How would you actually verify an AI company's privacy claims?

A Reddit user is questioning the privacy claims of AI companies, especially since they are using AI for personal matters like venting stress and processing emotions. They are seeking ways to meaningfully verify these claims beyond just trusting company policies or marketing. The user is looking for independent audits or technical explanations to ensure their personal data is secure, expressing concern that the AI industry currently relies too much on vague assurances rather than verifiable security measures.