The thesis is one sentence, and it belongs on the wall of every marketing team piloting AI avatars this year. The laws that landed this month did not ban synthetic likeness. They moved its cost from production to compliance, and for the one use case B2B teams actually reach for, cloning a real person from your own team, the math no longer clears.
What actually happened
Three things converged in the last three weeks, and together they change how a careful B2B team should price synthetic video.
On May 19, the enforcement section of the TAKE IT DOWN Act went live. Covered platforms now have to remove flagged non-consensual intimate imagery, including AI-generated deepfakes, within 48 hours of a valid request, and the FTC is the enforcer, with civil penalties running north of $53,000 per violation. Say this plainly so nobody panics for the wrong reason: that law is about intimate imagery, not your product explainer. It is not the statute that will bill your marketing team.
The one that will is the NO FAKES Act, and a revised version landed back in Congress this month. It would create the first federal right of publicity in the United States and a “digital replica” right: protection over a newly created, computer-generated, highly realistic representation that is readily identifiable as the voice or visual likeness of an individual. Read that definition twice. It is a precise description of what an enterprise avatar tool produces the moment you clone a real person on your team. The right survives death and runs up to seventy years.
And on May 5, an actress sued a major studio and its director, alleging her face was used as the structural basis for a digital character without proper consent. Different scale, same principle now being argued in public: a real person's likeness has owners, and “we generated it” is not the same as “we cleared it.”
Why this is a B2B problem, not a Hollywood one
Most marketing leaders will read those headlines and file them under entertainment law. That instinct is the expensive one.
The avatar pitch for B2B was always cost. Clone the founder once, generate a hundred localized product videos, skip the shoot. The savings are real on paper. What the pitch never priced is that the asset you just created is a digital replica of a real, identifiable person, and that category is exactly what the new legal regime is built to govern.
Here is the part that catches teams. The consent your avatar vendor collects protects the vendor, not you. A face and a voice are biometric data under GDPR and a growing list of state laws, and consent given to a platform does not automatically flow to its business customers. When that executive leaves, or simply changes their mind, you inherit the obligations: deactivate the avatar for new use, honor sunset clauses on everything already published, delete the underlying biometric data, and document that you did. Those duties sit with the brand that published the video. They do not sit with the SaaS tool that rendered it.
Read your avatar vendor's terms before you assume they carry this for you. In most of them the indemnification runs the other way: you warrant that you hold the rights to every likeness you upload, and you agree to hold the vendor harmless if you did not. The tool renders the face. You own the exposure.
So the cost did not disappear when you skipped the shoot. It moved. It moved from a line item you understand, a production day, to one most marketing teams have never run, a likeness-licensing and data-retention workflow with legal in the loop.
The math nobody put in the deck
Two numbers, one ours and one from outside.
Ours first, framed honestly. Across our retainer book, north of twenty active B2B SaaS engagements, every client that piloted a cloned-spokesperson avatar in the last eighteen months has paused or killed it. For most of last year I would have told you the reason was quality, the uncanny stiffness everyone can spot. It is not anymore. On the three most recent, the reason logged was legal review.
And when a client asked us to scope a cloned-executive program properly last quarter, we built the wrapper a defensible version actually needs: a written likeness license with a defined term, a revocation-and-deletion clause, a disclosure standard for every output. That scoping and paperwork ran more billable hours than the half-day live shoot it was meant to replace. The shoot also handed the client a year of owned B-roll. The avatar would have handed them a license that expires the day the executive updates their LinkedIn.
Run the comparison on durability, not just unit cost. A shoot produces assets whose legal status is settled the day you wrap. You have the footage, the release, the usage rights, and none of it evaporates when someone resigns. A cloned avatar produces assets whose legal status is conditional and ongoing, a license you have to maintain, monitor, and unwind on demand. One is a purchase. The other is a subscription with a compliance tail. Teams keep comparing the first render cost and missing that difference entirely.
The outside number sets the stakes. Sixty-one percent of B2B content marketers plan to increase video investment in 2026, and more of them rank video highest for ROI than any other format. Demand is climbing, which is exactly why the cheap-scale promise of avatars is seductive, and exactly why getting the compliance math wrong scales too. The FTC penalty framework on the adjacent TAKE IT DOWN regime already runs past $53,000 per violation. You do not want to be the team that learns what the NO FAKES equivalent looks like by testing it.
The counter-argument, steelmanned
The strongest case against this thesis is that I am overcooking a risk the tools already handle. Enterprise avatar platforms do collect consent. They do offer disclosure features. The EU and FTC frameworks lean heavily on labeling, and labeling is cheap. If you clone your own willing CEO and tag the output as AI-generated, where is the exposure?
Two places. Disclosure law and likeness law are different bodies of law. A clear “this is AI” label answers the deception question the FTC cares about. It does nothing for the publicity-and-biometric question, which asks whether you hold a real, scoped, revocable license to that person's face and voice, and whether you can prove it. A Slack message saying “sure, use my avatar” is not that license.
Second, the willing CEO is the easy case until it is not. People leave. Acquisitions happen. A founder's amicable yes becomes a departed founder's lawyer asking why their face is still selling software. The entire value of the avatar was durability, a hundred videos from one capture. That durability becomes the liability the moment consent turns out to be a moment rather than a contract.
The honest version of the counter-argument survives in one place, and I will concede it. A fully licensed synthetic presenter who is not a real, identifiable member of your team, used for high-volume, low-trust work like course localization or internal training, is a fine tool with a manageable risk profile. That is a genuine use case. It is just not the one most B2B teams reach for when they get excited about avatars. They want the founder's face. That is the expensive one now.
What to do Monday
Start with an inventory. Find every place a real person's cloned face or voice already appears in your published video, and identify who at the company can produce the signed license that covers it. If the answer is nobody, you have just found this quarter's quiet liability.
Split your avatar use into two buckets and treat them differently. Fully licensed synthetic strangers for volume work can stay. Cloned real teammates move behind a legal gate before another clip ships.
Reprice the program with the real costs in it. Add the consent paperwork, the term-and-revocation workflow, the deletion obligations, and the disclosure standard to the line item, then compare it honestly to a half-day live shoot you can repurpose across a year. For the founder-and-exec content that anchors trust, the shoot usually wins on both cost and risk once the wrapper is priced in.
Default your highest-trust content to real footage captured in a repeatable system. The face that builds buyer confidence should also be a face you own outright, not one you license back from a vendor on terms that can change.
And if you do scale avatars, put the license term, the revocation-and-deletion clause, and the disclosure standard in the contract before the first render, not after the first complaint.