As digital technologies evolve, GRESB has integrated Artificial Intelligence (AI) tools in a responsible and controlled manner to enhance efficiency, consistency, and scalability across a range of activities, including data analysis to support insight generation, analytical workflows, and operational processes. AI is used to support human expertise and decision-making rather than replace it.
This statement outlines how GRESB applies AI, the principles guiding its use, and the safeguards in place to protect participant data and ensure fairness.
Our Guiding Principles For Responsible AI Use
Transparency and accountability
At present, AI outputs are advisory and supplemental to human judgment and do not replace existing processes. Human oversight remains embedded in workflows across GRESB products and operations. As AI capabilities evolve, GRESB will continue to evaluate the appropriate balance between automation and expert review, ensuring that quality, fairness, and assurance standards are maintained.
Human oversight
AI tools may assist GRESB teams in tasks such as extracting, classifying, translating, and summarizing information, identifying patterns in datasets, or supporting internal analytical workflows, including within validation processes. This includes assisting in the review of participant-submitted evidence and generating high-confidence acceptance recommendations for straightforward data points. This enables human reviewers to focus more time on edge cases and gray areas and supports consistency across decisions, assets, and managers.
AI tools do not have the authority or system permissions to take actions that lead to point deductions that could reduce a participant’s score without human review and approval. Any AI-supported output that could negatively affect a participant’s score (including rejections or point deductions) is reviewed by qualified human reviewers, including GRESB’s assurance partner (SAS) where applicable, before a decision is finalized.
In addition, GRESB performs ongoing quality control over AI-supported acceptance outcomes, including reviewing a portion of high-confidence acceptance cases through sampling and targeted spot checks, to validate accuracy and continuously improve controls and guidance.
Purpose-driven application
The application of AI tools focuses on areas where they can demonstrably enhance efficiency, consistency, or linguistic accuracy within the relevant operational or product processes. Examples include language translation and automated relevant text extraction and summarization to support reviewers’ understanding of participant-submitted evidence.
Data protection and security
GRESB enters into contractual agreements with AI service providers to ensure appropriate data protection, confidentiality, and information security safeguards consistent with its internal data governance policies.
All data processed through AI systems remains protected under GRESB’s data governance policies and terms and conditions. For language translation, GRESB uses the Google Translate API, which ensures that data is transmitted securely and is not retained for model training. For large language model (LLM)-based processing, GRESB uses providers with published data governance commitments (e.g., OpenAI’s API and Gemini for Google Cloud). These providers state that prompts and outputs submitted via their business/API services are not used to train models by default (unless an organization explicitly opts in), and they describe retention and security controls applicable to the service. GRESB does not opt in to having its data used for training purposes.
Continuous evaluation
Given the rapid pace of technological development, GRESB regularly reviews AI-enabled processes to ensure they remain reliable, ethical, and aligned with its data-quality objectives. Any new AI deployment undergoes internal review before integration into workflows across GRESB products and operations.
Through these principles, GRESB ensures that AI strengthens, rather than replaces, the integrity and consistency of its products and operational processes, helping to deliver fair, high-quality benchmarks to the market. GRESB monitors relevant regulatory developments, including the EU Artificial Intelligence Act and applicable ESG ratings regulations. GRESB’s governance framework and public disclosures will evolve as necessary to remain aligned with legal and industry standards.
FAQ: GRESB’s Use of AI
-
GRESB uses AI to improve the efficiency and consistency of its products and operational processes, including data validation. AI assists in areas such as document translation, information extraction, and quality-control checks, helping relevant GRESB teams manage large volumes of data and evidence without compromising accuracy or fairness.
-
No. AI is used to support the validation process (e.g., extraction, classification, translation, and summarization of evidence) and may automatically apply high-confidence acceptance outcomes for straightforward data points under defined controls. Any outcome that could negatively affect a participant’s score (including rejections or point deductions) is reviewed by qualified human reviewers, including GRESB’s assurance partner (SAS) where applicable, before it is applied. GRESB also performs ongoing quality control over AI-supported acceptance outcomes, including the review of a portion of high-confidence acceptance cases through sampling and targeted spot checks.
-
Before integration into operational workflows, AI tools undergo internal review to assess their reliability, data protection safeguards, and operational value. GRESB continues to monitor AI-enabled processes and may adjust or discontinue their use if they do not meet quality or governance standards.
-
GRESB primarily uses established AI technologies provided by trusted third-party providers and integrated into secure operational environments. These tools support internal workflows and are selected based on their ability to meet GRESB’s data protection, reliability, and governance requirements.
-
All AI-enabled processes follow GRESB’s strict data protection standards. GRESB only uses systems that meet secure data-handling requirements and contractual safeguards. For example, Google Translate API does not store or use submitted text for model training, and ESGDS services operate within protected environments where data access and retention are tightly controlled.
-
GRESB uses the Google Translate API for automated document translation and ESGDS-integrated LLM services for text interpretation and classification tasks. These are established, secure technologies provided by industry-recognized models such as Claude, ChatGPT, and Gemini.
-
No, AI models do not have the autonomy to determine the validity of any scheme, standard, or certification that has not been pre-approved by GRESB. Any new application for a scheme or certification currently involves a human validator, with AI tools serving as aids to identify patterns, locate key information, or translate text. Only human validators have the ability to verify and decide on the final outcome.
-
GRESB continuously reviews the quality and reliability of AI outputs through sample testing, manual review, and alignment with human scoring results. GRESB retains only AI processes that meet quality standards and demonstrate measurable benefits without introducing bias.
-
Since AI is used only in supportive, non-deterministic steps (e.g., translation and text extraction), Members’ data privacy and validation outcomes remain unaffected. There is therefore no need for opt-out mechanisms, as AI LLM outputs do not and cannot override human review in the current validation framework.
-
GRESB continues to explore responsible AI applications that improve operational workflows, including validation quality and efficiency. Any future use of AI will remain consistent with its guiding principles: human oversight, data protection, transparency, and fairness.
-
Yes. GRESB provides training and guidance to employees on the responsible use of AI tools and maintains internal policies and controls for AI use within its operations. These practices cover appropriate use cases, data handling expectations, quality standards, and escalation processes where human review is required. GRESB applies human oversight not only in our workflows across GRESB products and operations, but also across internal use of AI to help ensure consistency, accountability, and alignment with its responsible AI principles.
-
GRESB monitors regulatory developments, including the EU Artificial Intelligence Act and other applicable regulations relevant to ESG ratings and digital governance. GRESB’s current AI applications support internal workflows across GRESB products and operations and do not involve fully automated decision-making that produces legal or similarly significant effects. AI outputs operate within a human-led framework. GRESB will continue to review its governance approach and public disclosures as regulatory guidance evolves and as its AI capabilities develop.