AS-2606_Requirements for High-Impact AI Systems_SIGNED_04-23-2026.pdf

PDF 343 KB Posted

Attached to
Issuance of HUD Acquisition Instruction 26‑02 – Acquisition of Artificial Intelligence and New Agency‑Specific Clauses
Solicitation number
Not on record
Issued by

About this file

This is a federal contract clause (AS-2606) that establishes requirements for High-Impact AI Systems in all HUD solicitations and contracts above the micropurchase threshold.

The clause mandates that contractors comply with comprehensive risk management and mitigation requirements upon designation of an AI system or service as a High-Impact AI System. Key requirements include: conducting and documenting pre-deployment testing and validation that reflects real-world outcomes, with contractors providing access to system functionality, documentation, and technical assistance for government review; implementing continuous post-deployment monitoring to identify adverse performance and security impacts, including potential violations of privacy, civil rights, or civil liberties laws; developing role-based training and assessments for users and operators on interpreting outputs, managing risks, and complying with applicable policies; ensuring the AI system enables appropriate human oversight, intervention, override, or suspension of automated outputs, and supporting implementation of fail-safe mechanisms to minimize harm. Contractors must also provide comprehensive technical and non-technical documentation including intended purpose, expected benefits, cost information, data characteristics, potential impacts on privacy and civil rights with mitigation measures, implementation evidence of risk mitigation requirements, and processes for periodic reassessments. Additional obligations include maintaining sufficient logging and traceability of system inputs and outputs to support government oversight and investigation, and promptly reporting incidents, failures, or identified harms with corrective actions. Contractors must notify the government in advance of material updates, retraining events, architectural changes, or other modifications that could alter the system's risk profile or performance characteristics. All requirements must be flowed down to subcontractors providing AI components or services, and contractor non-compliance may result in disincentives, liquidated damages, actual damages, or work stoppage until compliance is demonstrated.

View the file

Other files

On GovTribe

Work with this file

  • Download the original file
  • Contacts named in this file
  • Similar government files
  • Ask GovTribe AI about this file

Text version

PRESCRIPTION: USE THIS CLAUSE IN ALL SOLICITATIONS AND CONTRACTS

ABOVE THE MICROPURCHASE THRESHOLD.

AS-2606 Requirements for High-Impact AI Systems

Contractors shall comply with all applicable federal and agency risk mitigation requirements for any AI system or service (collectively referred to as AI System) determined by the Government to be a High-Impact AI System. These requirements apply regardless of whether the High-Impact

AI System designation occurs prior to or after contract award.

A list of AI Definitions can be found on the AI@HUD page on hud.gov.

1. Risk Management and Mitigation Requirements:

Upon designation as a High-Impact AI System, the Contractor shall implement and document risk management and mitigation measures proportionate to the AI System’s impact, including, at a minimum:

A. Support the Government in conducting and documenting pre-deployment testing and validation that reflect expected real-world outcomes and identify expected benefits of the AI System. Such support shall include providing access to system functionality, documentation, evaluation results, and technical assistance necessary for Government review. If the Government does not have access to the underlying AI System source code, models, or data, the Contractor shall support alternative test methodologies, such as querying the AI System and observing the outputs or executing Government-provided evaluation data and providing results.

In addition, the Contractor shall conduct and document its own pre-deployment testing and validation activities prior to operational use, including performance evaluation, robustness testing, and identification of known limitations and risks.

The Contractor shall provide the Contracting Officer Representative with the results of such testing.

B. Implement continuous monitoring of the High Impact AI System performance post-deployment. Identify any adverse impacts to the performance and security of the High Impact AI System, including potential violations of laws governing privacy, civil rights, or civil liberties. Ongoing monitoring must be designed to detect unforeseen circumstances, changes to the High Impact AI System after deployment, and changes to the context of use or associated data. The Contractor shall support the Government in developing and implementing appropriate mitigations, maintaining accurate High Impact AI System and use documentation, and developing processes that enable traceability and transparency of the monitoring and evaluation activities.

https://www.hud.gov/ai

C. Support the Government in developing role-based training and training assessments that instruct High Impact AI System users and operators how to interpret and act on High Impact AI System outputs, manage associated risks, and comply with applicable policies. Training shall be tailored to each user role and include guidance on human oversight, error handling, and risk mitigation procedures.

D. Ensure the High Impact AI System enables appropriate human oversight and review, intervention, override, or suspension of automated outputs, as directed by the Government. Support the Government in implementing fail-safe mechanisms that minimize the risk of significant harm.

E. Provide technical and non-technical documentation sufficient to support

Government oversight and audit of the High Impact AI System including, but not limited to:

i. Intended purpose and expected benefits of the AI System and supporting metrics and/or qualitative analysis that show expected improvement from existing process(es)

ii. Cost information, including costs associated with the AI System (e.g., development, licensing, infrastructure / hosting, security, etc.)

iii. Data characteristics, including data used, suitability of the data, how the data was gathered and prepared for use by the AI System, whether the data will be publicly disclosed as open government data, and whether the data includes information about groups protected by Federal nondiscrimination laws and, if so, provide further information.

iv. Potential impacts on privacy, civil rights, and civil liberties of the public of using the AI System, and planned mitigation measures for anticipated negative impacts.

v. Documentation demonstrating the implementation of the risk mitigation requirements as defined in sections 1A-D of this clause, AS-2606

vi. Processes, procedures, and schedule for conducting periodic reassessments of the AI System. This will include reassessments following any significant system changes.

F. Maintain sufficient logging and traceability of High Impact AI System inputs, outputs, and material High Impact AI System actions to support Government oversight, investigation, and audit.

G. Promptly report High Impact AI System incidents, failures, or identified harms to the Government and implement corrective actions as directed.

2. Changes and Updates:

The Contractor shall notify the Government in advance of material updates, retraining events, architectural changes, or other modifications that could alter the High Impact AI

System’s risk profile or performance characteristics.

3. Subcontractor Compliance:

The Contractor shall flow down all Requirements for High-Impact AI Systems to its subcontractors that are providing AI components, AI Systems, or AI Services under the contract and/or utilizing AI components, AI Systems, or AI Services in performance of the contract.

The Contractor shall provide an estimate of the cost to bring the AI System or AI Service into compliance with applicable Government requirements. Based on this information, the

Government will determine whether to continue, modify, or discontinue use of the AI System and/or AI Service.

The Contractor’s failure to adhere to the requirements of this Agency Specific clause could result in the Government exercising any rights available under the terms of the contract, including assessment of disincentives, liquidated damages, and/or actual damages, as appropriate. The

Government reserves the right to stop work on the contract pursuant to this clause until the

Contractor demonstrates compliance with this requirement.

End Clause

X Francine L. Hemphill

Chief Procurement Officer

2026-04-23T09:11:43-0400
FRANCINE HEMPHILL

File details come from the government source that posted it. Updated .