Relying too much on ML models seems risky. If an AI model is established as authoritative in the field, a forger who gets access to the model data could create

Relying too much on ML models seems risky. If an AI model is established as authoritative in the field, a forger who gets access to the model data could create adversarial fakes that look absolutely genuine to the AI. Bribe the right person and nobody knows that the ML data has leaked, so the fake’s credibility isn’t easily challenged.