Privacy Policy

ParliView — European Parliament Transparency Platform

Effective Date: 20 February 2026
Last Updated: 19 March 2026
Version: 2.1


At a Glance

ParliView is a university research project that helps you explore European Parliament information using AI. Here is a short summary of how we handle your data:

  • You need an account to use ParliView. Registration requires an email address. Your account is managed by AWS Cognito, and your activity on the platform is linked to a pseudonymous user identifier (a UUID) derived from your account.
  • Your data is pseudonymised, not anonymous. We link your platform activity to a pseudonymous identifier rather than your name or email. However, authorised members of the research team can link this identifier back to your account if necessary (for example, to fulfil a data subject rights request).
  • We do not track you across the web. We use essential cookies only and do not use third-party analytics or tracking tools.
  • Your queries are used for academic research only, in pseudonymised or aggregated form. We never sell data or use it for advertising.
  • AI responses are generated by Anthropic Claude (see Section 8). Responses may contain errors. We provide links to original European Parliament sources so you can verify information.
  • You have rights under GDPR. Because your data is pseudonymised (not anonymous), we can identify your records and fulfil data subject rights requests. Details are set out below.

For full details, please read the sections that follow.


1. Who We Are

Data Controller:
University College Dublin (UCD)
Belfield, Dublin 4, Ireland

Principal Investigator:
Dr James Cross
School of Politics and International Relations
University College Dublin
Email: admin@parliview.org

UCD Data Protection Officer:
Email: gdpr@ucd.ie
Phone: +353 1 716 8743

Funder:
Silicon Valley Community Foundation (EUR 1.5 million grant). The funder has no role in research design, data collection, analysis, or publication decisions. The research team maintains full academic independence.

Project Partners:

PartnerLocationRole
University of StrathclydeGlasgow, UKCollaborative research partner
Transparency International EUBrussels, BelgiumCollaborative research partner

Ethics Approval:
This research has been approved by the UCD Human Research Ethics Committee (Reference: 025-HS-26-C-Cross).


2. What Data We Collect

2.1 Account Data (Required for Registration)

To use ParliView, you must create an account via AWS Cognito. We collect:

  • Email address (required)
  • Name (optional, if provided during registration)
  • Phone number (optional, if provided during registration)
  • Email verification status

Your account is assigned a pseudonymous user identifier (a UUID, known as a Cognito sub). This identifier is used throughout the platform to link your activity without routinely exposing your email address or name. Your email address is also used to send platform access and onboarding communications (for example, welcome emails and access confirmations) via Amazon Simple Email Service (SES).

2.2 Platform Interaction Data (Collected Automatically)

When you use ParliView, we automatically collect:

  • Conversation history: The questions you submit and the AI-generated responses you receive, along with navigation state. Conversations are stored and linked to your pseudonymous user identifier. We advise users not to include personal information in their queries. If personal data is inadvertently submitted in a query, it will be treated in accordance with this policy. We will delete identifiable information upon request where technically feasible.
  • Token consumption metrics: Per-user, per-day records of AI token usage (broken down by feature), request counts, and success/failure rates.
  • Usage data: Pages visited, session duration, timestamps, and UI analytics events (action name, timestamp, metadata).
  • Technical data: Browser type, device type, and screen resolution.
  • Group and experiment assignments: Users may be assigned to groups or sub-groups (for example, for A/B testing of platform features), with assignment history tracked.

2.3 Feedback Data

If you submit feedback through the in-app feedback mechanism, we collect:

  • Your feedback text
  • A snapshot of the full conversation history at the time of submission (all preceding messages in that conversation)

This feedback is linked to your pseudonymous user identifier.

2.4 Access Request Data

If you request access to ParliView before being added to the allowlist, we collect:

  • Email address, name, occupation, affiliation, and intended use

This data is retained for the duration of the project plus 5 years (consistent with other data types; see Section 11), even after access is granted or denied.

2.5 Content Policy Data

If your input triggers an automated content policy check (for example, keyword matching), an excerpt of up to 300 characters of the input may be stored along with the matched keyword and user context.

2.6 Infrastructure and Logging Data

The platform generates operational data that is stored in logging and monitoring systems:

  • Application logs (shipped to AWS CloudWatch and OpenSearch): These contain your pseudonymous user identifier, session identifier, request identifier, group/sub-group identifiers, query text, AI response metadata, and latency data.
  • Distributed traces (shipped to AWS X-Ray via OpenTelemetry): These contain your pseudonymous user identifier, session identifier, request identifier, and group identifiers as trace attributes.
  • AWS infrastructure logs (CloudFront, ALB, WAF, CloudTrail): These contain your IP address and HTTP request metadata. These logs are generated automatically by AWS infrastructure services; your IP address is not hashed or anonymised before being logged.
  • Moderation data: If your account is suspended or banned, the reason for the action (ban_reason) and any suspension expiry date (suspended_until) are stored as part of your Cognito account record.

2.7 Survey Data (Optional)

We occasionally distribute opt-in surveys to gather feedback on the platform. If you choose to participate, we collect:

  • Your responses to survey questions (demographic ranges, Likert-scale ratings, open-text feedback)
  • Submission timestamp

Survey responses are pseudonymised. The survey instructs participants not to include identifying information in open-text responses. Any inadvertently provided identifying data will be redacted during analysis.

At the end of the survey, you may voluntarily provide an email address if you wish to be contacted about future research.


3. Voluntary Nature of Data Provision

You are not required by law or by contract to provide any personal data to ParliView. Use of the platform is entirely voluntary. The only consequence of not providing data is being unable to use the platform. Registration requires an email address; if you choose not to provide one, you will not be able to create an account.


4. What We Do NOT Collect

  • Your physical address
  • Financial or payment information
  • Precise location data (country-level location may be inferred from IP address ranges in infrastructure logs)

5. How and Why We Use Your Data

PurposeData UsedLawful Basis
Provide the platform service (authentication, session management, conversation storage)Account data, conversation history, session dataLegitimate interests (Article 6(1)(f))
Improve the platform and evaluate usabilityPseudonymised interaction data, UI analytics events, feedback dataLegitimate interests (Article 6(1)(f))
Understand user needs and query patternsPseudonymised query data, token consumption metricsLegitimate interests (Article 6(1)(f))
Academic research and publicationPseudonymised or aggregated dataLegitimate interests (Article 6(1)(f))
Platform security and performance monitoringTechnical data, infrastructure logs, application logs, distributed tracesLegitimate interests (Article 6(1)(f))
Content moderation and abuse preventionContent policy flags, moderation data (ban reason, suspension status)Legitimate interests (Article 6(1)(f))
A/B testing and platform experimentationGroup and experiment assignmentsLegitimate interests (Article 6(1)(f))
Process access requestsAccess request data (email, name, occupation, affiliation, intended use)Legitimate interests (Article 6(1)(f))
Send platform access and onboarding communicationsEmail addressLegitimate interests (Article 6(1)(f))
Analyse survey feedbackSurvey responsesConsent (Article 6(1)(a))
Contact about future research (if opted in via survey)Email addressConsent (Article 6(1)(a))

Automated access restrictions: The platform includes automated systems that may restrict your access in certain circumstances:

  • An automated content policy that blocks queries matching certain keywords or length rules
  • A pre-registration check that restricts sign-up to email addresses on an approved allowlist
  • Group-based access gating that may deny platform access if you are not assigned to an active group

These are standard platform access controls. They do not constitute automated decision-making producing legal or similarly significant effects under GDPR Article 22, because they do not affect your legal rights or status; they control access to a voluntary research platform.

We do NOT use your data for:

  • Commercial purposes
  • Advertising, marketing, or profiling
  • Sale or rental to third parties
  • Sharing with the funder in any form that could identify individuals

6. Data Minimisation

In accordance with Article 5 of the GDPR, we apply the principle of data minimisation throughout this project:

  • We collect only the data necessary for the stated research purposes.
  • Platform activity is linked to a pseudonymous identifier (UUID) rather than directly to your name or email address.
  • Email addresses are stored in limited locations: the Cognito User Pool, the access request records, and the email allowlist. They are not routinely included in application logs or analytics data (see Section 12 for exceptions and planned remediation).
  • Optional data fields (name, phone number) are not required for registration.
  • We regularly review data holdings to confirm that no unnecessary data is retained.

7. Legal Basis for Processing

7.1 Legitimate Interests (Article 6(1)(f))

For platform interaction data, our lawful basis is legitimate interests. We have conducted balancing tests as required under Article 6(1)(f), grouped by purpose cluster. The following assessments reflect the different data types and risk profiles involved.

7.1.1 Core platform operation

Purposes: Provide the platform service (authentication, session management, conversation storage); process access requests.

Purpose test: The research team has a legitimate interest in operating the platform so that users can access European Parliament information. Processing access requests is necessary to manage participation in the research.

Necessity test: Account data, session data, and conversation history are essential to providing an authenticated, stateful service. Access request data (email, name, occupation, affiliation, intended use) is necessary to evaluate eligibility. We have minimised processing by pseudonymising user identifiers and collecting optional personal details (name, phone) only where voluntarily provided.

Balancing test: The processing does not override users' rights and freedoms. Platform activity is linked to a pseudonymous UUID; only authorised members of the research team can link this to an individual, and only when necessary (for example, to fulfil data subject rights requests). Users are informed of this processing and retain the right to object (see Section 14.2).

7.1.2 Research and analysis

Purposes: Understand user needs and query patterns; academic research and publication.

Purpose test: The research team has a legitimate interest in conducting academic research in the public interest to evaluate tools for democratic transparency and citizen engagement with EU institutions. This interest is shared by the project partners and the wider public.

Necessity test: Understanding how users interact with the platform, the types of queries submitted, and patterns of use is essential to evaluating the platform's effectiveness. Without this data, the research objectives cannot be achieved. Research outputs use pseudonymised or aggregated data only.

Balancing test: The processing does not override users' rights and freedoms. Query data may incidentally reveal political interests, which increases the sensitivity of this processing. To mitigate this: data is pseudonymised; access to raw query data is restricted to authorised researchers; research publications use aggregated or anonymised data; and the enhanced safeguards described in Section 7.3 apply. Users can stop participating at any time, are informed of this processing, and retain the right to object (see Section 14.2).

7.1.3 Platform improvement

Purposes: Improve the platform and evaluate usability; A/B testing and platform experimentation.

Purpose test: The research team has a legitimate interest in improving the platform's design and functionality to better serve its research objectives and user needs.

Necessity test: Pseudonymised interaction data, UI analytics events, feedback data, and group/experiment assignments are necessary to identify usability issues and evaluate design changes. Without this data, platform improvements would be uninformed.

Balancing test: The processing does not override users' rights and freedoms. The data used is pseudonymised interaction data and UI events, which present a low risk to individuals. Group assignments affect only the presentation of platform features and do not impact users' legal rights or status. Users are informed of this processing and retain the right to object (see Section 14.2).

7.1.4 Security and moderation

Purposes: Platform security and performance monitoring; content moderation and abuse prevention.

Purpose test: The research team has a legitimate interest in maintaining the security, integrity, and availability of the platform, and in preventing misuse.

Necessity test: Technical data, infrastructure logs, application logs, distributed traces, content policy flags, and moderation records are necessary to detect and respond to security incidents, monitor performance, and enforce platform rules. Infrastructure logs necessarily contain IP addresses, which are generated automatically by AWS services.

Balancing test: The processing does not override users' rights and freedoms. Security and moderation processing protects all users of the platform. Infrastructure logs contain IP addresses, which are more directly identifying than pseudonymous UUIDs; to mitigate this, these logs are subject to defined retention periods (see Section 11) and are not used for research purposes. Users are informed of this processing and retain the right to object (see Section 14.2).

7.2 Consent (Article 6(1)(a))

For survey participation (see Section 2.7), we rely on your freely given, specific, informed, and unambiguous consent. Consent is obtained through a consent confirmation presented before any survey questions. You may withdraw consent at any time before submitting the survey by closing the survey. Once submitted, survey responses are pseudonymised and may not be identifiable for withdrawal; this limitation is explained before you provide consent.

For optional email provision for future contact, separate consent will be obtained. You may withdraw this consent at any time by contacting admin@parliview.org, and your email will be deleted.

7.3 Special Category Data: Political Opinions (Article 9)

Why this matters: When you ask ParliView questions about parliamentary topics, such as how MEPs voted on particular issues, your queries may incidentally reveal your political opinions or interests. Political opinions are special category data under Article 9 of the GDPR.

Our approach:

We process this data under the research exemption in Article 9(2)(j) (processing necessary for archiving purposes in the public interest, scientific research, or statistical purposes), read together with Article 89(1), which requires appropriate safeguards. This processing is authorised under Irish law by Section 54 of the Data Protection Act 2018, which permits the processing of special categories of personal data for scientific research purposes subject to appropriate safeguards, read together with Section 42 of that Act.

Enhanced safeguards we apply:

  • Queries are linked to a pseudonymous identifier (UUID) rather than directly to the user's name or email address. Only authorised members of the research team can link this identifier back to an individual.
  • Research analysis is conducted on aggregate or pseudonymised data only
  • Access to raw query data is restricted to authorised members of the research team
  • A Data Protection Impact Assessment has been conducted for this processing (see Section 17)
  • Technical and organisational measures are in place as described in Section 12

We recognise the sensitivity of political expression and apply these enhanced safeguards to protect users' rights and freedoms.


8. AI-Generated Content and Transparency

8.1 About ParliView's AI System

ParliView uses a retrieval-augmented generation (RAG) system combined with a large language model (LLM) to generate responses to your queries. This means:

  • When you ask a question, the system retrieves relevant information from official European Parliament data sources (including voting records, plenary proceedings, committee documents, and parliamentary questions)
  • The retrieved information is used to generate a natural language response
  • The system provides citations and direct links to the source European Parliament documents

8.2 AI Model and Provider

  • Language model: Anthropic Claude (Sonnet family), provided by Anthropic
  • Embeddings model: Cohere embed-multilingual-v3, accessed via AWS Bedrock. Cohere embedding models are accessed exclusively through AWS Bedrock and are covered under the AWS data processing agreement. Cohere does not receive data directly from ParliView and is not engaged as a separate data processor.
  • Where inference happens: Language model inference is routed through AWS Bedrock (EU region) where available. A direct Anthropic API integration also exists and may be used for certain requests; in this case, queries are sent to Anthropic's servers in the United States. This transfer is governed by Standard Contractual Clauses (SCCs) under Anthropic's data processing agreement (see Sections 9.1.1 and 10).
  • Training: Your queries are not used to train the language model. Under both the AWS Bedrock and Anthropic API terms of service, customer inputs and outputs are not used for model training.

8.3 Training Data Sources

The AI model is grounded using official, publicly available European Parliament data accessed through the EP Open Data Portal and related institutional sources. This includes structured data on MEP activities, voting records, legislative procedures, and plenary debates.

8.4 Limitations and Verification

  • AI-generated responses may contain errors, omissions, or inaccuracies despite grounding in official sources
  • Responses are not legal, political, or professional advice
  • ParliView is a research tool, not an authoritative source of parliamentary information
  • We display prominent disclaimers alongside AI-generated responses
  • Direct links to original European Parliament source documents are provided so you can verify information independently

8.5 EU AI Act Transparency (Article 50)

In accordance with the transparency obligations of the EU AI Act (Regulation (EU) 2024/1689), we inform you that:

  • Content you receive from ParliView is generated by an artificial intelligence system
  • The AI system uses Anthropic Claude (Sonnet), a general-purpose AI model applied to the specific domain of European Parliament information retrieval
  • The system is designed to assist with information access and does not make decisions that affect your legal rights or status
  • Outputs are clearly labelled as AI-generated on the platform interface

9. Data Sharing

9.1 Project Partners

We share pseudonymised or aggregated data with our project partners under the terms of the ParliView collaboration agreement:

PartnerLocationPurposeData SharedSafeguards
University of Strathclyde Glasgow, UK Collaborative research and analysis Pseudonymised interaction data, aggregated survey data UK adequacy decision (originally adopted 28 June 2021, renewed 19 December 2025, valid until 27 December 2031); data sharing agreement
Transparency International EU Brussels, Belgium Collaborative research on democratic transparency Pseudonymised interaction data, aggregated survey data GDPR applies directly; data sharing agreement

Personal identifiers (email addresses, names) are not shared with partners.

9.1.1 Data Processors

We engage the following data processors to provide infrastructure and services for the platform:

ProcessorServicesData ProcessedSafeguards
Amazon Web Services (AWS) Hosting, authentication (Cognito), AI inference (Bedrock), logging, email delivery (SES), CDN (CloudFront), search (OpenSearch), caching (ElastiCache), and related infrastructure services All platform data, including account data, conversation history, logs, and infrastructure data Data processing agreement under Article 28 GDPR; primary data region eu-west-1
Anthropic Large language model inference via direct API User queries and AI-generated responses. Queries may contain text that incidentally reveals political opinions. Data processing agreement (Article 28 GDPR) incorporating Standard Contractual Clauses (SCCs) for transfers to the United States. Anthropic's API terms prohibit use of inputs/outputs for model training. API inputs are retained by Anthropic for up to 30 days for safety and abuse monitoring, then automatically deleted.

We are working to consolidate all language model inference through AWS Bedrock (EU region). While the direct Anthropic API integration remains active, Anthropic is listed as a separate data processor.

Data transmitted to Anthropic via the direct API: Each API request contains the user's query text, retrieved European Parliament source documents relevant to the query, and system instructions that define the assistant's behaviour. No user identifiers (pseudonymous or otherwise), email addresses, IP addresses, or session identifiers are included in API requests. Queries may incidentally contain text that reveals political opinions (a special category under Article 9 GDPR); the safeguards described in Section 7.3 apply to this processing.

9.2 Academic Archiving

Anonymised research datasets may be deposited in academic repositories (such as the UK Data Service and Zenodo) for future research, in accordance with academic open data principles. Archived data will be fully anonymised (with pseudonymous identifiers removed or replaced) and subject to data access agreements where applicable.

9.3 Publication

Aggregated, anonymised findings may be published in academic journals, conference papers, and research reports.

9.4 We Will NEVER

  • Sell or rent your data to anyone
  • Share identifiable data without consent
  • Share data with the funder (Silicon Valley Community Foundation) in any form that could identify individuals
  • Voluntarily share data with law enforcement; we will only disclose data where compelled by law, and will inform you of any disclosure to the extent we are legally permitted to do so

10. International Data Transfers

User account data, conversation history, and query content are stored within the European Economic Area (EEA), hosted by Amazon Web Services (AWS) in the eu-west-1 (Ireland) region.

However, certain AWS global infrastructure services may process infrastructure-level data (IP addresses, HTTP request metadata, security events) outside the EEA. Specifically:

  • CloudFront (content delivery network) operates globally; access logs may be stored in us-east-1.
  • AWS WAF (web application firewall), when associated with CloudFront, operates globally; WAF logs may be stored in us-east-1.
  • CloudTrail (audit logging) records global service events in us-east-1 regardless of trail configuration.
  • Route 53 (DNS), if used, processes DNS query logs globally.

The data processed by these global services is limited to infrastructure-level metadata (IP addresses, HTTP headers, security events). User query content and account data remain in eu-west-1. These transfers are governed by Standard Contractual Clauses incorporated into AWS's data processing agreement, in accordance with Article 46(2)(c) GDPR.

The only intentional international transfer is to the University of Strathclyde in the United Kingdom. The UK benefits from an adequacy decision by the European Commission, originally adopted on 28 June 2021 and renewed on 19 December 2025 (valid until 27 December 2031), under Article 45 of the GDPR.

Where the direct Anthropic API is used for language model inference (see Section 9.1.1), user queries and retrieved European Parliament source documents are sent to Anthropic's servers in the United States. This transfer is governed by Standard Contractual Clauses (SCCs) incorporated into Anthropic's data processing agreement, in accordance with Article 46(2)(c) GDPR. Anthropic has also certified under the EU-US Data Privacy Framework, providing an additional layer of adequacy commitment. We have conducted a transfer risk assessment and concluded that, given the nature of the data transferred (query text and publicly available parliamentary documents, with no user identifiers), the SCCs and Anthropic's technical and organisational measures provide appropriate safeguards. No user identifiers, email addresses, or IP addresses are included in API requests.


11. Data Retention

Data TypeRetention PeriodAfter Retention
Account data (Cognito)Duration of the project + 5 years (the project is funded until 2028; data will be retained until 2033 at the latest)Securely deleted
Conversation history (DynamoDB)Duration of the project + 5 years. No automatic time-to-live (TTL) is configured; data is retained until manual deletion.Archived in academic repositories (fully anonymised) or securely deleted
Token consumption metrics (DynamoDB)Duration of the project + 5 years. No TTL configured.Securely deleted
Access request data (DynamoDB)Duration of the project + 5 yearsSecurely deleted
Feedback data (DynamoDB)Duration of the project + 5 years. No TTL configured.Securely deleted
Survey responsesDuration of the project + 5 yearsArchived in academic repositories (fully anonymised) or securely deleted
Future contact email addresses (from survey)Until the contact opts out, or end of project + 5 years (whichever is sooner)Securely deleted
Application logs (CloudWatch)90 daysAutomatically purged after retention period
Application logs (OpenSearch)90 daysAutomatically purged after retention period
Session data (Redis/ElastiCache)24-hour TTL (automatically purged)Automatically purged
Distributed traces (X-Ray)30 days (AWS default)Automatically purged
Infrastructure logs (CloudFront, ALB, WAF)90 daysAutomatically purged after retention period
Audit logs (CloudTrail)1 yearAutomatically purged after retention period
Data held by Anthropic (direct API)Up to 30 days (for safety and abuse monitoring)Automatically deleted by Anthropic
Archived research datasetsIndefinitelyRemain in academic repositories in fully anonymised form

After the retention period, data that has not been archived will be securely deleted using industry-standard methods.


12. Data Security

We protect your data through the following technical and organisational measures:

Technical measures:

  • Encryption in transit (TLS/HTTPS) and at rest (AES-256)
  • Authentication via AWS Cognito with secure credential management
  • Pseudonymisation: platform activity is linked to a UUID rather than directly to the user's identity
  • Secure, access-controlled servers within the EEA (Amazon Web Services, eu-west-1 region)
  • Data transmitted to Anthropic's API is encrypted in transit via TLS. Anthropic's data processing agreement includes technical and organisational security measures; details of Anthropic's security practices are published at trust.anthropic.com.
  • Two-factor authentication for all researcher access to data systems
  • Regular security audits and vulnerability assessments

Pseudonymisation model and known exceptions:

The platform is designed so that the pseudonymous user identifier (UUID) is the primary key used in application logs, conversation records, and analytics. Email addresses and other directly identifying information should not appear in these systems. However, the following exceptions currently exist:

  • Application logs: Certain log statements generated during login events and administrative actions currently include email addresses. We plan to remediate these before public launch by replacing email with the pseudonymous identifier.
  • Email allowlist and access request tables (DynamoDB): These store email addresses as a primary key by design (the allowlist must contain the email to function, and access requests are submitted before a pseudonymous identifier exists).
  • Session store (Redis/ElastiCache): The full Cognito user profile (including email) is stored server-side in the session. This data is transient (24-hour TTL), encrypted in transit, and not accessible to other users.

Organisational measures:

  • Data processing agreements with all data processors and project partners
  • Access to research data restricted to authorised team members on a need-to-know basis
  • Staff training on data protection and information security
  • Records of processing activities maintained in accordance with Article 30 of the GDPR
  • Incident response procedures for potential data breaches (see Section 16)

13. Cookies and Storage

13.1 Essential Cookies

ParliView uses a limited number of cookies that are strictly necessary for the platform to function:

Cookie NamePurposeDurationTypeAttributes
sidMaintains your authenticated session. Session data (including your Cognito user profile) is stored server-side in Redis; only a random identifier is stored in the cookie.24-hour rolling TTLEssentialHttpOnly, Secure, SameSite
Cognito Managed Login cookiesSet by AWS during the OAuth authentication flow. These cookies are managed by AWS and are necessary for the login process to function. ParliView does not control their content or duration.Varies (set by AWS)EssentialSet by AWS

Essential cookies do not require consent under Article 5(3) of the ePrivacy Directive because they are strictly necessary for the provision of the service you have requested.

13.2 Client-Side Storage

The platform uses browser storage mechanisms for functional purposes:

  • sessionStorage: Used to preserve group assignment parameters across the OAuth login redirect (for example, pendingGroupUuid, pendingGroupId, pendingEventCode). This data is cleared automatically when you close the browser tab.
  • localStorage: Used as an offline queue to retry failed conversation saves. This data persists until successfully sent or manually cleared. The offline retry queue ensures that conversation data you have already submitted is not lost due to transient network failures. We consider this strictly necessary for the reliable provision of the service, as without it your submitted data could be silently discarded.

13.3 Analytics

We do not use any third-party analytics tools on this website. No analytics data is shared with external services, and no third-party analytics scripts are loaded when you visit the site. UI analytics events (see Section 2.2) are collected by our own infrastructure.

13.4 Third-Party Resources

The ParliView application does not load resources from third-party servers. All fonts, stylesheets, and scripts are bundled and self-hosted. However, the Cognito Managed Login pages (used during the authentication flow) are served by AWS and may load their own resources. We do not control the content of these pages.

13.5 What We Do NOT Use

  • Advertising or marketing cookies
  • Third-party tracking cookies
  • Social media cookies or plugins
  • Fingerprinting or other cross-site tracking technologies

14. Your Rights Under GDPR

14.1 Summary of Your Rights

Under the General Data Protection Regulation, you have the following rights. Because your data is pseudonymised (not anonymous), we can identify your records using your Cognito account and fulfil these rights. We will respond to any valid request within one month of receipt, in accordance with Article 12(3) of the GDPR. If your request is complex or we receive a high volume of requests, we may extend this by up to two further months, and we will inform you of the reason for any delay within one month.

RightGDPR ArticleDescription
AccessArticle 15Request confirmation of whether we process your data, and obtain a copy
RectificationArticle 16Request correction of inaccurate personal data
ErasureArticle 17Request deletion of your personal data (“right to be forgotten”)
RestrictionArticle 18Request that we limit processing of your data
Data PortabilityArticle 20Receive your personal data in a structured, commonly used, machine-readable format. Note: this right applies only to data processed on the basis of consent (Article 6(1)(a)) or contract (Article 6(1)(b)). Most ParliView data is processed under legitimate interests and may not be subject to portability.
ObjectArticle 21Object to processing based on legitimate interests (see Section 14.2)
Withdraw ConsentArticle 7(3)Withdraw consent for survey participation or email contact at any time (see Section 14.4)
ComplaintArticle 77Lodge a complaint with a supervisory authority

14.2 Your Right to Object (Article 21)

You have the right to object to our processing of your data where that processing is based on legitimate interests (Article 6(1)(f)).

If you object, we will stop processing your data unless we can demonstrate compelling legitimate grounds for the processing that override your interests, rights, and freedoms, or the processing is necessary for the establishment, exercise, or defence of legal claims.

To exercise your right to object, contact admin@parliview.org.

14.3 Erasure: Technical Limitations

We are committed to fulfilling erasure requests. However, you should be aware of the following technical limitations:

  • Conversation data uses a soft-delete mechanism: records are marked as deleted but may remain in the database until a purge process removes them. We are implementing an automated purge job to permanently remove soft-deleted records.
  • Access request records are retained for the duration of the project plus 5 years for audit purposes.
  • Application logs (CloudWatch, OpenSearch): Individual records cannot be selectively deleted from these logging systems. Log data will be automatically purged when the configured retention period expires.
  • Infrastructure logs (CloudFront, ALB, WAF, CloudTrail): Individual records cannot be selectively deleted. These logs are purged according to their configured retention periods.

When you submit an erasure request, we will: delete your Cognito account, soft-delete your conversation and feedback data (and permanently purge it as soon as technically feasible), remove your email from the allowlist, and confirm what data remains in logging systems along with its expected retention period.

14.4 How to Withdraw Consent

Email contact consent: You may withdraw your consent to being contacted about future research at any time by emailing admin@parliview.org. Your email address will be deleted within 30 days of your request.


15. How to Exercise Your Rights

To exercise any of your data protection rights, you may contact:

Principal Investigator:
Dr James Cross
School of Politics and International Relations
University College Dublin
Email: admin@parliview.org

UCD Data Protection Officer:
Email: gdpr@ucd.ie
Phone: +353 1 716 8743

Process: We will acknowledge your request within 5 working days and provide a substantive response within one month (extendable by up to two months for complex requests, with notification). We may ask you to verify your identity before processing your request. There is no fee for exercising your rights.

Supervisory Authority:
If you are unsatisfied with our response, you have the right to lodge a complaint with:

Data Protection Commission (Ireland)
21 Fitzwilliam Square South, Dublin 2, D02 RD28
Website: www.dataprotection.ie
Email: info@dataprotection.ie
Phone: +353 1 765 0100 / 1800 437 737

You may also complain to the supervisory authority in your EU/EEA Member State of residence or place of work.


16. Data Breach Notification

In the event of a personal data breach that poses a risk to your rights and freedoms, we will:

  • Notify the Data Protection Commission (Ireland) within 72 hours of becoming aware of the breach, in accordance with Article 33 of the GDPR
  • Notify affected individuals without undue delay where the breach is likely to result in a high risk to rights and freedoms, in accordance with Article 34 of the GDPR
  • Document the breach, its effects, and remedial actions taken in our internal breach register

Because user data is pseudonymised, the impact of a breach involving application-level data (conversation history, usage metrics) would be limited, as the data is not directly identifiable without access to the Cognito User Pool. However, we maintain full breach notification procedures as a matter of legal compliance and good practice.


17. Data Protection Impact Assessment

A Data Protection Impact Assessment (DPIA) has been conducted for this project in accordance with Article 35 of the GDPR. A DPIA is required because the processing involves:

  • New technologies (AI-generated content based on large language models)
  • Data that may incidentally reveal political opinions (a special category under Article 9)
  • Large-scale processing of interaction data from a public-facing platform

The DPIA identified the risks associated with special category data processing and confirmed that the safeguards described in this policy (pseudonymisation of user identifiers, restricted access, encryption, data processing agreements) reduce residual risks to an acceptable level.

The DPIA is being reviewed to ensure it fully reflects the current data model, in which users are authenticated via Cognito and data is pseudonymised rather than anonymous. This review will be completed before public launch. The DPIA specifically considers the risks arising from the transfer of user query data to Anthropic in the United States via the direct API path, including the adequacy of Standard Contractual Clauses as a transfer mechanism and the implications of a third-party processor handling data that may incidentally reveal political opinions (a special category under Article 9).

The DPIA is available for review by the Data Protection Commission upon request.


18. Policy Acknowledgement

At present, access to ParliView is restricted to pre-approved research participants via an email allowlist. Participants are informed about data processing through this privacy policy and the participant information sheet.

For the planned public launch, we will implement a post-login acknowledgement interstitial that requires users to acknowledge this privacy policy before using the platform. Acknowledgement records will include the version of the policy acknowledged and the date of acknowledgement. This section will be updated once the acknowledgement mechanism is in place.

This acknowledgement is a transparency mechanism; it does not constitute consent under Article 6(1)(a) GDPR. The lawful basis for processing remains legitimate interests as described in Section 7.


19. Children

ParliView is not intended for use by individuals under 18 years of age. This threshold reflects both the requirements of the ethical approval granted for this research project and a precautionary approach to ensure compliance with data protection requirements for minors across all EEA jurisdictions, where the digital age of consent varies between 13 and 16. We do not knowingly collect data from children. If we become aware that data has been collected from a person under 18, we will take steps to delete it.


20. Changes to This Policy

We may update this privacy policy from time to time to reflect changes in our practices, legal requirements, or project activities. When we make changes:

  • The “Last Updated” date at the top of this page will be revised
  • Significant changes will be highlighted on the platform with a notice
  • Previous versions of this policy will remain accessible on request

We encourage you to review this policy periodically.


21. Contact Us

For questions about this privacy policy or your data:

Dr James Cross
School of Politics and International Relations
University College Dublin
Email: admin@parliview.org

For data protection queries:

UCD Data Protection Officer
Email: gdpr@ucd.ie
Phone: +353 1 716 8743

For complaints about data processing:

Data Protection Commission (Ireland)
21 Fitzwilliam Square South, Dublin 2, D02 RD28
Website: www.dataprotection.ie
Email: info@dataprotection.ie
Phone: +353 1 765 0100 / 1800 437 737


This research has been approved by the UCD Human Research Ethics Committee (Reference: 025-HS-26-C-Cross).

ParliView website: https://www.parliview.org