AI and Your Data

Abuse Monitoring Retention When an AI provider keeps prompts and answers for a set time, such as up to 30 days, to look for misuse; it is the main exception to zero data retention. AI Connector (MCP Integration) An AI connector, such as one built on the Model Context Protocol (MCP), is a link that lets an AI assistant read data from or take actions in one of your business apps, using a user's permissions. AI Copyright Indemnity An AI copyright indemnity is a vendor's promise to defend you, and cover certain costs, if someone claims the AI's output infringes their copyright, as long as you meet its conditions. AI Features Enabled by Default AI features enabled by default are new AI tools a vendor switches on in your account without asking first, so your data can start flowing to them before anyone has reviewed the terms. AI Output Ownership AI output ownership is who holds the rights to what an AI tool produces for you. A vendor's contract decides this between you and the vendor, while copyright law decides whether the output is protected at all. AI Subprocessor An AI subprocessor is an outside AI company, such as OpenAI, Anthropic or Google, that one of your vendors uses to process your data when you use the vendor's AI features. AI Training on Customer Data AI training on customer data is when a vendor uses the information you put into its product, such as files, prompts and AI answers, to teach or improve its AI models. AI Training Opt-Out An AI training opt-out is a setting, request or contract clause that tells a vendor not to use your data to train its AI models when training would otherwise be on by default. AI Usage Policy An AI usage policy is a vendor's set of rules for what you may and may not do with its AI features or models. It is different from your own internal policy for staff use of AI. Automated Decision-Making Automated decision-making is when software, often AI, makes a decision about a person, such as a hiring, lending or pricing decision, with little or no human involvement. Fine-Tuning on Customer Data Extra training of an existing AI model on a smaller set of your examples, either to build a model only you use or to improve a model shared with other customers. Human Review of AI Data Human review of AI data means people at an AI vendor, or its contractors, can read your prompts and the AI's answers, for example to check flagged misuse, fix problems or answer legal demands. "Improve Our Services" Clause A contract line that lets a vendor use your data to make its product better; unless AI training is excluded, it can cover training models on your data. ISO 42001 ISO 42001 is an international standard for managing how a company builds and uses AI. A certificate shows an outside auditor checked the company's AI governance process, not that any AI product is safe. Model Provider (LLM Provider) The company that builds and runs the large language model behind a vendor's AI feature, such as OpenAI, Anthropic or Google. Responsible AI Policy A responsible AI policy is a vendor's public statement of how it intends to build and use AI. It often has no contract force, so do not treat it as a promise about your data. Shadow AI Shadow AI is any AI tool people use for work without the company knowing or approving it, such as personal chatbot accounts, browser extensions, note-takers or AI features turned on without review. Zero Data Retention (ZDR) Zero data retention (ZDR) is an AI provider's promise not to store your prompts or its answers after a request is processed, with possible exceptions for misuse checks, saved features and legal holds.