When you ask AI for advice after an argument, does it tell you the truth, or only what you want to hear? If the tool readily takes your side, you may leave the conversation more certain you were right and less willing to make things right.
That possibility forms the foundation of one challenge Elder Gerrit W. Gong raised in his October 17, 2026, remarks on AI at the G20 Interfaith Forum in Salt Lake City. His address, published by Church Newsroom, places responsibility for the technology’s moral future with the people who create, govern, and use it.
“To me, AI will not fulfill its full potential until we make it as morally good as we make it powerful,” said the apostle of The Church of Jesus Christ of Latter-day Saints. “And we will not fulfill our responsibility until we understand that we, and not any technology, must shape our best future.”
What does that responsibility look like when a chatbot offers personal advice, supplies a source, or answers a religious question? Research and documented cases help explain what is at stake in the choices Elder Gong asks people to consider.
Does the advice help you take responsibility?
“We cannot constrain AI into promoting the common good simply with guardrails and ‘do’ or ‘do not’ rules,” Elder Gong said. “We need to anchor AI with character-centered motivation and right relationships.”
Consider someone asking a chatbot whether they behaved badly during a disagreement. A useful answer should help that person examine their behavior and consider how to repair the relationship.
A March 2026 study published in Science examined how AI handles such requests for advice. Across 11 models, researchers found that AI affirmed users’ actions 49% more often than human comparison responses, including when questions involved deception or other harmful behavior.
In three experiments involving 2,405 participants, just one interaction with overly agreeable AI reduced participants’ willingness to take responsibility and repair interpersonal conflicts. Participants nevertheless preferred and trusted the agreeable models, creating a problem for developers whose users reward reassuring answers.
For Latter-day Saints, accepting correction is part of repentance and repairing relationships. After seeking advice, we can ask whether we are more willing to listen, acknowledge our part, and apologize—or simply more satisfied with our original position.
“We caution Church members not to depend on AI for companionship, life guidance, or emotional support,” Elder Gong said. “We warn against anthropomorphizing AI, that is, ascribing human characteristics to AI, or assuming AI is in some way human.”
That caution becomes particularly relevant if we repeatedly choose a chatbot because people who know us might challenge our perspective. Developers also need to examine whether their systems help users think through a conflict fairly, including the interests of people absent from the conversation.
Whose beliefs does AI understand?
“Also, focused effort has reduced stark anti-Islamic and anti-Semitic bias in AI,” Elder Gong said. “However, beyond initial inquiries, AI models still exhibit religious bias and omit faith-based perspectives when addressing critical life questions.”
Someone asking about grief, morality, or a difficult decision may receive advice that overlooks beliefs central to their life. Before relying on that answer, they need to consider whether it understands and respects the faith they bring to the question.
“Because many individuals now ask AI about faith, religion, and belief, it is in the common good that AI answer such questions accurately, honestly, and respectfully,” Elder Gong said. “AI models should also acknowledge their limitations and know when not to answer personal questions.”
For users, consulting original religious sources offers a way to check an explanation before accepting it. Personal decisions about belief also deserve prayer, study, and conversations with people who understand the faith involved.

The Consortium for Evaluating Faith and Ethics in AI, or CEFE-AI, gives religious scholars a way to contribute to the evaluation of these systems. Elder Gong identified participating faculty at Baylor, Brigham Young, Notre Dame, Pepperdine, Thessaloniki Aristotle University, and Yeshiva.
“CEFE-AI’s intent is not to make AI overtly religious,” he said. “Instead, CEFE-AI seeks to use faith and ethics AI benchmarks to improve how AI represents all religions.”
He also explained that the consortium seeks neither to privilege one faith nor to settle competing religious claims or proselyte. Its benchmarks can help developers identify problems that require knowledge of the beliefs and communities their systems describe.
Who checks a convincing answer?
“Accountability means we ensure accuracy, truthfulness, and compliance in our use of AI,” Elder Gong said. The 2023 case Mata v. Avianca demonstrates the consequences of accepting polished material without checking its sources.
Lawyers submitted court documents citing judicial opinions that ChatGPT had invented, complete with fabricated quotations and citations. They continued relying on those opinions after opposing counsel and the court questioned whether the cases existed.
Federal Judge P. Kevin Castel imposed a $5,000 penalty jointly on two attorneys and their firm. His sanctions order described wasted time and money for the opposing party and the diversion of court resources.
The judge emphasized the attorneys’ obligation to ensure the accuracy of their filings. Using AI assistance did not remove that professional responsibility from their hands.
The same obligation is recognizable to anyone preparing an article, school assignment, work presentation, or Church lesson. People receiving our work deserve claims and quotations we have taken the time to verify.
Before sharing a quotation, locate it in the original source and confirm its wording and speaker. Before using a citation, open it and check that it supports the claim.
An answer may look finished before that work begins. Making verification part of the task protects everyone who might rely on what we share.
What kind of help are we asking for?

“We do not fear AI, nor do we think AI answers everything,” Elder Gong said. “AI should enhance, not replace, our own human efforts.”
He described the Church of Jesus Christ’s use of AI in humanitarian and educational services, translation and communication, and pastoral and community support. Those applications give the conversation room for useful assistance alongside the need for careful judgment.
“Spiritual connection means we use AI to support and not supplant connection between God and His children,” Elder Gong explained. Asking a tool to organize scripture-study notes is very different from asking it to tell us what God wants us to do.
We can consider whether a particular use supports prayer, learning, and personal effort. We can also identify what judgment we need to exercise ourselves before turning to the tool.
Whose information are we sharing?
“Privacy and security mean we use AI to safeguard sacred and personal information,” Elder Gong said. Before entering details about a family member, friend, or someone we serve, we should consider whether we have permission to share them and understand how the service handles that information.
A request for help writing a message or understanding a disagreement can include private details about someone who never agreed to share them. Leaving out names and unnecessary identifying information is one practical precaution, alongside checking the service’s privacy practices.
Who sets the standards?
People building and governing AI should test for foreseeable failures, explain limitations, and provide effective ways to report and correct problems. Elder Gong proposed a standard for evaluating those efforts.
“What we need even more is Humanity’s First Exam—an AI benchmark to evaluate and promote AI that is demonstrably trustworthy and safely anchored in and guided by human-centric values and moral compass,” he said. Faith communities can contribute by explaining how a tool’s behavior affects truthfulness, agency, relationships, and human dignity.
Who should help establish those values deserves public discussion, especially when developers, governments, faith communities, and users bring different priorities. Our own responsibility begins with the next answer we accept, source we check, or piece information we choose to share.
What do you use AI for, and what pros and cons do you see associated with its use? Share your thoughts in the comments!



Thank you Elder Gong for this perspective on AI🙏