The Tool Desk
Outbyte PC Repair FREEClear out junk files and repair common Windows errorsFree Scan →Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →iTechGuides is reader-supported. When you buy through links on our site, we may earn an affiliate commission. As an Amazon Associate I earn from qualifying purchases. Learn more
Manage an employee’s AI-generated app like any other software: assign a business owner, record what it does, and scale review to the data it handles, who can access it, and the consequences of failure. AI assistance does not transfer accountability. A workable program gives employees an approved way to build while requiring stronger checks for apps that expose sensitive data or take consequential actions.
Start with an approved path and an app inventory
Make it possible for employees to build safely rather than relying on a blanket ban. Set out which AI tools, development environments, and hosting locations are approved, then require a short intake record before an app is deployed or connected to business data.
Record the app’s purpose, business sponsor, named owner, users, deployment location, data handled, external services, AI tools used, dependencies, and permissions. Also note whether AI only helped write the app or whether the running app itself calls an AI model or can take actions. This inventory gives IT and security a way to find the app, reach the person accountable for it, and review it when its use changes.
NIST’s Generative AI Profile (AI 600-1) includes governance considerations such as auditing, data protection, education, incident response, monitoring, and decommissioning. It is voluntary guidance, not a universal regulatory requirement.
Decide how much review the app needs
Use your organization’s existing risk tiers if they fit; adapt them if generative AI introduces risks your process does not capture. NIST says organizations may apply existing tiers to generative AI or revise them, and notes that additional human review, documentation, tracking, and management oversight may be warranted.
Consider these factors together rather than relying on whether an app was written by a professional developer or generated with AI:
- Data: How sensitive is it, and how much of it does the app handle?
- Access and exposure: Who can use the app, and is it available outside the organization?
- Consequences: Could an error affect people, business-critical work, or important records?
- Capabilities: Can the app edit or delete records, spend money, or send communications?
- Safeguards: Are testing, logging, access controls, and rollback adequate for the app’s role?
Route higher-impact apps to the relevant security, privacy, legal, or engineering reviewers. Applicable legal obligations depend on the jurisdiction, industry, data, and function; the fact that an app uses AI does not by itself establish which rules apply.
Do these 3 things before closing this tab:
1Clear out junk files and repair common Windows errors2Fix the driver behind crashes, sound loss and screen glitches3Repair Windows errors before they cause bigger problemsRank #2
Keep sensitive information out of AI prompts and restrict app data
Tell employees which AI services they may use and what information may be submitted to them. Microsoft’s Windows development security guidance advises against putting API keys, passwords, or connection strings into prompts. It also warns against sharing real customer names, email addresses, usage data, or other personal information, and recommends checking organizational policy before sharing proprietary code. Synthetic data is safer for examples.
For the deployed app, define what data it can access and how long it retains that data. Require appropriate authentication and authorization, and give each app only the access it needs. Role- or group-based access can be more appropriate than broad access when an app serves a specific team.
Review and test generated code as software
The employee responsible for the app should be able to explain and test the changes before they are deployed. Microsoft’s guidance puts it plainly: “AI tools don’t remove the need for code review. They change what you’re reviewing, not whether you review.” Generated code needs review; authorship is not a substitute for understanding its behavior.
Rank #3
At minimum, review and test the app’s inputs, data access, permissions, and failure behavior. Check that user input has appropriate type, length, and range limits. Do not pass raw input into shell commands, file paths, or database queries, and ensure resolved file paths stay within permitted directories. Remove capabilities and permissions the app does not need.
Review each suggested software dependency: verify its publisher, consider whether it is maintained, and scan for known vulnerabilities. For a NuGet project, Microsoft documents dotnet list package --vulnerable as one scanning example; it is not a universal command for every language or package ecosystem. Keep credentials out of source code and use an appropriate vault or secret-management service.
NIST SP 800-218A adds AI-specific secure software development practices to the Secure Software Development Framework. It complements SP 800-218 rather than replacing it, and is aimed at producers of AI models and systems and their acquirers; it is lifecycle guidance, not a checklist every employee must implement alone.
Add controls if the app uses AI at runtime
An app created with AI is not necessarily an AI-powered app. If the deployed app calls a model or agent, document its purpose, boundaries, data flows, and expected behavior. Test representative and edge-case inputs, consider disclosing material AI use to affected users, and keep people involved in consequential decisions.
Do not allow an AI feature to delete data, make purchases, or send messages autonomously without explicit confirmation. If an agent can call tools or APIs, set action limits in deterministic controls rather than relying only on instructions to the model. Grant only the tools and data required, require approval for high-risk or irreversible actions, and provide a reliable way to pause or stop the agent. Show people the intended action and keep logs of tool calls, data accessed, approvals, and outcomes. These controls help address risks such as unintended actions, sensitive-data exposure, supply-chain compromise, and excessive agent permissions identified in Microsoft’s guidance on human review and approvals for AI agents.
Recommended Free Tools
Monitor changes, respond to incidents, and retire unused apps
Deployment is not the end of the review. Track app use, permissions, dependencies, incidents, and material changes; reassess the app when its data, audience, integrations, or capabilities change. Name the person or team responsible for responding to a suspected exposure or unsafe action, and define how to revoke access or disable integrations.
Periodically confirm that the app still has an owner and a business purpose. When it is no longer needed, remove the app and revoke its credentials and integrations. NIST AI 600-1 includes monitoring, incident response, change management, auditing, and decommissioning among its governance considerations.
Use guidance proportionately
The NIST AI Risk Management Framework 1.0 is voluntary, and NIST says it is being revised. Its AI Risk Management Framework page provides status information. NIST released the Generative AI Profile on July 26, 2024; its announcement describes the profile. NIST SP 800-218A was released July 26, 2024, and NIST’s announcement records an update on June 25, 2025.
These documents can help shape a company’s process, but they do not establish one universal approval threshold or risk-scoring formula for citizen-built apps. Set review depth according to the app’s actual exposure and potential impact.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
