London-based Synthesia, one of Europe’s best-funded artificial intelligence companies, has begun letting enterprises generate avatar versions of named employees who can deliver sales pitches and presentations in up to 40 languages. The feature, rolled out to corporate customers this month, effectively allows a company to clone its entire sales team as photorealistic ‘digital twins’ — a development that blurs the already-thin line between legitimate training tool and what critics are increasingly calling deepfake-as-a-service.
The capability builds on Synthesia’s existing avatar technology, which has long allowed users to create AI presenters from a short video sample. What is new is the explicit framing around real, named staff: a top-performing salesperson recorded once in English can now appear to pitch fluently in Japanese, Arabic, German or Portuguese, lip-synced and inflected to sound native. For multinational firms, the appeal is obvious. For privacy and employment specialists, the implications are considerably more fraught.
From training videos to production-grade clones
Synthesia made its name supplying corporate learning and development teams, where avatar-led videos replaced expensive studio shoots for compliance training and internal communications. The shift towards customer-facing sales content marks a significant escalation. These are no longer generic stock presenters but recognisable employees whose likeness and voice are being deployed at scale, often without the audience knowing they are watching a synthetic version.
The company says the feature is gated behind consent controls. Employees must actively agree to have their likeness captured, and Synthesia maintains that avatars cannot be generated covertly. The firm has historically positioned its strict content moderation — blocking news-style impersonation and political content — as a differentiator in an industry plagued by misuse.
Yet the move into multilingual production at enterprise scale tests whether existing consent frameworks can keep pace. “There’s a meaningful difference between consenting to a training video and consenting to a perpetual, multilingual digital twin used in live commercial settings,” said Dr Priya Nandakumar, a researcher in digital identity at the fictional Centre for Synthetic Media Ethics. “Once your avatar is in production, who controls how it’s used, for how long, and what happens when you leave the company?”
The consent question nobody has fully answered
The thorniest issues are likely to surface around employment. If a salesperson’s avatar continues generating pitches after they resign, are they entitled to compensation? Can a company retire an avatar on request, and how quickly? Synthesia says likeness rights remain with the individual and avatars can be deleted, but the contractual reality inside large organisations is rarely so clean.
Employment lawyers note that consent obtained from an employee is inherently complicated by the power imbalance of the workplace. A junior staffer asked to be ‘cloned’ for the good of the team may not feel able to refuse.
“Consent given under economic pressure is a grey area in UK and EU law,” said Marcus Veldt, a fictional senior analyst at advisory firm Halberd Insight. “GDPR treats biometric data as a special category, and a face and voice are about as biometric as it gets. Companies deploying this at scale are taking on regulatory exposure they may not fully appreciate.”
There are also reputational risks. An avatar that performs flawlessly in English may produce subtle errors or culturally tone-deaf phrasing in another language — errors that nonetheless appear to come directly from a named, trusted employee.
A booming market with blurry guardrails
Synthesia is far from alone. The synthetic media market has expanded rapidly, with rivals offering similar avatar and voice-cloning tools. The company’s bet is that responsible, consent-based deployment can capture enterprise demand while keeping it on the right side of an increasingly nervous regulatory environment. The EU AI Act, with its transparency obligations for AI-generated content, looms over the entire sector, as do emerging deepfake disclosure rules in several jurisdictions.
Industry observers point out that the same technology enabling a sales team to scale across languages could, in less scrupulous hands, normalise the idea that the person on screen may not be real — and may not have meaningfully agreed to be there.
- Transparency: Will customers be told they are watching a synthetic avatar rather than a live employee?
- Persistence: What happens to digital twins after an employee departs?
- Accuracy: Who is liable when a multilingual avatar misrepresents a product?
Synthesia maintains that its moderation and consent architecture addresses these concerns, but acknowledges the industry is moving faster than the rules governing it.
What this means
Synthesia’s multilingual avatar feature is a clear signal that AI-generated likenesses of real employees are moving from novelty to operational infrastructure. For businesses, the productivity case is compelling — instant global reach without the cost of reshoots or translation. But the launch forces a question the wider industry has so far dodged: whether consent frameworks designed for occasional training videos can hold up when a worker’s face and voice become reusable, perpetual corporate assets. As regulators sharpen transparency rules and employees grow wary of being cloned, the companies that win will be those that treat consent not as a one-off checkbox but as an ongoing, revocable relationship. The technology is ready for production. The governance around it is not.
Photo by panumas nikhomkhai on Pexels
