AI SECURITY
Prevent out-of-policy agent actions. Varonis Atlas detects when an agent drifts from its intended task and takes action in real time.
Trusted by 8,000+ customers
OVERVIEW
The question is no longer “Can the user access this?” Now the question is "Can this agent, in this context, for this purpose, using this data, take this action, right now?" Agent Intent answers that question for every action, so agents stay on task.
Atlas can detect an agent's purpose automatically. Add an agent manifest for agents that interact with sensitive data or powerful tool by design.
Atlas can detect an agent's purpose automatically. Add an agent manifest for agents that interact with sensitive data or powerful tool by design.
Atlas compares every tool call to the agent’s intent and decides in real time: allow it, or block it.
Atlas compares every tool call to the agent’s intent and decides in real time: allow it, or block it.
Write down each agent’s purpose, owner, data boundaries and the actions it must never take, in plain language.
Write down each agent’s purpose, owner, data boundaries and the actions it must never take, in plain language.
Focus on the permissions and exposed data that could increase the impact of a breach.
Bring data discovery, access visibility, and activity monitoring together to support your compliance program.
DETECTION
Atlas uses AI and contextual analysis to classify the intent behind every agent action, then flags anything that strays from the agent’s purpose. Tune the sensitivity to fit each agent.
Manifest
The Agent Manifest is a declared record of each agent’s purpose and limits, set by its author. Anything outside it shows up as a deviation, so you can prevent drift before it happens.
GUARDRAILS
Alert, block, modify, log or require human-in-the-loop approval on every intent-based detection. Show a custom message to the agent or user, and quarantine when needed.
INVESTIGATIONS
Review agent intent, agent actions, and runtime guardrails across the entire session. Answer the questions auditors will ask: who was this agent acting for, what was it supposed to do and what did it actually do?
FAQS
Agent Intent has 2 parts. The Agent Manifest declares what each agent is for and what it may do. Agent Intent-Based Access Control (IBAC) checks every tool call against that intent at runtime and allows or blocks it.
Static permissions answer whether a user can access something. IBAC asks whether this agent should take this action, for this purpose, with this data, right now. It blocks calls that exceed data scope or volume limits, perform prohibited actions or send data outside approved destinations.
No. Atlas can infer an agent's purpose automatically and judge each call against it. Use the Agent Manifest for high-risk agents, or agents with powerful tools or sensitive data access by design, to make intended behavior explicit.
Ownership and context, operating expectations like autonomy level and human oversight, the expected workflow, data boundaries and prohibited actions, and runtime identity expectations such as per-user delegated OAuth.
You decide. Policies can alert, block, modify, log or require human approval. You can show a custom message to the agent or user and optionally quarantine.
Each agent gets a declared, reviewable record of its purpose, owner, data boundaries and oversight model. Every allowed or blocked action is logged as evidence, tied to a real user and agent.
Agent Intent is part of Atlas AI Detection and Response, alongside Browser and Endpoint Protection and the AI Gateway. Intent violations appear in the same timeline as prompts, responses and tool calls, correlated with identity and data context.
