User-submission principles
For your submission, AI can reduce paperwork. It cannot replace your confirmation or proposal review.
These principles govern AI assistance shown while a person prepares their own submission. They describe limits—not a promise that every possible assistance feature is currently available—and are separate from Claimitt's internal editorial scanner.
No silent submission
In user submission flows, AI-assisted tools do not submit a public-event claim, event proposal, field report, or supporting answer on your behalf. You review and intentionally submit your own information.
No invented answers or promises
AI output must not fill a gap by inventing facts, sources, eligibility, legal conclusions, or promised outcomes. A suggestion is not evidence and is not legal or financial advice.
Extraction requires confirmation
When an AI-assisted feature organizes or extracts information from material you provide, the result remains a draft. You confirm or correct it before it becomes part of a submission.
Risk flags lead to human review
For user-submitted proposals and field reports, an automated risk or missing-information flag can ask for clarification or route work to a reviewer. It does not autonomously reject a user or publish that user's submission.
Sources stay visible
AI summaries do not replace the underlying record. Claimitt's evidence standard prioritizes primary and attributable sources, and uncertainty remains visible when the record is incomplete.
A manual path remains available
Users can provide information directly and request support. AI assistance is not permission to bypass review, safety, rights, privacy, or content rules.
Safety and security controls are separate
These AI limits do not prevent Claimitt from blocking malware, prohibited content, abuse, unsafe uploads, or unauthorized access under its terms and safety rules. Those controls are not evidence that a claim is true or false.
Automated source monitoring is labeled separately
Claimitt's administrator-controlled scanner can create event drafts from configured public-source feeds. When an administrator selects autonomous mode, it can publish a qualifying Claimitt-created event record without a person approving each individual draft. Manual and assisted modes send publication actions to an approval queue, and a missing or unreadable mode setting fails back to that assisted queue.
Scanner-created editorial records are not user submissions and do not speak for a user. Qualification uses configured status, uncertainty, and source-count controls; an autonomous fallback can use an explicit single-source override. Read the resulting record with its sources and uncertainty—not as a guarantee that every reported fact is settled.