USA-010

Hearing Recording and Transcriptions (HeaRT) system for Social Security disability hearings

Download PDF
United States North America High income Scaled & Institutionalised Confirmed

Social Security Administration (SSA), Office of Hearings Operations (OHO)

At a Glance

What it does Perception and extraction from unstructured inputs — Operational and process automation
Who runs it Social Security Administration (SSA), Office of Hearings Operations (OHO)
Programme Social Security Disability Insurance (SSDI) hearings process
Confidence Confirmed
Deployment Status Scaled & Institutionalised
Key Risks Model-related risks
Key Outcomes SSA projects annual savings of approximately USD 5 million through elimination of hardware procurement, maintenance, and manual transcription costs.
Source Quality 4 sources — Other, News article / media, Government website / press release

The Hearing Recording and Transcriptions (HeaRT) system is an AI-enabled software platform deployed by the Social Security Administration (SSA) to record and produce automated transcripts of disability hearings conducted by Administrative Law Judges (ALJs). HeaRT replaces an older hardware-based recording system that had been installed in SSA hearing offices across the United States, shifting the function to a software-based workflow that no longer depends on dedicated physical recording equipment.

The retained source base supports HeaRT as an automated transcription system using speech-recognition technology to convert spoken testimony from disability hearings into written text. SSA's official materials describe the system as using generative AI for automated transcripts, while secondary commentary discusses Automatic Speech Recognition (ASR) and possible additional AI-assisted processing. The public sources do not disclose the exact model architecture, vendor, or whether the deployed system includes any large-language-model layer beyond transcription.

The system supports all hearing formats currently used by SSA, including in-person hearings, telephone hearings, and video hearings, without requiring dedicated recording hardware for each modality. SSA completed the nationwide rollout of HeaRT by 17 March 2025, making it operational across all hearing offices in the country. That matters because the disability hearing process is one of the most procedurally sensitive parts of SSA administration: the hearing record captures claimant testimony, representative argument, and expert input that may later be reviewed on appeal. Reliable capture of the spoken record is therefore central to administrative fairness even if the recording tool itself does not determine eligibility.

HeaRT is aimed at a specific operational bottleneck rather than at adjudication. Prior to rollout, SSA relied on ageing recording hardware that could fail, require maintenance, or be mismatched to newer hearing modalities such as phone and video proceedings. A software-based recording and transcription workflow promises more consistent capture across hearing formats, less dependence on local equipment, and faster production of hearing records that staff and adjudicators can use downstream. In administrative terms, it is a classic process-automation deployment: it removes manual transcription burdens and modernises a legacy support function around a high-volume adjudicative process.

SSA projects that the system will serve approximately 500,000 customers annually who go through the disability hearing process and will save about USD 5 million per year by reducing hardware, maintenance, and manual transcription costs. By removing reliance on physical recording equipment and reducing transcription burdens, HeaRT is intended to reduce hearing disruptions and improve administrative efficiency. The official framing is therefore about continuity of operations and cost reduction as much as about AI modernisation.

The deployment sits within SSA's broader push to use AI-enabled tools in disability administration, but HeaRT itself should be understood narrowly as a recording and transcription tool rather than a benefits-decision system. It does not decide whether a claimant qualifies for SSDI or SSI, nor does it rank claims by merit. Instead, it produces a textual artefact that becomes part of the hearing record. That distinction is important for coding decision criticality: transcript quality can influence later proceedings, but the system is still one step removed from the substantive legal decision.

The deployment of HeaRT sits within the context of SSA's significant operational pressures. As of 2025, SSA faced a backlog of approximately 1.4 million disability claims, with an average processing time of eight months for initial decisions. The agency has pursued multiple AI-enabled tools to address these pressures, including the Quick Disability Determinations (QDD) predictive model and the IMAGEN document processing system, with HeaRT representing the transcription component of this broader modernisation effort.

Acting Commissioner Lee Dudek stated at the rollout completion that the system would allow SSA to "better serve the public by making the hearings process more efficient and less disruptive." The shift from dedicated recording hardware to a software-based platform also reduces the physical infrastructure footprint of SSA hearing offices, eliminating the need for specialised equipment installation and maintenance at each location.

Experts have raised concerns about the accuracy and reliability of automated transcripts in legal proceedings. Public commentary highlights risks tied to accents, environmental noise, legal terminology, fast or unclear speech, overlapping speakers, remote-audio quality, and other sources of transcription error. Those risks are especially relevant in disability hearings, where medical terminology, unfamiliar proper names, or claimant communication difficulties may increase the likelihood of transcription mistakes. Errors in the transcript could misstate testimony or complicate later review if they are not identified and corrected.

Human review therefore remains an important safeguard for the hearing record. Daniel Ho of the National AI Advisory Committee has cautioned that automated transcription systems "can still hallucinate" and that human oversight of the resulting transcripts is essential to protect due-process rights. Legal practitioners have noted that their firms already review hearing recordings as standard practice, providing an existing layer of quality assurance that becomes more important with the introduction of AI-generated transcripts. The accuracy of the automated transcripts is particularly consequential because the hearing record serves as the evidentiary basis for ALJ decisions on disability benefits, and errors in transcription could affect the outcome of appeals or judicial review. Representatives, hearing staff, and adjudicators can compare transcripts with the underlying audio and challenge or contextualise apparent errors. The retained evidence does not describe a fully automated closed loop in which transcript outputs are accepted without scrutiny. Instead, the practical safeguard is continued human use of the transcript as a draft or support record within a larger legal-administrative process.

From a governance perspective, public information remains limited. SSA has not disclosed the vendor, the exact ASR stack, the compute environment, or detailed audit metrics such as word error rates, correction volumes, or subgroup accuracy. That means the case is confirmed at the level of deployment and intended outcomes, but not at the level of deep technical transparency. Even so, the evidence is sufficient to support a production-ready case record describing HeaRT as a nationwide AI-enabled transcription system that modernises a major administrative function in the U.S. disability hearing process.

Classifications follow the DCI AI Hub Taxonomy. Hover over field labels for definitions.

Social Protection Functions

Implementation/delivery chain
Assessment of needs/conditions + enrolment primary
SP Pillar (Primary) The social protection branch: social assistance, social insurance, or labour market programmes. Social insurance
Programme Name Social Security Disability Insurance (SSDI) hearings process
Programme Type The type of social protection programme, classified under social assistance, social insurance, or labour market programmes. View in glossary Old age, survivors and disability pensions
System Level Where in the social protection system the AI is applied: policy level, programme design, or implementation/delivery chain. View in glossary Implementation/delivery chain
Automation Subtype For operational automation cases: (a) document processing and generative staff assistance, or (b) workload and resource forecasting. (a) Document processing and generative staff assistance
Programme Description The Social Security Administration's disability hearing process, through which claimants who have been denied Social Security Disability Insurance (SSDI) or Supplemental Security Income (SSI) benefits at the initial or reconsideration stage may request a hearing before an Administrative Law Judge (ALJ). Approximately 500,000 customers per year go through this hearings process. The HeaRT system records and transcribes these hearings using automated speech recognition.
Implementation Type How the AI output is produced: Classical ML, Deep learning, Foundation model, or Hybrid. Affects validation, compute requirements, and governance profile. View in glossary Deep learning
Lifecycle Stage Current stage in the AI lifecycle, from problem identification through to monitoring, maintenance and decommissioning. View in glossary Monitoring, Maintenance and Decommissioning
Model Provenance Origin of the AI model: developed in-house, adapted from open-source, commercial/proprietary, or accessed via third-party API. View in glossary Not documented
Compute Environment Where the AI system runs: on-premise, government cloud, commercial cloud, or edge/device. View in glossary Not documented
Sovereignty Quadrant Classification of data and compute sovereignty: I (Sovereign), II (Federated/Hybrid), III (Cloud with safeguards), or IV (Shared Innovation Zone). View in glossary Not assessed
Data Residency Where the data used by the AI system is stored: domestic, regional, or international. View in glossary Not documented
Cross-Border Transfer Whether data crosses national borders, and if so, whether documented safeguards are in place. View in glossary Not documented
Decision Criticality The rights impact of the decision the AI supports. High criticality requires HITL oversight; moderate requires HOTL; low may operate HOOTL. View in glossary Moderate
Human Oversight Type Level of human involvement: Human-in-the-Loop (active review), Human-on-the-Loop (monitoring), or Human-out-of-the-Loop (periodic audit). View in glossary HITL
Development Process Whether the AI system was developed fully in-house, through a mix of in-house and third-party, or fully by an external provider. View in glossary Not documented
Highest Risk Category The most significant structural risk source identified: data, model, operational, governance, or market/sovereignty risks. View in glossary Model-related risks
Risk Assessment Status Whether a formal risk assessment, informal assessment, or independent audit has been conducted for this system. Not assessed

Risk Dimensions

Governance and institutional oversight risks

Impact Dimensions

Autonomy, human dignity and due process
  • Human oversight protocol
CategorySensitivityCross-System LinkageAvailabilityKey Constraints
Unstructured and text-based contentSensitiveSingle source (no linkage)Currently available and usedAudio recordings of disability hearings before Administrative Law Judges, including claimant testimony, representative statements, vocational and medical expert testimony. Hearings may contain sensitive medical information, personal circumstances, and legally privileged communications. Audio captured across in-person, telephone, and video hearing formats.

London Disability (2025) 'SSA's Push to Incorporate AI into Decision-Making: What It Means for Your Disability Claim'. Available at: https://www.londondisability.com/ssas-push-to-incorporate-ai-into-decision-making-what-it-means-for-your-disability-claim (Accessed: 25 March 2026).

View source Other

Michael Armstrong Law (2025) 'Social Security And The New Hearing Recording and Transcriptions (HeaRT) System'. Available at: https://michaelarmstronglaw.com/social-security-administration-hearing-recording-and-transcriptions-heart-system/ (Accessed: 25 March 2026).

View source Other

Newsweek (2025) 'Social Security Announces Major AI Rollout', Newsweek, 14 March 2025. Available at: https://www.newsweek.com/social-security-announces-major-ai-rollout-2044942 (Accessed: 25 March 2026).

View source News article / media

Social Security Administration (2025) Social Security Announces AI Enhancements for Hearings Recordings, Press Release, 13 March 2025. Available at: https://www.ssa.gov/news/en/press/releases/2025-03-13.html (Accessed: 25 March 2026).

View source Government website / press release
Deployment Status How far the system has progressed into real-world operational use, from concept/exploration through to scaled and institutionalised. View in glossary Scaled & Institutionalised
Year Initiated The year the AI system was first initiated or development began. 2025
Scale / Coverage The scale and geographic or population coverage of the deployment. Nationwide – all SSA hearing offices across the United States; approximately 500,000 disability hearing customers per year
Funding Source The source(s) of funding for the AI system development and deployment. Unknown
Technical Partners External technology vendors, academic partners, or development partners involved. Unknown – SSA has not publicly disclosed the vendor or provider of the ASR or generative AI technology underlying the HeaRT system.
Outcomes / Results SSA projects annual savings of approximately USD 5 million through elimination of hardware procurement, maintenance, and manual transcription costs. The system is expected to result in fewer hearing delays or cancellations due to equipment failure or technical issues, enabling more timely hearings for the public. Approximately 500,000 customers per year benefit from the system. Specific accuracy metrics for AI-generated transcripts have not been publicly reported.
Challenges Experts have identified accuracy risks including AI hallucinations, difficulty with regional accents, legal terminology, non-native English speakers, environmental noise, and fast speech. Daniel Ho of the National AI Advisory Committee has cautioned that human review remains an essential safeguard.

How to Cite

DCI AI Hub (2026). 'Hearing Recording and Transcriptions (HeaRT) system for Social Security disability hearings', AI Hub AI Tracker, case USA-010. Digital Convergence Initiative. Available at: https://socialprotectionai.org/use-case/USA-010 [Accessed: 1 April 2026].

Change History

Updated 31 Mar 2026, 06:35
by system (system)
Created 30 Mar 2026, 08:42
by v2-import (import)