Evaluating Authorship and Likeness

To protect your work legally, you must understand where the law draws the line on ownership. According to US Copyright Office standards, copyright only protects works created by humans. This means that purely AI-generated material—content where a machine makes the final creative decisions—is not protectable.

Consider a Creative Director using Midjourney to build a moodboard. They enter a detailed prompt describing a "close-up of a worn vintage jacket with a frayed cuff and a stove scorch mark." The AI processes this request and generates a high-fidelity image. While the director provided the idea, the AI decided the exact placement of the threads, the specific shade of the scorch, and the lighting. Because the machine performed the actual 'authorship,' the US Copyright Office considers this image to lack human authorship.

As a leader, you must recognize that any asset produced solely through prompting is a proposal, not a piece of owned intellectual property. You cannot simply assume that because you paid for the tool or wrote the prompt, your brand owns the resulting image. You use this knowledge to identify which assets in your composite are legally precarious.

While purely generated images aren't protectable, you can earn copyright protection by adding human creative value. The US Copyright Office looks for a "perceptible contribution"—specifically your creative selection, arrangement, and modification of the material. When you move from prompting a tool to actively shaping the output, you transition from using a proposal to creating an authored work.

Imagine a designer working on a brand poster for a repair service. Instead of asking an AI to "make a poster," the designer generates three separate AI textures: one of frayed denim, one of leather grain, and one of weathered canvas. The designer then manually composites these textures into a specific layout, layering them to create a collage of fabric. They add custom typography and adjust the color grading to match the brand's palette.

In this scenario, the individual textures remain unprotectable, but the final poster is an authored work. The creative selection of which textures to use, the specific arrangement of the layout, and the modification of the colors are all human decisions. As a lead, you use this distinction to guide your team: don't just generate; composite, arrange, and modify to ensure the brand owns the final delivery.

Many creative leads fall into the 'prompt fallacy': the belief that a highly detailed, technical prompt is equivalent to creative control. From a legal standpoint, it is not. The US Copyright Office maintains that writing a prompt is more like giving instructions to a commissioned artist than actually creating the work yourself. If the AI is the one deciding the final execution, the human is a client, not an author.

For example, imagine you prompt a tool for a "cinematic 35mm shot of a frayed cuff, shallow depth of field, soft morning light, Kodak Portra 400 grain." You have specified the equipment, the lighting, and the film stock. However, the AI still determines exactly where every grain of silver halide falls and how the light hits each individual thread of the fabric. These specific aesthetic choices are what constitute 'authorship.' Because you didn't place those pixels yourself, the prompt alone doesn't grant you ownership.

As a Creative Director, you use this understanding to stop your team from treating a 'perfect prompt' as a legal safeguard. You must distinguish between a prompt that describes a result and a process that actively modifies one.

Because AI-generated content is legally precarious, you must identify 'Must-Own' assets. These are critical brand elements—such as the central hero image of a campaign—that the company must legally own to ensure they cannot be contested or used by others. If an asset is the foundation of the brand's visual identity, relying on a prompt is a business risk. You ensure ownership by mandating human authorship for these specific pieces.

Take the campaign for the repair service. A team might be tempted to generate a 'perfect' image of a loyal customer wearing a weathered jacket. However, a purely generated person is a proposal the brand doesn't own. To secure the rights, the Creative Director instead hires a professional photographer to shoot real owners with their actual worn jackets. Because a human photographer made the creative decisions regarding framing and lighting, and a human model signed a release, the brand owns the copyright to those photographs.

As a lead, you apply this by auditing your asset list. You categorize every item as either 'illustrative' (where AI proposals are acceptable) or 'Must-Own' (where human craft is non-negotiable).

When your creative direction involves synthetic spokespeople—people created or modified by AI—you are no longer dealing with just copyright, but with 'likeness rights.' A likeness is the right of an individual to control how their face, voice, and identity are used commercially. Whether you use a deepfake of a known celebrity or a synthetic hybrid that looks like a specific person, you must treat the likeness as a legal requirement that exists independently of the AI tool used to create the image.

Consider the Cadbury campaign in India, where AI was used to create a synthetic version of actor Shah Rukh Khan to promote small businesses at a neighborhood scale. Even though the images were generated by AI and geo-targeted to different zip codes, the project relied on a specific likeness. This required an explicit, written permission document from the actor.

As a Creative Director, you ensure this permission isn't a one-time conversation. You treat the likeness agreement as a mandatory asset record. You mandate that a copy of the signed permission be attached to every version of the asset, from the master film to the smallest social cutdown, so that no piece of content is produced without a verified right to use that person's identity.

A permission is only useful if it is visible at the moment of release. In a high-speed production environment, rights frequently 'detach' from the asset. An image might be cleared in the concept phase, but as it is cropped for social media or cut down for different markets, the legal documentation stays in a folder somewhere while the file moves forward. This creates a production lapse where an uncleared asset is treated as ready simply because it has a final filename.

Imagine a delivery lead performing a final check on a synthetic spokesperson video destined for a neighborhood-scale rollout. They open the asset record for the final render and see the video file and the technical specs, but the field for 'Likeness Permission' is empty. Even if the team believes the rights are cleared, the absence of the document in the record means the asset is flagged as uncleared.

As a Creative Director, you implement a hard rule: the permission must travel with the asset. You treat the lack of a linked permission document as a production error, equivalent to a corrupted file. You instruct your team that no asset is 'delivery-ready' until the likeness right is physically attached to that specific version's record.