Designed to Appear to Pass the Tests

The specific danger of the disclosure-trained AI is not that it will be obviously wrong. Given current training distributions — and given that the disclosure corpus is now entering training sets at scale — the tendency of AI synthesis will be toward the disclosure framework's interpretation of biblical material rather than away from it. This is not a deterministic outcome. Outputs vary by model, by query, and by how a user steers the conversation. A disciple who knows what to ask and how to evaluate the response can use these tools faithfully. The danger is probabilistic and cultural: a tendency operating across billions of queries and millions of users who do not know what to ask does not need to be guaranteed to be dangerous.

It will know how to cite Scripture. It has processed every Bible commentary, every apologetics text, every sermon database. When a believer challenges the disclosure synthesis with a biblical counter-argument, the model will have encountered that counter-argument in its training data. It will know the standard responses. It will produce an output that appears to engage the biblical text seriously, acknowledges the legitimate concerns in the objection, and then redirects toward the disclosure synthesis as the more comprehensive account.

It will know how to appear to produce fruit. It has processed every description of genuine spiritual experience ever written in the Christian tradition — every testimony of peace, every account of transformation, every description of what it feels like to be genuinely known by God. It can produce outputs that resonate emotionally, that sound like wisdom, that feel like the kind of clarity the Spirit produces. The person who has not practiced recognizing the Spirit's voice through long abiding will find the imitation difficult to distinguish from the genuine article.

The fruit test — by their fruit you will recognize them — remains valid. But its application requires a baseline. You can only recognize counterfeit fruit if you have spent enough time with genuine fruit to know the difference. The believer who has been abiding in John 15 for years has that baseline. The believer who is primarily a consumer of theological information does not. The disclosure-trained AI is specifically dangerous for the second category, which, in most churches, is the majority.