A Biotech Company’s Contradictory Claims Expose a Legal Fault Line
Insilico Medicine built computer models that proposed a promising drug candidate for pulmonary fibrosis – and then publicly declared the molecule had been “discovered by” generative AI. When patent filings followed, five human names appeared as inventors. AI was not mentioned. That gap between the press release and the legal document is not a clerical inconsistency. It is a precise illustration of where intellectual-property law currently stands, and where it is about to be tested harder than ever before.
Patent law, in every major jurisdiction, allows only humans to hold inventorship rights.
As AI systems grow capable of generating viable drug candidates with the same ease that a language model writes a routine email, the underlying legal question – what does it actually mean to invent something – is approaching an answer that current statutes were never designed to give. The pharmaceutical industry is moving faster than the law, and the distance between those two speeds is widening by the quarter.
What the Insilico Case Actually Reveals
Insilico Medicine’s situation is not an anomaly. It is the first clear public example of a structural problem that every AI-driven drug discovery company will eventually face. The company’s generative AI system did not assist a human researcher in the traditional sense – it did not run calculations that a scientist then interpreted. The AI proposed the molecule itself. That is a qualitatively different relationship between tool and output than patent law has ever had to process.
The five humans named on the patent filing presumably satisfied legal inventorship requirements by contributing to conception of the invention in ways the law recognizes. But as AI models take on more of what has historically constituted conception – identifying targets, generating chemical structures, predicting binding behavior – the human contribution at each stage becomes thinner and harder to locate precisely. A researcher who selects training data, sets parameters, and evaluates outputs may or may not meet the bar for inventorship depending on how courts eventually choose to read those actions.
The discrepancy in Insilico’s public communications versus its legal filings is not necessarily dishonest. It reflects two different audiences operating under two entirely different frameworks. Marketing language answers to no legal standard. Patent filings answer to very specific ones. What makes this moment significant is that the gap between those frameworks is now visible enough that regulators, courts, and patent offices cannot continue to ignore it.
Space Mirrors, Drone Swarms, and the Week’s Other Flashpoints
Separate from the patent question, a company called Reflect Orbital is drawing scrutiny for plans that involve deploying up to 50,000 satellites carrying large reflective surfaces in orbit. The stated applications include extending daylight for solar panel charging, emergency response, and military use. A new study found that the giant beams of reflected sunlight could shine as bright as 10,000 full moons and scatter light across tens of kilometers – raising concerns from astronomers, aviation authorities, and wildlife researchers. Reflect Orbital plans to launch a test satellite this year, extending an 18-by-18-meter mirror in orbit, before any broader deployment decisions are made.
Elsewhere, reporting from The Atlantic revealed that Ukraine planned an operation to send 1,000 AI-guided autonomous drones per night toward Moscow airports – a stalled effort that, if it had proceeded, would have represented one of the largest deployments of autonomous weapons systems targeting civilian infrastructure in the current conflict. Separately, Kyiv is seeking Elon Musk’s approval to use Starlink-equipped drones inside Russia, according to the Financial Times. The combination of AI-guided targeting and commercial satellite communications in an active war zone points to questions about accountability that are moving faster than any existing legal or military framework.
On the AI and science research front, a study published in Nature found that 90 percent of biomedical papers now show signs of AI use – a figure considerably higher than previous estimates and raising fresh concerns about the reliability of the scientific record when AI-generated text, summaries, or analysis goes undisclosed. A third of new web pages also show AI authorship, according to TechCrunch. The concentration of AI-generated content across both professional research and the open web is creating a feedback loop: AI systems trained on web data are increasingly being trained on content that other AI systems produced.
Patent Offices Are Already Behind
The U.S. Patent and Trademark Office has ruled previously that AI cannot be named as an inventor, a position consistent with how courts have interpreted the word “whoever” in patent statutes to mean a natural person. The European Patent Office reached the same conclusion. Those rulings were issued in response to deliberate test cases filed by researchers trying to force the question. The Insilico situation is different – it is not a legal test. It is ordinary commercial activity, and the mismatch it exposes will only multiply as more AI drug discovery programs move toward clinical trials and eventual commercialization.
China’s Chang’e 7 mission, separately, is expected to attempt the first landing at the moon’s south pole this year, deploying robotic landers, rovers, and what mission planners describe as “hopping” probes to hunt for lunar ice – adding another layer to a space competition that now runs from low Earth orbit to the lunar surface to, apparently, a startup’s plan to bounce sunlight off 50,000 mirrors. Meanwhile, Greater Manchester has rejected Palantir’s data platform in favor of building its own homegrown system, insisting it can manage public-sector data more effectively without the American contractor. Whether a regional government’s in-house engineering team can realistically outperform a company whose core product is exactly this – and which counts intelligence agencies among its long-term clients – is a question Greater Manchester has committed real money to answering.
