Google Gemini Was Allegedly Used to Process Student Data Amid India's Central Board of Secondary Education Portal Vulnerabilities

The Hindu reported that vulnerabilities in the OnMark exam-marking portal used by India's Central Board of Secondary Education (CBSE) allegedly exposed sensitive student data, including answer-sheet images. Ethical hacker Nisarga Adhikary also alleged that COEMPT Eduteck quality-assurance scripts processed students' personal information through Google Gemini. CBSE said the vulnerabilities had been contained.

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 processing of non-public sensitive student data and answer sheets by Google Gemini without student consent, violating data privacy and sovereignty.

Additional risk subdomains

• 6.5 Governance failure: The third-party vendor processed sensitive student data using a US-based AI service without student consent, violating data privacy laws and demonstrating a failure of governance.

Causal factors

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

The privacy risk was caused by human developers intentionally integrating Google Gemini into automation scripts, leading to the unexpected exposure and unauthorized transfer of sensitive student data.

EU AI Act risk tier

• Risk tier: 2 High Risk

High Risk: The AI was used in CBSE's digital evaluation ecosystem to process student exam data and answer sheets. This falls under Risk Level 2 as it involves 'Educational and vocational training systems' that affect student grading and academic outcomes.

AI system and alleged parties

• AI system: Gemini (Google)
• AI purpose: Automatic Skill Assessment; Content Verification
• Behaviour type: Tool
• Alleged developer: Large language model developers, Google
• Alleged deployer: Government of India, COEMPT Eduteck, Central Board of Secondary Education
• Alleged harmed parties: Students in India, Students, Privacy, Minors in India, Minors, Educational communities, Central Board of Secondary Education students

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 Negligible
• Epistemic: direct Negligible, indirect Negligible
• Child sexual exploitation and abuse: direct Negligible, indirect Negligible

Privacy

Reported: The report explicitly describes privacy violations where sensitive student data was processed by Google Gemini without consent.

Directly caused: The report states that sensitive personal information of students was processed by Google's Gemini in automation scripts prepared by quality assurance engineers of COEMPT without student consent, violating data privacy laws.

Inferred additional harm: It is likely that the personal data of approximately 1,500,000 students was transmitted to servers in the United States, violating national data sovereignty principles, though the exact number of affected individuals is not confirmed.

People affected

• Occurrences reported: 1
• People reportedly exposed: 1500000

Potential causes

Management

• Poor Vendor Oversight: CBSE did not monitor COEMPT Eduteck's insecure cloud and AI usage.
• Delayed Incident Response: CBSE initially rejected claims and ignored follow-up reports.
• Inadequate Infrastructure Management: CBSE poorly managed its IT infrastructure and security safeguards.

Technology

• Unsecured Public AWS Buckets: Answer sheets stored in public AWS buckets without authentication.
• Weak Password Security: The passwords used across the portal were easy to guess.
• Vulnerable Portal Architecture: Active, exploitable vulnerabilities left open in the OnMark portal.

Data Inputs

• Unprotected Sensitive Student Data: 9.3 million rows of sensitive student data were left unprotected.
• Raw Student Data in QA Scripts: Personal student data fed directly into Gemini automation scripts.

Human Factors

• Weak Password Selection: Users/admins set easy-to-guess passwords on the ecosystem.
• QA Engineers Sending Data to Gemini: Engineers used Gemini to process sensitive student data in scripts.

Process and Methods

• Inadequate Vulnerability Patching: CBSE did not respond to or patch active vulnerabilities for months.
• Insecure Data Storage Practices: Vendor stored answer sheets on public AWS buckets to save cost.
• Lack of AI Data Privacy Controls: No checks before sending sensitive student data to Gemini in the US.

Regulatory Environment

• Data Sovereignty Violations: Student data was sent to the US for processing, violating local laws.
• Lack of Consent for Data Processing: Third-party AI processed student exam data without student consent.

Information quality

• Classification confidence: High
• Reason for confidence: The report provides clear, direct statements from the ethical hacker who discovered the vulnerability, along with official statements from the CBSE acknowledging the security flaws. The role of Google Gemini in the automation scripts is explicitly described, leaving little ambiguity about the nature of the privacy and security issues.
• Ambiguities identified: The exact number of students whose data was processed by Gemini (as opposed to just being stored in the unsecured AWS bucket) is not explicitly quantified, only the total database size of 9.3 million rows and columns is provided.
• Alternative interpretations: One could argue this is primarily an IT security incident rather than an AI safety incident, but the integration of Gemini to process sensitive student data without consent directly links it to AI data governance risks.

An ethical hacker exposed vulnerabilities in India's CBSE digital evaluation system, revealing that a third-party vendor used Google Gemini scripts to process sensitive student data, sending it to US servers without consent. While the technical flaws have been contained, the incident highlights critical data sovereignty and third-party AI integration risks for government entities.

• Overall national security impact: Minor
• Scope: Single nation
• Primary target: India
• Alleged perpetrator: COEMPT Eduteck

Threat characteristics

• Imminence: Long-term. The immediate technical vulnerabilities have been contained, leaving long-term strategic concerns regarding data governance and third-party AI integration.
• Autonomy: Human-controlled. The AI (Google Gemini) acted as a tool integrated into automation scripts designed and executed by human quality assurance engineers.
• Novelty: Evolved capability. Represents an evolved threat where traditional data leaks are compounded by the silent integration of external AI APIs that transfer sensitive data across borders.

Impact by dimension

• Physical security: Negligible. No kinetic threat, physical systems compromise, or critical infrastructure impact occurred in this digital data exposure incident.
• Information security: Minor. While sensitive student data was exposed and processed by a US-based AI model, it did not involve classified intelligence, military systems, or active disinformation campaigns.
• Sovereignty: Minor. The incident involved unauthorized processing of national education board (CBSE) data by a foreign AI system, highlighting data sovereignty concerns, but was contained without disrupting core government operations.
• Economic security: Negligible. The exposure did not target strategic industries, financial systems, or critical economic supply chains, resulting in negligible economic security impact.
• Societal stability: Minor. The incident compromised the privacy of millions of students whose data was processed without consent, representing a notable privacy failure but without causing civil unrest or systemic human rights abuses.