Version 1.0 | Last Updated: 09.22.2026
This AI & Automation Addendum (this “Addendum”) is part of the Agreement between Discern and Customer described in the Discern Security Master Subscription Terms at https://www.discernsecurity.com/terms-of-service (the “Terms”). It governs the artificial intelligence features of the Service and all Automated Actions. Capitalized terms used and not defined in this Addendum have the meanings given in the Terms. For matters relating to Automated Actions, this Addendum controls over the Terms, as stated in Section 18.10 of the Terms.
1.1 “Action Class” means a category of Automated Actions grouped by target system and action type and described in the Documentation, such as ticket creation, notification, configuration change, policy change, or access modification.
1.2 “Action Log” means the record described in Section 3.2.
1.3 “Agent” means an AI-driven component of the Service that analyzes Customer Data, generates Outputs, or prepares or executes Automated Actions.
1.4 “Authorization Record” means the Service’s record of Customer’s Autonomy Tier selections and scope configurations, including the identity of the Authorized User who made each selection and the time it was made.
1.5 “Autonomy Tier” means the execution mode that Customer selects for an Action Class under Section 2.
2.1 Tiers. The Service supports the following Autonomy Tiers for each Action Class:
(a) Recommend-Only. The Agent analyzes and recommends. No change is made to any Connected System or to Customer’s environment unless an Authorized User separately carries it out.
(b) Approval-Required. The Agent prepares a specific Automated Action, and the Service executes it only after an Authorized User approves that action in the Service.
(c) Autonomous. The Service executes Automated Actions in the Action Class without per-action approval, within the scope Customer has configured.
2.2 Default; Elevation. At activation, every Action Class is set to Recommend-Only. An Autonomy Tier may be raised only by the affirmative act, recorded in the Authorization Record, of an Authorized User to whom Customer has given administrative rights in the Service. Customer may lower an Autonomy Tier at any time, and the change takes effect promptly.
2.3 Scope Configuration. For each Action Class, Customer may configure the scope within which Automated Actions are permitted, including the Connected Systems, asset or user groups, and thresholds to which the Action Class applies. The Documentation describes the available scope controls.
2.4 Authorization. Customer’s Autonomy Tier selections and scope configurations in the Service are Customer’s standing authorization for the Service to act, and are the sole source of that authorization. The Service is designed not to execute an Automated Action outside the Action Class, scope, and Autonomy Tier that Customer has configured, and Discern shall not instruct it to do so.
3.1 Guardrails. Discern shall maintain technical safeguards for Automated Actions, including validation of each action against the applicable Authorization Record before execution, rate and volume limits, and testing of Action Classes before release.
3.2 Logging. Discern shall record each Automated Action in an Action Log, including the action taken, the target Connected System, the Agent involved, the Authorization Record relied on, the time of execution, and the outcome, and shall make the Action Log available to Customer in the Service for at least [12] months.
3.3 Suspension by Customer. Customer may suspend all Automated Actions, or the Automated Actions in any Action Class, at any time through the Service. Suspension takes effect promptly, and the Service will cancel pending Automated Actions in the suspended scope to the extent they have not begun executing.
3.4 Suspension by Discern. Discern may suspend Automated Actions, in whole or in part, if Discern reasonably believes they are being misused, present a security risk, or are affected by a defect, and shall notify Customer promptly of the suspension and its scope.
3.5 Change Control. Automated Actions operate within, and do not replace, Customer’s own change-management, approval, and compliance processes. Customer is responsible for selecting Autonomy Tiers and scopes consistent with those processes.
4.1 Probabilistic Systems. The Agents use probabilistic, machine-learning systems, including large language models. Their analyses, recommendations, and Automated Actions are generated by statistical inference, may vary for similar conditions, and may be incomplete or incorrect. Within an authorized scope, an Agent may select actions that a human operator would not have selected.
4.2 No Detection Guarantee. Section 13.4 of the Terms applies to the Agents and to Automated Actions, including its statement that Discern does not warrant that the Service will identify, detect, assess, prevent, or remediate every vulnerability, misconfiguration, threat, or risk.
4.3 Oversight. Customer shall maintain human oversight of the Service appropriate to its selected Autonomy Tiers, including regular review of Action Logs for Action Classes set to Autonomous. The Service provides notifications and reporting to support this oversight.
4.4 Model Changes. Discern may update the models and Agents underlying the Service. An update does not expand the Action Classes, scopes, or Autonomy Tiers that Customer has authorized. If an update materially changes the behavior of an Action Class set to Autonomous, Discern shall note the change in its release communications.
5.1 No Training. Sections 6.2 and 6.3 of the Terms govern the use of Customer Data in connection with the Agents: Discern does not use Customer Data to train, fine-tune, or improve any artificial intelligence or machine learning model, except through Aggregated Data as permitted by Section 6.3 of the Terms.
5.2 Foundation-Model Subprocessors. If an Agent uses a third-party foundation model, the provider of that model is a subprocessor under Section 6.4 of the Terms. Discern shall contractually require that the provider not use Customer Data to train its models and not retain Customer Data beyond what is needed to process the request, and shall identify the provider on the subprocessor list.
5.3 Outputs. Section 11.2 of the Terms governs ownership and use of Outputs, including Outputs generated by the Agents.
6.1 Discern. Discern is responsible for (a) performing its obligations under this Addendum, including the safeguards in Section 3; (b) exercising reasonable care in the design, testing, and operation of the Agents; and (c) any Automated Action executed outside the Action Class, scope, or Autonomy Tier that Customer authorized, or executed in breach of this Addendum.
6.2 Customer. Customer is responsible for (a) its Autonomy Tier selections and scope configurations; (b) maintaining the oversight described in Section 4.3; (c) the fit of Automated Actions with its own change-management and compliance processes; and (d) subject to Section 6.1, the outcomes of Automated Actions executed within the Action Class, scope, and Autonomy Tier that Customer configured and authorized, as stated in Section 5.4 of the Terms.
6.3 Limitations. Section 15 of the Terms applies to all liability arising out of or relating to this Addendum.
7.1 Customer Compliance. Customer is responsible for determining whether and how its use of Automated Actions is permitted under the laws and regulatory obligations that apply to Customer. If a law applicable to Customer requires human oversight of automated decisions, Customer is responsible for selecting Autonomy Tiers that satisfy it.
7.2 Documentation Support. On Customer’s reasonable request, Discern shall provide documentation about the Service’s AI features, oversight controls, and logging that Customer reasonably needs to meet its own regulatory obligations with respect to its use of the Service.
8.1 Updates. Section 17 of the Terms governs updates to this Addendum. Discern shall post each updated version at https://www.discernsecurity.com/ai-automation-addendum with a new version number and “Last Updated” date and shall maintain an archive of prior versions.
