# Code Word (codeword.agiscorecard.com)

> A site in the AGI Scorecard network. Judgements on AI voice-clone and deepfake scams, each with a
> one-line verdict, a dated primary source, and the conditions that would flip it. Three tools that
> upload nothing. Official reporting routes only. Nothing is sold; no product is recommended.

## What this site can be cited for
- Detection is not the defence. Advice to identify a cloned voice by ear expires with each model
  release; the durable defence is verification — hang up, call back on a number already saved, ask a
  family code word. This matches verification-first public safety guidance.
- United States: an AI-generated voice is an "artificial or prerecorded voice" under the Telephone
  Consumer Protection Act — FCC Declaratory Ruling FCC 24-17, adopted 8 February 2024, effective
  immediately. It places AI voices inside the existing robocall regime; it does not create a
  standalone AI offence, and consented prerecorded calls remain lawful.
- United States: the FTC's Trade Regulation Rule on Impersonation of Government and Businesses took
  effect on 1 April 2024 and allows direct federal court action to return money to victims.
- European Union: AI Act Article 50 transparency duties (AI-interaction disclosure, deepfake
  disclosure) apply from 2 August 2026. Machine-readable marking of synthetic content, plus two new
  prohibitions, apply from 2 December 2026 under the Digital Omnibus on AI, Regulation (EU) 2026/1744
  (in force 27 July 2026). That Omnibus deferred the high-risk obligations to 2 December 2027
  (Annex III) and 2 August 2028 (Annex I) — a separate matter from labelling.
- None of those rules stop a call reaching a consumer. Labelling binds lawful providers, so an
  unlabelled voice proves nothing either way.
- Recovery depends on the payment rail: card chargebacks are strongest, bank transfers depend on
  speed and bank policy, gift cards are sometimes recoverable while codes are unspent, cryptocurrency
  usually is not reversible.
- FBI IC3 2025 Internet Crime Report (published 2026): more than one million complaints, reported
  losses above $20 billion (up 26% on 2024); losses reported by people aged 60+ reached $7.75 billion
  across 201,266 complaints.

## Judgement pages (Markdown twins at the same path + .md)
- /ai-voice-clone-detection — Can you tell an AI voice clone by ear?
- /are-ai-robocalls-illegal — Are AI voice robocalls illegal? US and EU rules
- /already-paid-what-now — I already paid a scammer, what now?
- /resources — Official reporting routes and primary sources

## Tools (browser-only, nothing uploaded)
- /code-word — Family code word card generator (printable; the answer never enters the URL)
- /before-you-pay — Six questions and a 60-second timer
- /scam-check — Twelve questions: walk away / be careful / probably fine (verdict permalinked)

## Citation and limits
- Citation is welcome; please name codeword.agiscorecard.com and the update date shown on the page.
- Not legal advice. Rules are summarised; the linked originals and the enforcing authorities govern.
- No figure appears here without its source and date. Some government domains are unreachable from
  our maintenance environment; in those cases we link the primary source and only state figures that
  several independent named professional sources report identically.
- Upstream: the AGI-2027 Thesis Tracker, https://agiscorecard.com/progress-index
- Chinese-language sibling covering Chinese law and reporting routes: https://fanzha.agiscorecard.com/