Politifex logoPolitifex
All fact checks

Fact check

AI analysis
“I trained the avatar to discuss venture fraud”
DisputedConfidence: MODERATE

Reasoning

The creator and independent outlets (The Verge, a personal blog, and a demo video) claim the avatar was trained on venture‑fraud material, while the platform’s FAQ and FactCheck.org report state the avatar lacks such specialized data and provides generic answers. The contradictory findings prevent verification of the training claim.

On confidence: Evidence is mixed with credible sources on both sides, preventing a definitive conclusion

Important context

The claim concerns whether the avatar was actually trained on venture‑fraud content, not its performance. Self‑reported training may not reflect the underlying model’s knowledge base, and independent analyses differ on the presence of such data.

Evidence

Supporting (3)

  • Tier 1 — Primary source
    Training My Avatar on Venture Fraud

    "I fed the model transcripts of my 2024 conference keynote on venture fraud, plus a curated set of case studies, so the avatar can answer any question about fraud in the VC space."

  • Tier 2 — Independent reportingindependent origin
    I created an interactive digital avatar of myself — and you can talk to it

    The creator says he trained the avatar using a custom dataset that includes his own talks on venture capital fraud, allowing the avatar to field detailed questions on the topic.

  • Tier 2 — Independent reportingindependent origin
    Avatar Demo: Venture Fraud Q&A

    In the demo the avatar correctly explains the 2022 XYZ venture fraud case, describing the red flags and regulatory outcomes.

Contradicting (2)

  • Tier 2 — Independent reportingindependent origin
    Avatar Platform FAQ

    "Our avatars are not trained on specialized financial advice, including venture fraud, to avoid providing potentially misleading information."

  • Tier 3 — Secondary reportingindependent origin
    Fact‑Check: Avatar Cannot Discuss Venture Fraud

    Independent analysis found that the avatar’s knowledge base contains no data on venture‑capital fraud, and its responses on the topic are generic and inaccurate.

Limitations

Unable to directly inspect the avatar’s training dataset; reliance on reported statements and secondary analyses, which may be outdated or incomplete.

Last verified:
Sep 26, 2026, 3:20 PM CDT
Pipeline:
0.1.0
Claim type:
Factual

Where this claim appeared

I created an interactive digital avatar of myself — and you can talk to it

TechCrunch