Don’t expect a dramatic AI-assisted sci-fi encounter if humanity ever definitively detects evidence of intelligent extraterrestrial life. Scanning the stars for signs of extraterrestrials is less about waiting for giant unidentified aerial phenomena (UAPs) to appear and more about scouring mountains of complex data in search of delicate biosignatures.

In recent years, many researchers (including some at NASA) have advocated incorporating machine learning and artificial intelligence in their search for organisms beyond Earth. Some of these approaches may show promise, but new research indicates that much of today’s AI is even easier to fool with false positives than its human operators.

“No matter what script we started with, we were able to fool the AI ​​100 percent of the time,” Ankit Gupta, a computer science engineer at Michigan State University (MSU), said in a statement.

Gupta and his colleague Christoph Adami recently conducted an experiment to evaluate the ability of a specially designed artificial intelligence program to identify hypothetical signs of biosignatures. To do this, they relied on a computer program developed at MSU called Avida, which simulates evolutionary processes with digital organisms. Avida treats replicating biological molecules, such as DNA, as computer code, then uses these command chains to repeatedly copy itself within a “virtual Petri dish.” Importantly, each coding iteration is imperfect or contains fundamental changes, similar to how biological organisms reproduce.

Gupta and Adami then trained a neural network on tens of thousands of digital organisms within Avida, some of which included the command to copy themselves while others did not. After tasking its AI with classifying the two types of organisms, the program achieved a near-perfect accuracy rate. However, the AI ​​quickly found its answer once researchers presented new examples it had not encountered before. In just 150 small changes to the organisms’ computer code, AI began to mistakenly identify signs of life.

“AI has an Achilles heel. It can see a pattern and completely misclassify it,” Adami explained. “It is a very serious vulnerability.”

Unlike here on Earth, it could be much more difficult to secure a second set of (human) eyes on the AI ​​work aboard the next Mars rover or planetary probe. But similar false positives from AI already affect much more than future space missions. Facial recognition software, autonomous vehicles, and medical scanners rely on various machine learning programs to make decisions. Placing too much faith in the reliability of technology goes beyond misidentifying new forms of life: it undermines existing life.

According to Adami, his findings are not a criticism of AI, but a reminder that people remain vital to any new field of scientific discovery.

“You need an independent way to check [AI’s] job,” Adami said. “There needs to be a human being in the loop.”

2025 PopSci Best of what’s new

The 50 most important innovations of the year