AI use disclosure

Last updated: May 2026.

Aclamos uses generative AI to help operators run shows and ballots. This page explains where AI is involved, how it's constrained, and how to opt out. It supports our obligations under the EU AI Act (Reg. (EU) 2024/1689), Executive Order 14110, and emerging U.S. state laws on AI disclosure (e.g. Colorado AIA SB24-205, California ADMT).

Where AI is used

• Eligibility checks — when an admin opts in, AI summarizes a nomination against the show's eligibility rules. The admin always makes the final call.
• Judge briefs — produces a one-page summary for judges per nomination from the fields they'd otherwise read end-to-end.
• Polish & translation — proposes copy edits or translations on text the producer wrote. Output is suggested, not auto-published.
• Plagiarism & GenAI risk — flags suspicious nomination text. The platform never auto-rejects on this signal.
• Show Director Co-Pilot — generates a starter ballot, rubric, pricing tiers and form schema from a brief. The producer reviews and edits before publishing.
• Ballotis fraud detection — Claude reads vote patterns and flags coordinated/automated voting. Suspicious votes are quarantined for human review, not silently dropped.
• Voter outreach copy — drafts reminder emails for the producer to approve and send.

Always-human-in-the-loop decisions

Decisions that are never made by AI alone:

• Eligibility determinations.
• Judging scores and shortlist decisions.
• Vote tallies and election outcomes — counting is deterministic code, not LLMs.
• Account suspensions, content takedowns, fraud determinations.
• Refunds and payouts.

Where the EU AI Act applies and a decision could be construed as “significantly affecting” a person, you have the right to (a) be told an automated decision is being made, (b) obtain meaningful information about the logic, and (c) request human review. Email privacy@aclamos.app.

Model & data handling

• Default model: Anthropic Claude (Sonnet 4.6 / Opus 4.7 depending on task).
• Prompts and outputs are processed under Anthropic's zero-retention, no-training policy. They are not used to train Anthropic models.
• We send only the fields needed for the task. We never send payment information, password hashes, 2FA secrets, or audit logs to model providers.
• Outputs are reviewed by the producer before any visible-to-end-user publication, except for transparent operational helpers (e.g. a “suggest similar nominees” chip).

Opt-out

• Show producers: turn off AI features in /[org]/[show]/settings. The platform remains fully functional without them.
• Nominators & voters: AI is not directed at you personally. Your free-text fields may be processed by AI for plagiarism/fraud signals; the producer who runs the show decides whether to enable that. You can decline to participate or contact the producer to opt out.

Synthetic-content marking

Where Aclamos generates synthetic media on a producer's behalf (e.g. a winner announcement card draft), output includes provenance metadata (C2PA where supported) and a visible label by default.

Contact

Questions about how AI is used: ai@aclamos.app. Privacy or rights requests: privacy@aclamos.app or use /privacy/request.