incidents, from rogue AI agents to the widespread exploitation of infrastructure flaws, have left security teams on edge. SentinelOne also expanded its PartnerOne program so partners can build and scale AI Security practices on the Singularity Platform.

Weingarten's framing of the moment was blunt: security teams can no longer assume attackers are human or that defenders have hours. The agent-versus-agent era of cybersecurity has officially begun — and SentinelOne wants to be the company selling both the agents and the fences.

By Hamza Faisal