Alleged ChatGPT Misuse by Contractor Leads to Reported Data Exposure in New South Wales Resilient Homes Program

A former contractor of the New South Wales Reconstruction Authority reportedly uploaded a spreadsheet containing personal and health information of Resilient Homes Program applicants to ChatGPT during a three-day period in March 2025. Up to 3,000 people may have reportedly been affected.

Risk classification

• Primary risk domain: 2 Privacy & Security
• Primary risk subdomain: 2.1 Compromise of privacy by obtaining, leaking or correctly inferring sensitive information

The incident involved the unauthorized upload of non-public sensitive personal and health information of up to 3,000 individuals to a third-party AI platform, directly compromising their privacy.

Additional risk subdomains

• 5.1 Overreliance and unsafe use: The contractor unsafely used an unauthorized public AI tool to process sensitive government data without understanding the privacy implications.

Causal factors

• Entity: Human
• Intent: Unintentional
• Timing: Post-deployment

The data breach was caused by a human decision when a contractor uploaded sensitive files to ChatGPT post-deployment, resulting in an unintentional exposure of personal information.

EU AI Act risk tier

• Risk tier: 3 Limited Risk

Limited Risk (Level 3): The system involved is ChatGPT, a chatbot. The report states 'ChatGPT is an online AI tool... It allows users to ask questions or upload information to help generate written content or ideas', which fits the definition of a chatbot requiring transparency.

AI system and alleged parties

• AI system: ChatGPT (OpenAI)
• AI purpose: Chatbot; Writing Assistant
• Behaviour type: Assistant
• Alleged developer: OpenAI
• Alleged deployer: OpenAI
• Alleged harmed parties: Resilient Homes Program applicants, Resilient Homes Program, Government of New South Wales, General public of New South Wales, General public of Australia, General public

Harm severity

Severity ratings are based on this scale.

Highest direct severity in any category: Substantial. Severity is scored from Negligible to Catastrophic in each harm category, for harm the reports describe as caused directly or indirectly by the AI system.

• Physical: direct Negligible, indirect Negligible
• Infrastructure: direct Negligible, indirect Negligible
• Property: direct Negligible, indirect Negligible
• Financial: direct Negligible, indirect Negligible
• Environmental: direct Negligible, indirect Negligible
• Malicious content: direct Negligible, indirect Negligible
• Differential treatment: direct Negligible, indirect Negligible
• Civil rights: direct Negligible, indirect Negligible
• Democracy: direct Negligible, indirect Negligible
• Privacy: direct Substantial, indirect Negligible
• Psychological: direct Negligible, indirect Minor
• Epistemic: direct Negligible, indirect Negligible
• Child sexual exploitation and abuse: direct Negligible, indirect Negligible

Privacy

Reported: The report explicitly describes a major data breach involving the personal and health information of up to 3,000 individuals.

Directly caused: A former contractor uploaded a spreadsheet containing names, addresses, email addresses, phone numbers, and health information of up to 3,000 applicants to ChatGPT.

Inferred additional harm: Although there is no evidence of third-party access yet, the data remains on OpenAI's servers and could potentially be used for model training or accessed in future breaches.

Psychological

Reported: The report explicitly mentions potential distress caused to those engaged with the program.

Indirectly caused: The Reconstruction Authority acknowledged and apologized for the distress and concern the data breach may cause to the applicants.

Inferred additional harm: It is likely that some of the 3,000 affected individuals experienced anxiety or distress regarding the potential exposure of their personal and health information.

People affected

• Occurrences reported: 1
• People reportedly exposed: 3000

Potential causes

Management

• Insufficient Contractor Oversight: Lack of strict monitoring over contractor data handling and tool usage.
• Incomplete Cyber Security Training: Mandatory training on data obligations for AI tools was not fully updated.

Technology

• Unsecured AI Platform Use: Contractor uploaded sensitive data to ChatGPT, an unauthorized online tool.
• Lack of Technical Safeguards: No system blocks were initially in place to prevent uploads to AI tools.

Data Inputs

• Unencrypted Sensitive Spreadsheet: A Microsoft Excel sheet with 12,000 rows of personal data was uploaded.

Human Factors

• Unauthorized Contractor Action: A former contractor manually uploaded sensitive client data to ChatGPT.

Process and Methods

• Inadequate System Controls: Internal systems did not prevent the export and external upload of data.
• Lack of Clear AI Guidance: Clear guidance on non-sanctioned AI tools was not initially issued.

Information quality

• Classification confidence: High
• Reason for confidence: The reports are official government statements and reputable news articles detailing the specific dates, the nature of the data, the AI tool used, and the estimated number of affected individuals. There is little ambiguity regarding the sequence of events or the primary risk domain.
• Ambiguities identified: It is not fully clear how OpenAI processes or retains data uploaded by contractors, or if the data was actually used for training.
• Alternative interpretations: None. The incident is clearly a data privacy breach caused by human misuse of an AI tool.

A former contractor for the New South Wales Reconstruction Authority uploaded sensitive personal and health data of 3,000 flood victims to ChatGPT. While representing a significant domestic privacy breach and highlighting government data governance challenges, the incident has negligible national security implications and is manageable through standard administrative procedures.

• Overall national security impact: Minor
• Scope: Single nation
• Primary target: Australia
• Alleged perpetrator: Former contractor

Threat characteristics

• Imminence: Long-term. The immediate breach has been contained with system blocks, leaving only long-term data governance and privacy monitoring concerns.
• Autonomy: Human-controlled. The AI tool was passive; the data exposure was entirely initiated and executed by a human contractor.
• Novelty: Established threat. Data leakage through unauthorized employee uploads to public generative AI tools is a well-documented and common modern security risk.

Impact by dimension

• Physical security: Negligible. No physical systems, kinetic threats, or critical infrastructure were targeted or affected by this data upload.
• Information security: Minor. Involves the unauthorized transfer of sensitive citizen personal and health data to a foreign commercial entity, though no espionage or intelligence compromise is indicated.
• Sovereignty: Minor. Represents a minor lapse in government data control and contractor oversight, but does not threaten core state authority or electoral processes.
• Economic security: Negligible. No strategic technology, critical supply chains, or financial systems were compromised or targeted in this incident.
• Societal stability: Minor. Compromised the privacy of up to 3,000 citizens, causing potential distress, but did not result in mass surveillance or systematic human rights violations.