A year ago, Mithril Security exposed the risk of AI model manipulation with PoisonGPT. To solve this, we developed AICert, a cryptographic tool that creates tamper-proof model cards, ensuring transparency and detecting unauthorized changes in AI models.
