1. Overview

When data is not handled carefully, Artificial Intelligence tools can quietly expose information that {Company Name} never intended to share. Information typed into an Artificial Intelligence tool can be retained, used to train a model, or exposed outside the control of {Company Name} without anyone realizing it happened.

Since this policy was first written, federal and state regulators have moved quickly to govern Artificial Intelligence tools, adding new transparency, disclosure, and risk assessment requirements for Artificial Intelligence systems used in important decisions about people. The legal and reputational consequences of careless use of Artificial Intelligence tools, including the disclosure of trade secrets, client data, or biased decision-making, continue to grow.

This policy exists to protect {Company Name}, its workforce, and its clients by establishing requirements for the responsible, secure, and legally compliant use of Artificial Intelligence tools in business operations.

2. Purpose

The purpose of this policy is to establish the acceptable use of Artificial Intelligence tools at {Company Name}. This policy protects {Company Name} by preventing the unauthorized disclosure of Protected Information into Artificial Intelligence systems and by ensuring Artificial Intelligence tools are used accurately, ethically, and in compliance with applicable law.

3. Scope

This policy applies to all employees, contractors, consultants, temporary workers, and other workforce members of {Company Name} who use Artificial Intelligence tools to conduct {Company Name} business or to interact with {Company Name} business systems, whether the Artificial Intelligence tool is owned, leased, or provided by {Company Name}, an employee, or a third party. This policy applies to all equipment owned or leased by {Company Name}. Each individual is responsible for exercising good judgment regarding the appropriate use of Artificial Intelligence tools in accordance with this policy, {Company Name} standards, and applicable law. As used throughout this policy, “Artificial Intelligence tools” includes Artificial Intelligence systems, software, models, agents, agentic systems, agent platforms, and connectors, regardless of the degree of autonomy involved, whether or not a workforce member is actively directing the tool at the time it acts. Exceptions to this policy are documented in section 5.2.

4. Policy

4.1 General Policy Statement

4.1.1 The use of the {Company Name} electronic systems, including all forms of Artificial Intelligence technology, is for company business and other authorized purposes only.

4.1.2 The use of Artificial Intelligence capabilities embedded by the manufacturer within previously approved software on the Authorized Applications List is permitted without additional review, provided the embedded capability operates solely within the scope of that software and does not connect to external systems, services, or data sources beyond what the software was originally approved to access. Embedded Artificial Intelligence capabilities that introduce connectors, plug-ins, or agentic functionality enabling the software to access external systems, services, or data beyond its originally approved scope must be treated as a new connector under section 4.6.3 of this policy and require review and documentation by the Artificial Intelligence Oversight Team before use.

4.1.3 The use of third-party Artificial Intelligence applications, plug-ins, or connectors requires the same approval process required by any other software installation and use, as outlined in the Software Installation Policy, Software Usage Policy, and Authorized Applications List. This approval process includes, but is not limited to, approval to purchase licenses, justification for business use, and listing authorized users. An Artificial Intelligence application, plug-in, or connector that is used as a cloud service and installs no software on a {Company Name} device must be approved in the same way and listed on the Authorized Cloud Services List.

4.1.4 The approved use of Artificial Intelligence tools must be in accordance with all other security policies, including all provisions of:

Internet, Email, and Computer Use Policy, including, but not limited to:

• no inappropriate conduct;
• no unlawful or malicious activities;
• no violation of the laws and regulations of the United States or any other nation, or any state, city, province, or other local jurisdiction;
• no expectation of privacy;
• compliance with software license agreements;
• individual employee liability for any and all damages incurred as a result of violating company security policy, copyright, and licensing agreements;
• employee conduct provisions, including but not limited to: intellectual property, confidentiality, company information dissemination, standards of conduct, misuse of company resources, and information and data security.

Acceptable Encryption Policy:

• the use of encryption standards, and encryption for storage and transmission of data.

Electronic Data Protection Policy:

• use of authorized data storage locations;
• the prevention of unauthorized access, distribution, or data exfiltration;
• external and removable media devices.

Social Media Policy:

• governing the actions of individuals authorized to post content for {Company Name} on social media.

Data Classification Policy:

• requirements for classifying data as Public, Confidential, or Restricted, and the handling requirements that follow from that classification.

4.1.5 Ownership of results produced by Artificial Intelligence tools, and responsibility for their accuracy and consequences, are governed by section 4.3 of this policy.

4.2 Data Protection and Access Controls

4.2.1 No Restricted or Confidential data, as defined in the Data Classification Policy, is to be entered into a publicly accessible Artificial Intelligence interface or database without review and authorization from the Artificial Intelligence Oversight Team, in accordance with the Data Loss Prevention requirements of the Electronic Data Protection Policy.

4.2.2 Access to any account enabling the use of Artificial Intelligence software must comply with all provisions of:

Password Policy:

• a separate, unique password for each account, no shared use of passwords or accounts, and mandatory Multi-Factor Authentication;
• passwords used in company systems must be created and stored in the Password Manager software approved by {Company Name}.

Password Construction Policy:

• passwords must be properly constructed according to that policy to create secure passwords.

4.2.3 Where an Artificial Intelligence tool or connector authenticates using an API key, access token, or other credential instead of a username and password, that credential must be stored only in the Password Manager software approved by {Company Name}, or, where the platform does not support retrieval from the Password Manager at runtime, in the native secrets management mechanism of that platform, such as a script properties store or secrets manager. The credential must not be embedded as plain text within the script, application, or connector configuration itself.

4.2.4 API keys, access tokens, and other non-password credentials used by an Artificial Intelligence tool or connector must be issued separately to each tool or connector and not shared between them, scoped to the minimum permissions necessary, rotated on a defined schedule where the vendor supports it, and revoked immediately when no longer in use.

4.2.5 Before an Artificial Intelligence tool is used to collect the biometric identifiers of a workforce member, such as facial geometry or a voiceprint, {Company Name} must obtain the prior written consent of that individual and provide notice of how the biometric data will be stored, used, and retained, in accordance with applicable state biometric privacy law.

4.3 Ownership, Accuracy, and Accountability

4.3.1 Results produced by authorized use of Artificial Intelligence tools, including business content, graphics, images, and data analysis, remain the property of {Company Name}, as outlined in the Internet, Email, and Computer Use Policy.

4.3.2 Employees must independently verify the accuracy of any output generated by an Artificial Intelligence tool before relying on it for a business decision, client deliverable, public communication, financial record, compliance artifact, legal-related document, policy document, or other business-critical use, and must not represent Artificial Intelligence-generated content as independently verified fact without that review. The use of Artificial Intelligence software does not guarantee accurate results.

4.3.3 The use of an Artificial Intelligence tool does not transfer or diminish the responsibility of {Company Name} or any employee for the accuracy, legality, or consequences of work performed with that tool. {Company Name} and its employees remain fully accountable for decisions, communications, and deliverables produced with the assistance of Artificial Intelligence tools, including consequences arising from an error, limitation, or unexpected behavior in the tool itself.

4.3.4 Software code generated or modified using an Artificial Intelligence tool is also governed by the Secure Development Policy, including its requirements for review and testing before production use, the instructions given to the tool, and the third-party components the code depends on.

4.4 Client Disclosure and Automated Decision-Making

4.4.1 The use of Artificial Intelligence tools must be disclosed to clients when that use includes “automated processing” or “facilitated decision-making” using clients’ personally identifiable information (PII).

4.4.2 {Company Name} must include disclosures in its privacy policy describing the data being collected, the use of third-party software for automation, third-party analytics, and reporting through data collection, and must inform clients that content is being processed using Artificial Intelligence tools or that automated decision-making is being used.

4.4.3 {Company Name} must grant clients the right to opt out of automated decision-making, or provide a meaningful human review process, consistent with applicable state law.

4.4.4 {Company Name} must conduct an AI Impact Assessment before using Artificial Intelligence tools to make or materially influence an important decision about an individual, such as an employment, credit, insurance, or housing decision, documenting the purpose, the data used, the safeguards in place, and the degree of human review involved. AI Impact Assessments are not intended to be made public, but must be made available to an applicable regulator upon request.

4.5 Oversight and Training

4.5.1 {Company Name} must assign personnel to review the use of Artificial Intelligence tools within the company to ensure compliance with legal and ethical obligations.

4.5.2 {Company Name} must oversee the use of Artificial Intelligence tools in the company by implementing internal policies and procedures and providing training to employees, in accordance with the Security Awareness Training Policy.

4.5.3 The Artificial Intelligence Oversight Team is the designated function responsible for Artificial Intelligence tool review, approval, inventory maintenance, incident coordination, and policy oversight under this policy, consistent with the responsibilities assigned to it throughout this section.

4.6 Application, Connector, and Vendor Approval

4.6.1 Employees must not use an Artificial Intelligence tool that has not been approved in accordance with sections 4.1.2 or 4.1.3 of this policy, regardless of whether the tool is free, browser-based, or embedded in a personal device.

4.6.2 Before a new Artificial Intelligence application or service is added to the Authorized Applications List, the Artificial Intelligence Oversight Team must review the data handling, data residency, retention, and subprocessor practices of the vendor to confirm the vendor does not use {Company Name} data to train models available to other businesses, unless authorized in writing. Where applicable, the governing contract, order form, or terms of service should address confidentiality, breach notification, data deletion, subcontractor use, and data location requirements. An Artificial Intelligence application, plug-in, or connector that has no vendor accountable for it, such as open-source or community-maintained software, must be reviewed before approval for its origin, its maintainers, and its history of changes, in place of the vendor review above. A third party that hosts a connector for {Company Name} and has access to {Company Name} or client information through it is a service provider and must be managed in accordance with the Service Provider Management Policy.

4.6.3 Before an Artificial Intelligence connector, plug-in, or integration is enabled to link an approved application to another system, service, or data source, {Company Name} must document and obtain approval from the Artificial Intelligence Oversight Team for the identity of the connector, the system, service, or data source it connects to, and the scope of permissions granted to it, such as read, write, or delete access.

4.6.4 Permissions granted to an Artificial Intelligence connector must be limited to the minimum access necessary to perform its intended function.

4.6.5 {Company Name} must maintain an inventory of approved Artificial Intelligence connectors as part of the Authorized Applications List, including the individual responsible for each connector, the actions each connector makes available, the systems each connector accesses, and the permissions granted to it.

4.6.6 An Artificial Intelligence connector that is modified to add an action it can perform, expand its permissions, or change the systems, services, or data sources it accesses must be treated as a new connector requiring re-approval under section 4.6.3.

4.6.7 A new agentic capability, agent platform, or autonomous workflow feature added by a vendor to previously approved software must remain disabled and unused until reviewed and approved by the Artificial Intelligence Oversight Team under section 4.6.3, regardless of whether the vendor enables the feature by default.

4.6.8 {Company Name} must review its Authorized Applications List and approved connector inventory at least annually, and whenever a vendor announces a new agentic Artificial Intelligence capability or materially changes its model training terms, data practices, connector permissions, or subprocessor relationships for software already in use, to confirm that no approved tool has expanded its scope without documented review.

4.6.9 The Artificial Intelligence Oversight Team may suspend or revoke approval for any Artificial Intelligence tool, connector, or use case at any time if the tool presents unacceptable risk, violates policy, is no longer supported by its vendor, or is no longer required for an approved business purpose.

4.6.10 A workforce member who becomes aware that an approved connector has been modified as described in section 4.6.6, including by an update the vendor applied, must report the change to the Artificial Intelligence Oversight Team and must not use the added action, permission, or access until the Artificial Intelligence Oversight Team has reviewed and approved it.

4.6.11 A connector update that adds an action the connector can perform, expands its permissions, or changes the systems, services, or data sources it accesses is not a routine automated update under the Personal Computing and Network Device Policy and is not pre-authorized by that policy.

4.7 Autonomous and Agentic AI Systems

4.7.1 Artificial Intelligence systems capable of acting autonomously, such as initiating transactions, sending communications, or modifying records without step-by-step human direction, must not be authorized for use unless each such action is subject to review and approval by an individual with the authority to make or change that decision.

4.7.2 {Company Name} does not prohibit the use of Artificial Intelligence systems that operate with substantial autonomy. Regardless of the degree of autonomy granted to an Artificial Intelligence system, the following actions must always require review and approval by an individual with the authority to make or change the decision before being finalized:

• the transfer or disbursement of funds;
• the execution of a contract or other binding commitment on behalf of {Company Name};
• a communication sent on behalf of {Company Name} to a client, regulator, or the public;
• a decision affecting the employment of an individual;
• the deletion or destruction of Protected Information or records;
• a change to a security setting, access permission, or other security configuration of a {Company Name} or client system; and
• a sequence of actions that, taken together, would have an effect comparable to any action listed above, even where no single action in the sequence would require review on its own.

4.7.3 Accountability for actions taken by an autonomous or agentic Artificial Intelligence system is governed by section 4.3.3 of this policy and is not diminished by the autonomy of the system.

4.8 Synthetic Media and Social Engineering Risks

4.8.1 Employees must treat unexpected requests for funds, credentials, or Protected Information received by phone, video, or recorded message with heightened suspicion, and must verify the requestor’s identity through a separate, previously known communication channel before acting, due to the risk of Artificial Intelligence-generated voice or video impersonation.

4.8.2 Artificial Intelligence transcription or recording tools must not be used to record a meeting or call unless all participants have provided consent, in accordance with applicable two-party or all-party consent recording laws.

4.9 Public Disclosure of AI-Generated Content

4.9.1 Content drafted using Artificial Intelligence tools that has been reviewed and approved by an employee before being shared publicly or with clients is considered {Company Name} content and is not required to be labeled as Artificial Intelligence-generated solely on that basis.

4.9.2 Content that is published or communicated without human review prior to release must be labeled or disclosed as Artificial Intelligence-generated when required by applicable law.

4.9.3 A chatbot, virtual assistant, or other automated system that communicates with a customer, client, or website visitor must clearly disclose that the individual is interacting with an automated system and not a human, where required by applicable law.

4.9.4 Content designed to simulate a human speaker, presenter, or correspondent, such as a synthetic voice or video, must be labeled or disclosed as Artificial Intelligence-generated when required by applicable law.

4.9.5 Synthetic media depicting the voice, image, or likeness of a specific, identifiable real person must not be created or used for any {Company Name} purpose without the prior consent of that person, regardless of whether the content is labeled as Artificial Intelligence-generated. This requirement is separate from, and is not satisfied by, the disclosure required under section 4.9.4.

4.9.6 Consent obtained for an individual to create synthetic media of their own voice, image, or likeness, such as a personal avatar, does not by itself constitute consent for {Company Name} to use that synthetic media for a public, client-facing, marketing, or other commercial purpose. Separate, documented consent identifying the intended use must be obtained before any such use.

4.10 Data Retention, Logging, and Deletion

4.10.1 Artificial Intelligence tools must not be used in a manner that conflicts with the Record Retention and Destruction Policy, data deletion requirements, or legal hold obligations.

4.10.2 Workforce members must not input Protected Information into any Artificial Intelligence tool unless the retention, logging, model training, and deletion settings for that tool have been reviewed and approved by the Artificial Intelligence Oversight Team.

4.10.3 Where an Artificial Intelligence tool stores prompts, outputs, transcripts, uploaded files, conversation history, or associated metadata, that information must be retained only for the minimum period necessary to support the approved business purpose and in accordance with the Record Retention and Destruction Policy and applicable legal hold requirements.

4.10.4 If an Artificial Intelligence vendor permits the disabling or limitation of prompt retention, conversation history, model training, or similar data reuse settings, {Company Name} must disable or limit those functions unless a documented business need requires otherwise and the Artificial Intelligence Oversight Team approves that use.

4.10.5 Prompts, outputs, transcripts, and other Artificial Intelligence-generated or Artificial Intelligence-processed artifacts that become part of a business record, client file, compliance record, or other required company documentation must be retained in the authorized {Company Name} system of record in accordance with applicable policy and law.

4.10.6 If {Company Name} discontinues use of an Artificial Intelligence tool, or if an approved use case ends, associated accounts, credentials, prompts, outputs, uploaded files, and stored artifacts must be deleted, disabled, archived, or otherwise dispositioned in accordance with the Record Retention and Destruction Policy, contract requirements, and applicable law.

4.10.7 Activity performed through an Artificial Intelligence connector, including each approval or denial of an action by an individual under section 4.7, must be logged where the tool or connector provides a record of that activity, and those logs must be managed in accordance with the Security Logging Policy.

4.10.8 Where an Artificial Intelligence tool or connector provides no record of its activity, the Artificial Intelligence Oversight Team must document that limitation as part of the approval of the connector under section 4.6.3.

4.11 Incident Reporting and Response

4.11.1 Any suspected or actual misuse of an Artificial Intelligence tool must be reported immediately to the Information Security Team in accordance with the Incident Response Plan Policy. Reportable events include, but are not limited to, unauthorized disclosure of Protected Information, use of an unapproved Artificial Intelligence tool, unexpected autonomous action, materially inaccurate or harmful output used in business operations, suspected prompt injection, suspected credential compromise, or unauthorized connector activity.

4.11.2 Workforce members who become aware of a suspected Artificial Intelligence-related incident must take reasonable steps to contain it, including stopping use of the affected tool where feasible, disabling an affected connector and revoking the credentials issued to it where they have the access to do so, and preserving relevant evidence, and must not independently notify a client, regulator, vendor, media outlet, or other third party regarding the incident unless specifically authorized to do so by {Company Name}.

4.11.3 Reported events are handled in accordance with the Incident Response Plan Policy, including breach notification, client notification, legal hold, and regulatory reporting procedures as applicable.

4.12 Risk Classification and Use Approval

4.12.1 Artificial Intelligence use cases must be evaluated based on risk before approval for use.

4.12.2 Low-risk uses may include grammar correction, summarization of public or non-confidential materials, formatting assistance, or internal drafting that does not involve Protected Information, client-impacting decisions, or autonomous action.

4.12.3 Moderate-risk uses may include internal business automation, research using non-sensitive business information, or Artificial Intelligence-assisted content generation that will be reviewed by an employee before use.

4.12.4 High-risk uses are those described in sections 4.2.1, 4.2.5, 4.4, 4.7, and 4.9.5 of this policy, including uses involving Protected Information, biometric data, automated decision-making, autonomous action, connector access to business systems, or synthetic media involving an identifiable person.

4.12.5 High-risk uses require documented approval by the Artificial Intelligence Oversight Team and, where applicable, the designated business owner, system owner, or other authorized approver before deployment or use.

4.12.6 {Company Name} may establish additional written standards, testing requirements, validation procedures, monitoring requirements, and periodic reauthorization requirements for moderate-risk or high-risk Artificial Intelligence uses.

4.12.7 The Artificial Intelligence Oversight Team must classify the risk of an Artificial Intelligence tool, connector, or use case from the access it has and the actions it can perform, and must not base that classification on risk, safety, or permission labels supplied by the vendor or by the tool itself.

5. Policy Compliance

5.1 Compliance Measurement

The Information Security Team must verify compliance with this policy through various methods, including but not limited to, periodic walk-throughs, video monitoring, business tool reports, internal and external audits, and feedback to the policy owner.

5.2 Exceptions

Any exception to this Artificial Intelligence Acceptable Use Policy must be approved in advance and documented by the Information Security Team.

5.3 Non-Compliance

An employee found to have violated this policy may be subject to disciplinary action, up to and including termination of employment.

6. Related Standards, Policies, and Processes

Internet, Email, and Computer Use Policy
Password Policy
Password Construction Policy
Personal Computing and Network Device Policy
Software Installation Policy
Software Usage Policy
Authorized Applications List
Acceptable Encryption Policy
Security Logging Policy
Electronic Data Protection Policy
Security Awareness Training Policy
Incident Response Plan Policy
Social Media Policy
Record Retention and Destruction Policy
Data Classification Policy
Service Provider Management Policy
Secure Development Policy
CIS MCP Server Benchmark v1.0.0
CIS Controls v8.1 Model Context Protocol (MCP) Companion Guide v1.0

7. Definitions and Terms

Protected Information is electronically stored data that contains Personally Identifiable Information (PII), or other confidential or sensitive information.

Artificial Intelligence (AI) is intelligence — perceiving, synthesizing, and inferring information — demonstrated by machines, as opposed to intelligence displayed by humans or animals. Example tasks in which this is done include speech recognition, computer vision, translation between languages, as well as other mappings of inputs.

Generative Artificial Intelligence (Generative AI, or GenAI) is a type of Artificial Intelligence system capable of generating text, images, or other media in response to prompts. Generative AI models learn the patterns and structure of input data, and then generate new content that is similar to the training data but with some degree of novelty, rather than only classifying or predicting data.

Machine Learning is a subset of Artificial Intelligence technology. It is focused on teaching computers to learn from data and to improve with experience, instead of being explicitly programmed to do so. In Machine Learning, algorithms are trained to find patterns and correlations in large data sets and to make decisions and predictions based on that analysis.

Large Language Model (LLM) is a model consisting of a neural network with many parameters, typically billions of weights or more, trained on large quantities of unlabeled text using self-supervised or semi-supervised learning. LLMs are general purpose models which excel at a wide range of tasks, as opposed to being trained for one specific task.

Artificial General Intelligence (AGI) is the intelligence of machines that allows them to comprehend, learn, and perform intellectual tasks much like humans.

Agentic AI refers to Artificial Intelligence systems capable of independently planning and executing multi-step tasks, including invoking other software or services, with limited or no step-by-step human direction.

Agent Platform refers to a cloud-hosted environment that creates, runs, or orchestrates multiple Artificial Intelligence agents, often enabling those agents to act autonomously across connected systems with limited ongoing human direction.

Connector refers to a configured integration that allows an Artificial Intelligence tool or agent to access, read, or act upon a system, service, or data source other than the one in which the tool was originally installed or approved.

Automated Decision-Making is a decision made about an individual using Artificial Intelligence tools or other automated processing with little or no meaningful human involvement.

Shadow AI refers to the use of Artificial Intelligence tools or services that have not been reviewed, approved, or authorized by a business in accordance with its policies.

Synthetic Media refers to audio, video, or images generated or manipulated using Artificial Intelligence tools to realistically depict a person, often referred to as a deepfake.

Workforce Member refers to an employee, contractor, consultant, temporary worker, volunteer, trainee, or other person whose work for {Company Name} is under the direct control of {Company Name}, whether or not that person is paid by {Company Name}. A service provider engaged to perform work independently, and not under the direct control of {Company Name}, is not a workforce member and is governed by the Service Provider Management Policy.