← Course Map All Courses
Module 06 · 4 lessons · 5 quiz Qs

The Risks

What can be taken from you, and how to see it coming.

Long-exposure waterfall in deep forest

Photograph via Unsplash

Lesson 6.1 · Voice and Likeness Cloning

Voice. Anyone with five seconds of clean recorded audio of you speaking can produce a serviceable clone of your voice in 2026 using ElevenLabs or similar tools. The clone improves with more reference material, but the threshold for “convincing enough for an advert” is low. If you have ever appeared on a podcast, posted a YouTube video, or recorded a voice note that someone else kept, you are clonable.

Likeness. Image-to-image generators can take a photograph of your face and produce convincing variations of you in any setting, wearing any outfit, saying any words. Deepfake video is no longer the resource-intensive process it was in 2020.

The risks specifically. (1) Your voice or face used in adverts you did not consent to. (2) Your likeness used in fraudulent business contexts, such as fake testimonials. (3) Synthetic content of you used to harass, defame, or extort.

The defences. Watermarking and provenance metadata on every audio and video file you publish. C2PA Content Credentials are emerging as the standard. Public statements on every platform you use clarifying which channels are real. Legal: register your name and image as a trademark where commercially relevant.

Lesson 6.2 · Style Imitation and Authorship Theft

A model can be fine-tuned on any body of work that is publicly accessible. Once fine-tuned, it can produce variations on that work indefinitely. The cost of fine-tuning a small model dropped below US$100 in 2024 and continues to fall.

This is the painter-in-Mona scenario. It is also the photographer-style scenario, the writer-voice scenario, the producer-signature-sound scenario, and so on. Anything with a recognisable creator signature is targetable.

The defences in order of strength. Glaze and Nightshade on portfolio pieces (free, deployed pre-publication). Watermarking with traceable metadata. Public attribution on every piece. Brand recognition (the imitation is exposed when audiences can name the original). Legal recourse via copyright or unfair competition.

The most under-used defence is brand recognition. Most working creatives in Jamaica are not famous enough for an imitation to be obvious, so increasing your profile is now defence work rather than vanity.

Lesson 6.3 · Wage Compression and Lost Briefs

Wage compression. The lower-skilled portions of your work being done at one-fifth your rate by someone with AI tools and a Caribbean cost base. This is happening now in motion graphics, copywriting, basic graphic design, basic photography, and entry-level analysis.

Lost briefs. Clients putting early-stage briefs through AI to test ideas internally before commissioning anyone. Some of those briefs that would have come to you in 2022 are not coming in 2026 because the client got “good enough” output from ChatGPT for free.

The defences. Move up the value chain into work that requires direct presence, deep relationship, and authoritative judgement. Restructure your pricing to include the strategic layer your client cannot replicate with a chat tool. Rebuild your client base around fewer, higher-value relationships rather than many low-fee ones.

A note on contracts. Add the clause that your work cannot be used as input to train any model without separate compensation.

Lesson 6.4 · The Risk Map

Plot every risk on two axes: likelihood and severity.

The high-likelihood, high-severity risks for a Jamaican creative in 2026 (the “act first” quadrant): wage compression, lost briefs, and style imitation.

The high-likelihood, lower-severity risks (the “manage” quadrant): client edits to your work post-delivery, pitch leakage, and minor copyright violations on social platforms.

The lower-likelihood, high-severity risks (the “prepare anyway” quadrant): voice clone fraud, likeness theft for defamation, total replacement in your specific niche.

The lower-likelihood, lower-severity risks (the “monitor” quadrant): occasional accidental AI editing by a junior, isolated style-imitation by an obscure account.

The discipline. Spend defensive energy in the “act first” quadrant. Manage the lower-severity items with policy. Prepare for the high-severity tail risks even if they have not happened yet. Do not waste energy on the “monitor” quadrant.

Module 06 Quiz · pass mark 4 of 6 points

Q6.1 · How much recorded audio is required to produce a serviceable clone of someone's voice in 2026?
Q6.2 · Which is the strongest pre-emptive defence against style imitation for a visual artist?
Q6.3 · True or false. The “act first” quadrant for a Caribbean creative in 2026 includes wage compression, lost briefs, and style imitation.
Q6.4 · (2 points · scenario) A producer in Kingston has six possible defensive moves. Order them in priority for impact: (1) install watermark plugin, (2) update standard contract with AI clauses, (3) lobby for new copyright legislation, (4) build email list of fans, (5) apply Glaze to her portfolio, (6) write a public blog post critical of AI companies.
Q6.5 · A creative is choosing between a high-likelihood, high-severity risk and a low-likelihood, low-severity risk. The course advises which approach?