Provenance record

AI ‘deepfakes’ give US political attack ads a new twist

Financial Times: Artificial Intelligence (tier 1, news) 2026-10-09T04:00:12.000Z Original ↗

Source note: Headlines and standfirsts only; full text is paywalled.

Discourse valence

-70

Strongly adverse

confidence 50% · 4 items · range -100 to -45

-100

-50

+50

+100

-70

Adverse readingFavourable reading

Consensus of 4 models from different labs. Spread 55 points, agreement low.

Governance and regulation Synthetic media and manipulation Concentration of power

Excerpt as ingested

Manipulated images are filling the airwaves in the run-up to November’s midterm elections

Every model that read this

Model	Provider	Stage	Score	Conf.	Latency	Prompt	When
Llama 3.3 70B	Meta	analysis	-75	80%	5624ms	v1.0.0 / m1.0.1	2026-10-09 06:17
Claude Sonnet 5	Anthropic	consensus	-45	40%	3896ms	v1.0.0 / m1.0.1	2026-10-09 06:17
Gemini 2.5 Flash	Google	consensus	-100	0%	1697ms	v1.0.0 / m1.0.1	2026-10-09 06:17
Mistral Small 3.1 24B	Mistral AI	consensus	-60	80%	8533ms	v1.0.0 / m1.0.1	2026-10-09 06:17

Llama 3.3 70B · reading

AI deepfakes in political ads can mislead voters

evidence: reported horizon: n/a

Claude Sonnet 5 · reading

The item describes AI deepfakes proliferating in political attack ads, a known risk to electoral integrity and public trust. However, the text gives no detail on scale, verifiability, or actual voter impact, so the implication is concerning but not extreme on current evidence.

evidence: reported horizon: n/a

Gemini 2.5 Flash · reading

evidence: reported horizon: n/a

Mistral Small 3.1 24B · reading

The use of AI-generated deepfakes in political advertising can mislead voters. This undermines democratic processes. The increase in deepfakes in political ads is concerning.

evidence: speculative horizon: n/a

Evidence extracted

AI 'deepfakes' are used in US political attack ads
Manipulated images fill the airwaves in the run-up to November's midterm elections

The chain

SOURCE     Financial Times: Artificial Intelligence (tier 1)
   ↓
DOCUMENT   1cacb698-b179-4f7c-89ad-0db671776513
           https://ft.com/content/42fea52e-a4de-4613-b409-ca52d0a2aac2
   ↓
EVIDENCE   2 extracted excerpts
   ↓
MODEL RUN  4 runs, methodology 1.0.1
   ↓
SCORE      -70  (Strongly adverse)
   ↓
CONFIDENCE 50%