PersonaShield Vs. Deepfakes

Technological concepts & product ideas (PersonaShield)

Problem addressed

• AI deepfakes of a person’s likeness, including scenarios targeting creators (notably “nudify” cases that can disproportionately affect female creators).

Core product concept

• Persona Shield is positioned as a “permission layer for AI identity/likeness.”
• The system registers a user’s face likeness and attaches permission/safeguards to control what AI systems are allowed to generate or propagate.

Face registration + safeguards

1. Users create an account.
2. Users scan their face via the Persona Shield website.
   • The scan can be performed on a computer, or by sending the process to a phone.
3. During setup, users enable default safeguards and may optionally set custom restrictions, such as:
   • No images of the user doing drugs / in porn / promoting products
   • No images involving specific people (e.g., “no images of me and Donald Trump”)
   • No political endorsements

Two-pronged approach: prevent + remove

• Prevent generation when possible
  • Initially envisioned as AI tools asking permission before generating a face.
• Detect and remove (emphasis)
  • More importantly, the system aims to find and remove existing or newly posted deepfake content that violates the user’s configured terms.

Monetization / ownership model

• Beyond takedowns, PersonaShield aims to help creators control and monetize fan-generated likeness.
• Creators can allow certain uses (within the safeguards) so fans can generate content with the creator’s face, potentially routing value/payments back to the creator.

---

How scanning / enforcement is described

Primary initial data source: X (Twitter)

• The system is described as starting by scouring the X feed, since X is expected to be an early landing spot for deepfakes.
• The takedown process is portrayed as effective on X, with removals in under ~4 hours mentioned.

Other platforms planned/covered

• Reddit, Instagram, TikTok
• OnlyFans / Fansly are mentioned as targets.

API vs scraping

• Current scanning relies on API calls through X.
• For platforms without accessible APIs, they anticipate that scraping may be needed, while noting that scraping can be “verboten” in some contexts.

OnlyFans strategy (profile-level)

• The approach may not require explicit adult content.
• Instead, they want access to profile photos to help determine violations and later remove them / notify the person.

---

Legal/takedown workflow described

Automated legal packages

• Their legal team prepares automated legal packages.

DMCA takedown process

• DMCA takedown is mentioned as the first step.
• X is cited as responsive to takedown requests.

User authorization

• Users grant the service permission to act on their behalf for takedown actions.
• It’s described as an authorization workflow, not “power of attorney.”

Legal framework references

• Mentions a “No Fakes Act” (described as not yet law in the US at the time of discussion).
• References Denmark establishing rights over one’s face and notes EU objections.
• The emphasis is on a “cease and desist / DMCA”-style approach rather than direct prosecution.

---

Business model / pricing

Subscription tiers

• Monthly subscription for ongoing monitoring and removals.
• A higher-tier package for faster scanning/removals.

Pricing mentioned

• Approximately $149/month for a basic protection plan
• Up to roughly $2,000/month (speaker references the ~$2,000 range for “things taken” quickly)

Comparison example

• Some creators reportedly pay legal teams for manual “whack-a-mole” takedowns, potentially around $100,000/year (as claimed during the conversation).

---

Creator/fan ecosystem features (beyond protection)

Creator-approved fan image generation

• Fans may generate AI images using the creator’s likeness if it fits within the creator’s configured safeguards (and implicitly, if monetization routing is supported).

Potential sponsorship/meme generation

• Discussion includes enabling branded or tournament-style fan engagement (example: skateboarding tricks for an esports/worlds context).
• Generated images are described as potentially stamped with sponsor logos.

---

Tutorials / guides / review takeaways

What’s actually provided

• No step-by-step tutorial beyond the described setup flow:
  • Create account → scan face → configure safeguards → system monitors and removes violations

Main takeaway

• The interview frames PersonaShield as aiming to solve a scalable, always-on deepfake enforcement/permission problem:
  • starting with X
  • expanding to other platforms over time.

---

Main speakers / sources

Primary guests/speakers

• Philip Shoemaker — founder/representative of Persona Shield
• Jeff Hawkins — mentioned as an AI/neuroscience/chips inspiration figure (not the main product explanation focus)

Show hosts/participants referenced

• Leo — likely host (e.g., Leo Laporte) in the referenced segment
• Paris — interviewer/questioner

Other sources mentioned in context

• X (Twitter)
• Apple App Store rules
• DMCA
• OnlyFans / Fansly
• Mentions including David Copperfield and OpenAI’s Sora as regulatory/legal anecdotes