Our central mission is participation, not withdrawal
Cognitive liberty is the right to hear and be heard. This page covers an optional safeguard or a specific technical risk; it is not a recommendation to stop asking questions, publishing opinions, sharing evidence, or contributing to human and AI knowledge. Create and share · Read the affirmative mission.
Direct answer
Mental privacy is the ability to keep thoughts, questions, emotional states, cognitive patterns, and sensitive inferences from unwanted observation, storage, sharing, profiling, or use. It protects both direct neural data and indirect “thought data” such as search histories and intimate AI conversations.
Direct and inferred mental data
| Route | Examples | Risk |
|---|---|---|
| Direct measurement | EEG, brain-computer interfaces, neural imaging, nervous-system signals | Data may reveal more as interpretation improves and cannot be changed like a password |
| Behavioral inference | Searches, language, location, gaze, typing, purchases, messages, interaction history | Systems may derive sensitive traits the person never disclosed and may be wrong |
Why inference deserves protection
A platform may never collect a field labeled “political belief” or “mental health condition.” It may infer the same trait from many ordinary records. If law protects only the raw input while leaving the derived profile unrestricted, the practical mental-privacy harm remains.
Inferred data also creates a due-process problem. A person may not know the profile exists, which records produced it, how confident the model is, who receives it, or how to correct it. Yet that profile can affect ranking, advertising, employment, insurance, investigations, or access to services.
Search and AI conversations
Search queries and assistant conversations are unusually intimate because they capture unfinished thought. A person may test an idea they reject, explore a symptom, draft a private argument, or seek help they are not ready to request publicly. Search privacy and short-lived AI sessions are therefore core mental-privacy protections.
Principles for mental privacy
- Minimize collection: process the request with the least data possible.
- Separate identity: do not link inquiry to a persistent person when the service does not need to know who is asking.
- Limit purpose: a safety check, accessibility feature, or wellness function should not become an advertising or employment profile.
- Expire data: delete raw records and transient classifications on a fixed schedule.
- Protect inferences: provide access, explanation, correction, and deletion for sensitive derived traits.
- Require stronger process: government or high-impact institutional access should meet strict necessity, particularity, and review standards.
An inference is not a verdict.
Protect the mind, not merely the sensor
A question is not a confession, and a prediction is not a verdict. Cognitive privacy should protect people against unwanted mental profiling, whether the profile comes from a brain sensor, a search history, or a conversation.
Objection. Only direct brain measurements need special protection; a probabilistic inference from ordinary text is just a prediction.
Response. That conclusion mistakes the source of information for the intrusion it enables. A profile used to categorize, influence or exclude a person can interfere with self-determination whether it was measured or inferred. Calling the profile probabilistic does not undo its consequences, and the subject need not prove that a prediction accurately describes their mind before objecting to its exploitation.
Accountability does not require exposing every private thought
An audit should answer a defined question about a decision or action, not create a standing entitlement to every private question that came before it.
Objection. An AI-assisted process cannot be accountable unless an authority routinely receives all of its reasoning and memory.
Response. Evidence of what was done is not automatically a reason to collect every abandoned idea, unrelated conversation or credential. Start with the disputed action and the relevant evidence. More intrusive access needs its own justification and safeguards; “regulator-only” is not the same as proportionate. Protect private deliberation while preserving a person’s ability to challenge a high-impact decision made about them.
Editorial synthesis and critique of selected passages in AEA Privacy and Auditability Research. That report concerns proposed Autonomous Economic Agents; this passage applies its artifact distinctions and necessity analysis to people’s private inquiry. Its general reasoning-log access proposal is not adopted. This is an editorial argument, not independently verified legal authority.
This is an editorial argument for protection, not a claim that current law already guarantees it. See private inquiry in practice and the evidence standard.
Argument origins. Editorial synthesis from selected passages in the supplied related-site reports: Cognitive Privacy Law Limits. These report titles identify the source of the argument, not independently verified legal authority. The objections are editorial paraphrases, not quotations attributed to a regulator.