Snowflake's CEO criticised tech companies for not publishing AI hallucination rates, which can range from 1% to 30%. He emphasised that not knowing which part of an AI's answer is incorrect creates a trust problem, especially for critical applications. Snowflake's head of AI identified hallucinations as the biggest obstacle for generative AI use with external customers, suggesting improvements through guardrails and diverse data sources.

**For businesses relying on AI for critical applications:** You must demand transparency from AI vendors about hallucination rates. Without this information, you risk making decisions based on unreliable data, potentially leading to costly mistakes or reputational damage. Implement rigorous testing protocols to validate AI outputs before using them in customer-facing scenarios. **For AI solution providers:** Expect increasing pressure to disclose hallucination rates. Failing to do so may result in lost business to more transparent competitors. Invest in improving your AI models' accuracy and developing clear communication strategies about their limitations.

Expect a surge in demand for AI solutions with verifiable accuracy rates. You'll likely see the emergence of third-party auditing services for AI models, similar to financial audits. Prepare for potential regulations mandating disclosure of AI performance metrics, including hallucination rates. As the market matures, you may need to allocate resources to continuously benchmark your AI solutions against industry standards. Companies that can demonstrate consistently low hallucination rates will gain a significant competitive advantage, potentially reshaping market leadership in AI-driven industries.