AI Throws the Patent System into Turmoil

Artificial intelligence is overturning the conventional idea of intellectual property rights, especially in drug discovery.

Artificial intelligence or AI has been shaking up the established system of intellectual property rights (IPR) like nothing else, possibly since the time the British Parliament enacted the Statute of Monopolies in 1623.

The modern patent system has taken its time to evolve, the milestone development being the first Patent Statute passed by the US Congress in 1790, which was followed a year later by France’s patent system, created during the Revolution.

However, the AI revolution has thrown into a flap both policymakers and judges who are struggling to understand the wider implications of the new technology on the modern patent system that was established more than a century ago.

The case that has grabbed attention is the suit filed in February this year by stock photography giant Getty Images in the US against London-based Stability AI, the startup that created open-source AI art generator Stable Diffusion.

The photo company accused Stability AI of “brazen infringement of Getty Images’ intellectual property on a staggering scale” by copying more than 12 million images from its database “without permission…or compensation…as part of its efforts to build a competing business”.

Getty charges Stability AI with infringing its copyright and trademark protections.

AI art tools require illustrations, artwork and photographs to use as training data, which are taken from the web, usually without the creator’s consent. The startup, which describes itself as the world’s leading open source generative AI firm, is also facing a class action suit by artists in the US.

These artists have filed lawsuits against Stability AI and two other companies, alleging that their AI tools were “trained on billions of copyrighted images scraped from the internet” and contained in a dataset that was downloaded and used by the companies without compensation for or consent from the artists.

This may sound like the familiar old battle between artists and corporations on copyrights, but the new technology gives it a complex twist that will not be easy to unravel.

It is not just the judges who will be grappling with the implications of AI on diverse fields from art to medicine, but a host of others who will be drawn into the dispute. The fact that Getty has a sought a jury trial makes the outcome of the case even more unpredictable.

Stability AI contends its “goal is to maximize the accessibility of modern AI to inspire global creativity and innovation”.

That goal may be a long way off if the courts order, as Getty has sought, the withdrawal of the startup’s Stable Diffusion—the system for generating images from text inputs—and the image generator DreamStudio that was launched in August 2022.

However, the disruptive force of AI is beginning to be felt more strikingly in another field—medicine. As the legal drama over AI art tool generators escalates, the US Congress is wrestling with a more fundamental and vexing question: can AI systems involved in inventive processes like drug discovery be awarded patents?

This is a critical issue, because the US patent law, like those of Europe’s, recognises only humans as inventors and not machine-based or AI systems even though science and technology is demonstrating otherwise.

Most dramatically, AI has been able to invent new molecules within a short span, a feat that promises to upend the pharmaceutical market, healthcare and the patent system.

Although not the first AI-assisted breakthrough, a spectacular success came just a few weeks ago, with the announcement of the discovery of a new antibiotic that can kill a deadly hospital bacteria.

A study published on May 25 in the journal Nature Chemical Biology, by scientists from Canada’s McMaster University’s department of biomedicine and biochemistry and from the Massachusetts Institute of Technology, reported the discovery of a new antibiotic that can be used to kill a superbug.

The study focused on Acinetobacter baumannii, which heads the World Health Organization’s list of “priority pathogens”, a group of bacteria that pose the greatest threat to human life.

The way the two institutes went about finding the antibiotic to fight this superbug was to first screen around 7,500 molecules against the bacteria in a dish to find which ones could inhibit its growth. This data was used to train their machine learning model to identify chemical structures with anti-Acinetobacter baumannii qualities.

After the model was designed, it was used it to analyse the structures of 6,680 compounds from an open-access library of molecules with potential therapeutic benefits. This turned up a shortlist of hundreds of candidates, which was pruned and the team was able to zero in on a winner—a new antibacterial compound named abaucin.

That does not mean the end of the story is in sight. There is a still a critical phase before the drug can be brought to the market; the three-stage clinical trials which must prove the molecule will work effectively on humans.

What AI has accomplished is a sharp reduction in costs and time on the critical first phase of drug discovery. Industry sources are elated by the results of this study because AI has shaved off two years from the normal 4.5 years it takes, and slashed costs by a third.

These are parameters which are expected to improve as researchers hone their AI tools. Lawmakers and the courts, however, are behind the curve. It is still a million-dollar question whether the patent for abaucin will be granted to the AI system that discovered it or to the human behind it.

The case of computer scientist Stephen Thaler is instructive. Thaler had sought a US patent for a beverage holder and an emergency light beacon designed by his AI system without any human input.

The US Patent and Trademark Office (USPTO) had turned down his application because the rules say only humans can be inventors. It was a decision upheld by the lower courts, and when Thaler sought the intervention of the Supreme Court, it refused to do so.

But fresh winds may be blowing through the system. In April, Lantern Pharma Inc of Dallas, a clinical-stage biopharma company that uses its proprietary AI and machine learning platform to transform the cost and pace of oncology drug discovery said USPTO had allowed it file a claim on a new molecular entity it had discovered.

LP-284, its new molecule for treating non-Hodgkin’s lymphomas, is scheduled for Phase 1 trials later this year. It could make history in the US as the first AI-developed oncology drug to hit the market.