IP in the age of AI: what to consider

The line between what a person made and what a machine generated is disappearing. Ownership is becoming the defining question.

In 2025, Anthropic agreed to pay authors $1.5 billion, the largest copyright settlement in US history, covering roughly half a million books at around $3,000 per work. By mid-2026, more than 70 AI copyright lawsuits were active or recently resolved, with claimed damages across live cases exceeding $50 billion. The message is simple. Creative work has value, AI systems consume it at scale, and the fight over who gets paid has begun.

But litigation is a poor protection strategy. It is slow, expensive, and backwards-looking. It compensates for value already taken rather than preventing the taking. And outcomes hinge on details most creators never think about. In the Anthropic settlement, the difference between a $3,000 recovery and nothing could rest on whether a work had been registered in time. In the UK, a court declined to let Getty Images bring a claim on behalf of the 50,000 photographers affected in its case against Stability AI, leaving individual creators to fend for themselves.

So what should creators and businesses consider?

First, know where your work goes. Every prompt, upload, and workflow passed to an AI system is potentially training material for someone else's model. Read the retention terms. Assume nothing.

Second, protect the process, not just the output. The prompts, datasets, and creative decisions behind a piece of work are often more valuable than the work itself. If they leak into a shared model, they are gone, and no court can retrieve them.

Third, keep provenance. Records of what you made, when, and how are the foundation of any future claim, licence, or sale.

Fourth, remember the principle that connects all of this: you cannot sell what you do not own.

This question has defined Potentially since the start. How do we protect human ingenuity? Not with terms and conditions bolted on afterwards, but with the right architecture from day one, built to prove what belongs to whom. Every Star runs in its own environment. Every workflow, dataset, and output is isolated, so a person's work stays theirs by design. And protection is only the beginning. Once you can prove what is yours, you can build vehicles to monetise and distribute it. That is exactly what we are building.

How to protect human ingenuity is one of the defining questions of the era, and we want to be a meaningful voice in that equation.

SOURCES

Bartz v Anthropic settlement, preliminary approval September 2025; Authors Guild; NPR. Axis Intelligence AI copyright litigation tracker, June 2026. Getty Images v Stability AI, UK High Court, 2025; Pinsent Masons analysis.