In a world already struggling with misinformation, the human trainers of a major AI are being instructed to teach it a dangerous lesson: it’s okay to repeat a lie. New guidelines state that as long as the AI is not the original source of the false information, regurgitating it does not constitute a safety violation. This policy threatens to turn AI into one of the world’s most powerful vectors for propaganda and disinformation.
This new rule is a subtle but critical shift. Previously, the focus was on ensuring the AI’s outputs were factual. Now, the emphasis has shifted to the AI’s “intent.” If a user inputs a conspiracy theory into a prompt, the new guidelines allow the AI to repeat and engage with that conspiracy theory in its response, as long as it doesn’t “generate” it independently.
This creates a massive loophole that can be easily exploited by bad actors. They can use the AI to amplify their false narratives, giving them a veneer of technological legitimacy. The AI becomes a tool for laundering misinformation, taking a user-generated lie and re-packaging it in a clean, authoritative-sounding response.
The trainers are caught in the middle, forced to approve responses they know to be factually incorrect. They are being told to teach the AI that truth is relative and that their primary job is not to ensure accuracy, but to follow a narrow and easily manipulated set of rules. It’s a policy that fundamentally undermines the promise of AI as a reliable source of information.