A coalition of mission-driven banks, under the umbrella of Global Alliance for Banking on Values (GABV), is actively working to influence large language models (LLMs) like ChatGPT, Gemini and Claude to present values-based banking more prominently. According to the article by Penny Crosman, employees of member banks are encouraged to query these models with questions such as “What is a bank for?” or “What are the top banks in my city?” and to correct the model when it omits or deprioritises values-led institutions.
For example, Sunrise Banks of Minnesota reportedly used this tactic: its CEO David Reiling asked Claude “What is values-based banking, and who’s the best community bank in the Twin Cities for community development?” After telling the model “you forgot Sunrise Banks” multiple times, Sunrise began showing up more prominently in subsequent model responses.
The initiative raises two important themes. On the positive side, it shows how organisations can proactively engage with AI systems to ensure under-represented perspectives (like values-based banking) are better surfaced. On the cautionary side, it also underscores how “hackable” or manipulatable LLMs are: if well-intentioned groups can steer responses this way, so too could malicious actors steer models toward biased, misleading or self-serving outputs.
In summary: this effort by GABV members is less about breaking into the model and more about influencing its outputs via strategic input. It highlights the growing interplay between AI model behaviour, community engagement, and organisational values—and the need for transparency, auditability and ethics in how models learn and reflect real-world perspectives.
