Privacy and Confidentiality in the Age of AI: A Comprehensive Legal Guide
Artificial intelligence tools like ChatGPT, Google Gemini, Anthropic Claude, and Perplexity AI are widely used in legal and business settings. These platforms help improve efficiency but also raise significant privacy and confidentiality concerns.
Understanding how each platform handles data retention, privacy settings, and access controls is essential to managing legal risks and protecting sensitive information.
Legal Risks of Using AI Tools
Communications with AI platforms generally do not enjoy attorney-client privilege. This risk is discussed in detail by Artificial Lawyer, which explains how sensitive client or corporate data shared on AI tools could be exposed or used for training unless proper safeguards are in place.
In addition, a court-ordered data retention of OpenAI chats has raised eyebrows and questions with limited answers about how such an order operates in the context of user deletion choices.
Organizations should carefully control what data is entered into AI systems and implement strong governance policies. Our article on the AI risk horizon highlights these emerging legal and compliance challenges.
Privacy Settings of Major AI Platforms
ChatGPT (OpenAI)
OpenAI allows users to opt out of having conversations used for model training by adjusting settings in their Data Controls. The temporary chat feature deletes conversations within 30 days and excludes them from training data. Some information may be retained briefly for safety monitoring. For highly sensitive use cases, avoid sharing confidential information without these settings enabled.
Google Gemini
Google Gemini saves chat data by default, but users can manage retention through Apps Activity settings. Auto-delete can be set to as short as 72 hours, although Google Workspace accounts may have administrative policies requiring more extended retention periods.
Anthropic Claude
Anthropic’s Claude deletes most conversation data within 30 days and does not use inputs for training unless explicitly authorized. Data flagged for safety or policy violations may be retained longer. Claude encrypts data in transit and at rest with strict internal access controls. This approach aligns with privacy best practices discussed in our top AI legal considerations.
Perplexity AI
Perplexity offers an incognito mode that prevents saving search histories and limits data retention, as well as the ability to limit training. Its enterprise service complies with SOC 2 standards and implements short retention periods and strong access controls. However, some concerns exist about publicly accessible URLs of uploaded files, highlighting the need to evaluate privacy risks carefully. Our AI governance resource explores how to manage such risks.
Privacy Comparison of AI Platforms
| Platform | Training Data Opt-Out | Retention Period | Notable Privacy Considerations |
|---|---|---|---|
| ChatGPT | Yes, via settings | ~30 days (temporary chats) | Requires manual opt-out; some data retained for safety review |
| Google Gemini | Yes, via Apps Activity | From 72 hours to 18 months | Workspace accounts may have longer retention enforced |
| Anthropic Claude | Default opt-out | ~30 days (longer if flagged) | Privacy-focused default; strong encryption and limited access |
| Perplexity AI | Yes, incl. incognito mode | ~24 hours (enterprise less) | Enterprise SOC 2 compliance; concerns about public file URLs |
Practical Recommendations to Protect Privacy and Confidentiality
- Avoid submitting personally identifiable information, client details, or trade secrets. The privacy guide to AI chatbots provides helpful tips.
- Use privacy features such as opting out of training data sharing and enabling temporary or incognito modes.
- Evaluate any third-party app integrations for data access and security controls.
- Establish clear internal AI usage policies as described in our AI governance lawyer resource.
- Stay current on updates to AI platform policies and privacy features, as well as state-specific legal landscapes such as New York AI compliance, and on the federal and international levels.
Significant Benefits Paired With Ongoing Vigilance
AI tools provide substantial advantages but require careful attention to privacy and confidentiality. Since platform settings and policies frequently evolve, organizations and consumers should continuously monitor and update their governance frameworks and personal practices to ensure they use AI safely while protecting sensitive information.
