AI companies warn users not to rely on chatbot answers in contracts

In short: AI companies often say in their own terms and contracts that generative AI results can be wrong and that users must double-check them.

What's going on

AI companies are not only facing outside criticism about accuracy. Many also include direct warnings in their terms of service and business contracts that you should not blindly trust what their chatbots and other generative AI tools produce.

These documents commonly say AI output can be inaccurate, unreliable, or not suitable for important decisions. In practice, it is like a calculator that sometimes makes up numbers, so you are told to check the math yourself. TechCrunch points to Microsoft Copilot language that frames output as not something to rely on, and similar disclaimers appear across the industry.

Contracts also often shift responsibility to the customer. Some agreements say the provider does not guarantee error-free results, and the user must make sure prompts and outputs follow the law. Mailchimp, for example, warns that AI results may be imprecise or inappropriate, and it puts the burden on users to review what the tool produces.

Another common theme is liability and data use. Some vendors require users to handle risks like copyright problems or harmful content, and some promises to help with lawsuits only apply if users follow specific rules, such as using built-in filters exactly. Terms can also grant the company rights to use your inputs and outputs to improve the AI, which may include personal data, and some policies allow long retention even after an account ends.

What to watch

People and organizations should pay attention to what they agree to before using AI for work that has legal, financial, or health consequences. Regulators like the FTC have warned that quietly changing terms to expand data use for AI training could be unfair, especially if it applies to older data.

Source: TechCrunch AI