AGENCYSCRIPT
CoursesEnterpriseBlog
đź‘‘FoundersSign inJoin Waitlist
AGENCYSCRIPT

Governed Certification Framework

The operating system for AI-enabled agency building. Certify judgment under constraint. Standards over scale. Governance over shortcuts.

Stay informed

Governance updates, certification insights, and industry standards.

Products

  • Platform
  • Certification
  • Launch Program
  • Vault
  • The Book

Certification

  • Foundation (AS-F)
  • Operator (AS-O)
  • Architect (AS-A)
  • Principal (AS-P)

Resources

  • Blog
  • Verify Credential
  • Enterprise
  • Partners
  • Pricing

Company

  • About
  • Contact
  • Careers
  • Press
© 2026 Agency Script, Inc.·
Privacy PolicyTerms of ServiceCertification AgreementSecurity

Standards over scale. Judgment over volume. Governance over shortcuts.

On This Page

Export Control BasicsWhat Are Export ControlsKey Regulatory FrameworksThe Commerce Control ListAI-Specific Export Control IssuesThe October 2022 and Subsequent Semiconductor ControlsAI Model ControlsTraining Data ControlsDeemed ExportsCompliance Framework for AI AgenciesExport Control ClassificationRestricted Party ScreeningTechnology Access ControlsRecord KeepingTrainingRed Flags to Watch ForWorking with Government and Defense ClientsPenalties for ViolationsYour Next Step
Home/Blog/One New Hire's Codebase Access Triggered an Export Violation
Governance

One New Hire's Codebase Access Triggered an Export Violation

A

Agency Script Editorial

Editorial Team

·March 20, 2026·12 min read
ai export controlsai technology transferear ai regulationsai sanctions compliance

An AI agency in Virginia had been working with a defense contractor on computer vision models for autonomous inspection systems. When the agency hired a talented machine learning engineer who happened to hold dual citizenship with a country subject to US sanctions, nobody flagged the potential export control issue. The engineer was given access to the full codebase, model architectures, and training pipelines. Three months later, during a routine compliance review triggered by the defense contractor, the issue surfaced. The agency had potentially committed a "deemed export" violation—sharing controlled technology with a foreign national without the required license. The resolution involved an expensive voluntary self-disclosure to the Bureau of Industry and Security (BIS), legal fees exceeding $150,000, and the loss of the defense contractor relationship.

Export controls are the compliance area that AI agencies most commonly overlook. Most agency founders think of export controls as something that applies to weapons manufacturers and semiconductor companies. They do not realize that AI models, training methodologies, specialized datasets, and even technical consulting can be subject to export control regulations. As AI technology becomes increasingly strategic, export control enforcement is intensifying, and the penalties for violations include criminal prosecution, not just civil fines.

This post covers the export control framework as it applies to AI agencies, the specific triggers you need to watch for, and the compliance practices that keep you on the right side of these regulations.

Export Control Basics

What Are Export Controls

Export controls are government regulations that restrict the transfer of certain technologies, products, and information to foreign countries, foreign nationals, or designated entities. The purpose is to prevent adversaries from acquiring technologies that could be used against national security or foreign policy interests.

For AI agencies, "export" includes:

  • Sending AI models, code, or data to recipients outside the United States
  • Sharing technical information about AI systems with foreign nationals (even within the United States—the "deemed export" rule)
  • Providing AI consulting or training services to foreign clients
  • Hosting AI systems on infrastructure located outside the United States
  • Making AI tools available for download from the internet without access controls

Key Regulatory Frameworks

Export Administration Regulations (EAR): Administered by the Bureau of Industry and Security (BIS) within the Department of Commerce. The EAR controls the export of dual-use technologies—technologies with both civilian and military applications. Most AI technology falls under EAR jurisdiction.

International Traffic in Arms Regulations (ITAR): Administered by the Directorate of Defense Trade Controls (DDTC) within the State Department. ITAR controls defense articles and defense services. AI technology specifically designed for military applications may be ITAR-controlled.

Office of Foreign Assets Control (OFAC) Sanctions: Administered by the Treasury Department. OFAC maintains lists of sanctioned countries, entities, and individuals. Providing AI services or technology to sanctioned parties is prohibited regardless of whether the technology itself is controlled under EAR or ITAR.

Entity List: BIS maintains a list of foreign entities that are subject to specific export restrictions. Providing AI technology to Entity List parties typically requires a license that is often denied.

The Commerce Control List

The EAR includes the Commerce Control List (CCL), which categorizes controlled items by Export Control Classification Number (ECCN). AI-related items fall primarily under Category 4 (Computers) and Category 5 (Telecommunications and Information Security).

Key ECCNs for AI agencies:

  • ECCN 4A003-4A005: Computers and related equipment, including specialized AI accelerators
  • ECCN 4D004: Software specially designed for AI applications in certain controlled contexts
  • ECCN 4E001: Technology for the development, production, or use of items in 4A or 4D
  • ECCN 5D002: Information security software, relevant if your AI system includes encryption or cybersecurity components

EAR99: Items not specifically listed on the CCL are designated EAR99. Most commercial AI software and services fall under EAR99, which means they can generally be exported without a license—but not to sanctioned countries, Entity List parties, or for prohibited end uses.

AI-Specific Export Control Issues

The October 2022 and Subsequent Semiconductor Controls

BIS has implemented increasingly comprehensive controls on advanced semiconductor technology exports, particularly to China. These controls directly affect AI because advanced AI training and inference depend on high-performance computing hardware.

What is controlled:

  • Advanced AI chips (GPUs and AI accelerators above certain performance thresholds)
  • Semiconductor manufacturing equipment
  • Technology and software for designing advanced semiconductors

What this means for agencies:

  • You cannot provide AI services that require controlled hardware to be deployed in restricted countries
  • Cloud-based AI services hosted on controlled hardware may be subject to restrictions when accessed from restricted countries
  • Technical knowledge about optimizing AI for controlled hardware may itself be controlled

AI Model Controls

As of 2026, the landscape for AI model export controls is evolving rapidly.

Foundation models: There has been ongoing regulatory discussion about controlling the export of large foundation models, particularly those above certain capability thresholds. The specific thresholds and controls continue to be refined.

Application-specific models: AI models trained for specific applications (nuclear technology, advanced weapons systems, intelligence analysis) may be controlled based on their application rather than their underlying technology.

Model weights and architectures: Model weights (the trained parameters) and novel architectures may be classified as "technology" under the EAR and subject to control depending on their capability and intended use.

Training Data Controls

Training data can be subject to export controls if:

  • The data itself is controlled (classified information, controlled technical data, restricted government data)
  • The data was created using controlled processes or equipment
  • The data is specifically curated for developing controlled AI capabilities

Deemed Exports

This is the issue that catches most agencies. A deemed export occurs when controlled technology is released to a foreign national within the United States. "Release" includes visual inspection, oral exchange, or application of the technology.

Practical impact:

  • Hiring foreign nationals and giving them access to controlled AI technology may require a deemed export license
  • Having foreign national contractors or consultants work on controlled AI projects may require a license
  • Presenting controlled AI technology at conferences or meetings where foreign nationals are present may constitute a deemed export
  • Sharing controlled AI code repositories with foreign national employees or contractors is a potential deemed export

Who is a "foreign national": Any person who is not a US citizen, lawful permanent resident, or protected individual (asylee or refugee). This includes H-1B visa holders, L-1 visa holders, and other temporary residents.

Compliance Framework for AI Agencies

Export Control Classification

Step 1: Classify your technology. For each AI product, service, and technology your agency develops or provides, determine its export control classification.

  • Is it specifically listed on the Commerce Control List? If so, what is the ECCN?
  • If not listed, is it EAR99?
  • Could it be ITAR-controlled based on its military application?

Step 2: Know your clients and their end uses. For each client and engagement, determine:

  • Where is the client located?
  • Where will the AI technology be deployed?
  • Who are the end users?
  • What is the end use?
  • Is the client, any end user, or any involved party on a restricted party list?

Step 3: Determine licensing requirements. Based on the classification, destination, end use, and end user, determine whether an export license is required.

Restricted Party Screening

Screen every client, partner, and vendor against restricted party lists before engaging with them.

Lists to screen:

  • BIS Entity List
  • BIS Denied Persons List
  • BIS Unverified List
  • OFAC Specially Designated Nationals (SDN) List
  • OFAC Consolidated Sanctions List
  • State Department Debarred Parties List

When to screen:

  • Before onboarding a new client
  • Before hiring or contracting with any individual
  • Before establishing a partnership or vendor relationship
  • Periodically (at least annually) for existing relationships, as lists are updated frequently

How to screen:

  • Use a commercial restricted party screening service. Manual screening against multiple lists is error-prone and time-consuming.
  • Document all screening results and retain records for at least five years.

Technology Access Controls

Implement controls that prevent unauthorized access to potentially controlled technology.

Physical access: Control access to facilities where controlled AI work is performed. Use access badges, visitor logs, and clean desk policies.

Digital access: Control access to repositories, systems, and data that contain potentially controlled technology.

  • Role-based access controls aligned with need-to-know
  • Separate repositories for controlled and uncontrolled technology
  • Access logging and monitoring
  • VPN and secure access for remote work

Personnel controls: Before granting access to controlled technology, verify that each individual is authorized.

  • Citizenship and immigration status verification for deemed export purposes
  • Background checks appropriate to the classification level
  • Access agreements acknowledging export control obligations

Record Keeping

Maintain records of all export control compliance activities.

  • Classification determinations for each technology
  • Restricted party screening results
  • License applications and approvals
  • Technology access grants and revocations
  • Training records
  • Compliance assessments and audit results

Retain records for at least five years from the date of the export or deemed export.

Training

Train your team on export control requirements.

  • All employees: Basic awareness of export controls—what they are, why they matter, and how to identify potential issues
  • Technical staff: How to recognize controlled technology and when to escalate classification questions
  • Sales and business development: How to screen clients and recognize red flags in potential engagements
  • Leadership: Export control risk management, compliance program oversight, and liability implications

Red Flags to Watch For

Train your team to recognize these red flags that may indicate export control issues:

  • A client in a sanctioned country or with connections to sanctioned countries
  • Unusual routing: A client requests that AI technology be delivered to a third country rather than their own location
  • Military or dual-use applications: A client wants AI for applications that could have military uses
  • Reluctance to provide end-use information: A client is vague or evasive about how the AI technology will be used
  • Requests to bypass controls: A client asks you to remove access restrictions, export markings, or other controls
  • Unusual payment arrangements: Payment from third-party entities, shell companies, or in currencies inconsistent with the transaction
  • Prior violations: A client or partner has a history of export control violations

Working with Government and Defense Clients

If your agency serves government or defense clients, your export control obligations are significantly more complex.

Facility clearances: Some defense AI work requires a facility security clearance (FCL). Obtaining an FCL involves a lengthy application process and ongoing compliance obligations.

Personnel clearances: Individual security clearances may be required for team members working on classified or sensitive AI projects.

ITAR compliance: Defense-specific AI may be ITAR-controlled, which imposes stricter requirements than EAR. ITAR prohibits sharing controlled information with any foreign national without a specific license, with limited exceptions.

CMMC compliance: The Cybersecurity Maturity Model Certification (CMMC) framework applies to defense contractors and may be required for AI work involving controlled unclassified information (CUI).

Penalties for Violations

Civil penalties under EAR: Up to $364,992 per violation or twice the value of the transaction, whichever is greater.

Criminal penalties under EAR: Up to $1 million per violation and up to 20 years imprisonment.

OFAC penalties: Up to $356,579 per violation for civil penalties; criminal penalties up to $20 million and 30 years imprisonment.

ITAR penalties: Civil penalties up to $1,282,564 per violation; criminal penalties up to $1 million and 20 years imprisonment.

Debarment: Companies and individuals can be barred from participating in export transactions.

Your Next Step

Start with an export control audit of your agency. Identify the AI technologies you develop and provide, classify them under the EAR, and determine whether any of your current clients or engagements trigger export control requirements. Implement restricted party screening for all new clients and personnel. Train your team on basic export control awareness.

If you find potential issues during your audit, engage an export control attorney. Voluntary self-disclosure to BIS, while not pleasant, is far better than an enforcement action for an unreported violation.

The agency that understands export controls is the agency that can serve defense clients, international enterprises, and other high-value customers that require this compliance maturity. Ignoring export controls is not just risky—it locks you out of some of the most lucrative segments of the AI market.

Search Articles

Categories

OperationsSalesDeliveryGovernance

Popular Tags

prompt engineeringai fundamentalsai toolsthe difference between AIMLagency operationsagency growthenterprise sales

Share Article

A

Agency Script Editorial

Editorial Team

The Agency Script editorial team delivers operational insights on AI delivery, certification, and governance for modern agency operators.

Related Articles

Governance

Complete EU AI Act Compliance Guide — What Every AI Agency Needs to Know and Do

The EU AI Act is the most comprehensive AI regulation on the planet. Here is exactly what it requires from AI agencies, which of your systems are affected, and a step-by-step compliance roadmap you can start executing today.

A
Agency Script Editorial
March 21, 2026·15 min read
Governance

HIPAA Compliance Guide for AI in Healthcare — Building AI Systems That Protect Patient Data

Healthcare AI is booming, but one HIPAA violation can end your agency. Here is the complete guide to building HIPAA-compliant AI systems, from BAAs to technical safeguards to breach response.

A
Agency Script Editorial
March 21, 2026·15 min read
Governance

Question 14 Cost a Chicago Agency Its Fortune 500 Deal

ISO 27001 certification is becoming a prerequisite for enterprise AI contracts. Here is the complete implementation guide from gap analysis to certification audit, tailored for AI agencies.

A
Agency Script Editorial
March 21, 2026·14 min read

Ready to certify your AI capability?

Join the professionals building governed, repeatable AI delivery systems.

Explore Certification