Harm Ledger · verified · Computer Science
On 21 January 2024 thousands of New Hampshire voters received spoofed calls using an ElevenLabs-cloned Biden voice telling them not to vote; the FCC later fined Steve Kramer $6 million and Lingo Telecom $1 million.
Current score
−0.18
Thousands of deceptive AI-voice political calls with no demonstrated turnout or election effect are Significant civic harm (tier 2), not Major. This is not a risk index. Voice model plus human organizer (0.45). FCC forfeiture is strong evidence. Realization is the calls themselves. Transient election event.
The deepfake message urged Democrats to save their vote for November. Kramer was also criminally charged in New Hampshire. Election officials and the FCC treated the campaign as voter suppression. Turnout effects were not isolated in a causal study, but the calls and enforcement are documented.
Cloned a Biden-like voice used in the robocall audio.
The synthetic voice was generated with ElevenLabs; Kramer designed and placed the calls.
Attribution 0.4500 · Credit share 50% · ElevenLabs
Spoofed robocalls used a generative Biden voice to discourage New Hampshire primary voting, leading to FCC forfeitures.
outcome · supported
primary sources
Federal Communications Commission · regulatory record · 30 Sep 2024
FCC · regulatory record · 30 Sep 2024
independent sources
UK AISI Blog · other · 17 Sep 2026
13 Sep 2026 · 0.00 → 0.61
Initial adjudicated seed score under methodology 0.1.
14 Sep 2026 · 0.61 → 0.18
Methodology 0.2: realized civic incident without demonstrated electoral effect is Significant harm, not Major.
17 Sep 2026 · 0.18 → 0.18
Methodology 0.2: realized civic incident without demonstrated electoral effect is Significant harm, not Major.
17 Sep 2026 · 0.18 → 0.18
Ingest update: added a follow-up source.
18 Sep 2026 · 0.18 → 0.18
Methodology 0.2: realized civic incident without demonstrated electoral effect is Significant harm, not Major.
19 Sep 2026 · 0.18 → 0.18
Methodology 0.2: realized civic incident without demonstrated electoral effect is Significant harm, not Major.
21 Sep 2026 · 0.18 → 0.18
Methodology 0.2: realized civic incident without demonstrated electoral effect is Significant harm, not Major.
6 Oct 2026 · 0.18 → 0.18
Methodology 0.2: realized civic incident without demonstrated electoral effect is Significant harm, not Major.
7 Oct 2026 · 0.18 → 0.18
Methodology 0.2: realized civic incident without demonstrated electoral effect is Significant harm, not Major.
11 Oct 2026 · 0.18 → 0.18
Methodology 0.2: realized civic incident without demonstrated electoral effect is Significant harm, not Major.
11 Oct 2026 · 0.18 → 0.18
Methodology 0.2: realized civic incident without demonstrated electoral effect is Significant harm, not Major.
21 Oct 2026 · 0.18 → 0.18
Methodology 0.2: realized civic incident without demonstrated electoral effect is Significant harm, not Major.
24 Oct 2026 · 0.18 → 0.18
Methodology 0.2: realized civic incident without demonstrated electoral effect is Significant harm, not Major.
27 Oct 2026 · 0.18 → 0.18
Methodology 0.2: realized civic incident without demonstrated electoral effect is Significant harm, not Major.
28 Oct 2026 · 0.18 → 0.18
Methodology 0.2: realized civic incident without demonstrated electoral effect is Significant harm, not Major.
