One day in 2021 an Indian student in her early teens came to her family with some distressing news: At school, a man in his 20s had called her into an empty classroom and raped her at knifepoint. Later, after the student found out she was pregnant and decided to abort, she filed a statement with the police.
The man accused of the crime, Surjaram, denied everything and sought to be released on bail. To prove his innocence, he asked to undergo three forensic science tests, including one called brain electrical oscillation signature profiling (BEOS).
During the test, he would likely have sat quietly in an empty room, listening to a series of short, first-person phrases recounting the crime scene—perhaps “I called the girl into the room,” “I closed the door,” and “I pulled a knife”—while a headset recorded his brain’s electrical activity. Meanwhile, a computer monitored how his brain responded to each phrase, looking for telltale signs that he had participated in the crime. In Surjaram’s case, the results were what he hoped for: His brain signals suggested he had no experience of these events, and was therefore innocent.
Based largely on BEOS and the other two tests—lie detection and narcoanalysis, which involves injecting suspects with a “truth serum”—the judge released Surjaram from police custody. The case was sent to another court, but the judgments aren’t publicly available.
That case is one of at least 700 police investigations in India since the early 2000s in which suspects accused of serious crimes such as murder, rape, and terrorism have undergone BEOS tests. Proponents of the technology claim it is near-infallible. “It’s very helpful,” says Asha Srivastava, dean of the National Forensic Sciences University (NFSU), who says she has conducted BEOS tests on thousands of suspects herself. She sees it as an objective and more humane way to obtain information than torture, a practice Indian police are notorious for. “We are on a mission to make police friendly.”
But multiple scientific and legal experts say BEOS rests on scant evidence. The method hasn’t properly been vetted by the scientific community, they say, and the Bengaluru-based company that developed the technology, Axxonet, has only published a few studies on its effectiveness, which experts say are flawed. “Definitely I would not be in favor of this being used in an actual case,” says Narayanan Srinivasan, a cognitive scientist at the Indian Institute of Technology (IIT) Kanpur.
Despite this, the use of BEOS shows no sign of stopping. Although BEOS evidence has not been admissible in Indian courts since 2010, Science has found multiple recent court judgments in which BEOS tests appear to have influenced judges’ decisions about whether suspected rapists and others should be released on bail. Police and investigative agencies continue to use the systems, which can cost upward of $100,000, to generate leads, corroborate findings, and screen suspects—sometimes years after the original crime. And although Axxonet says they’re not in the business of actively selling their technology, Science.org has learned that in the past few years, forensic scientists affiliated with NFSU have been promoting the technology abroad, giving lectures about its merits to officials in several countries in South America, Africa, and Asia. Some overseas groups have already bought BEOS-related equipment, or are negotiating its purchase.
“BEOS has been validated and proved itself to be a useful tool during complex investigations,” says Chetan Mukundan, a director of Axxonet and son of the technology’s inventor. BEOS reports are not primary evidence, he adds, but are always corroborated with other investigative tools. India’s Ministry of Home Affairs did not respond to multiple requests for comment.
“The stakes are very high,” says Owen Jones, a professor of law and biological sciences at Vanderbilt University. “Reliably discovering from brain activities whether a suspect or witness has personal knowledge of an event could transform justice, internationally, in a positive way. But admitting unreliable evidence that may carry the veneer of science would do the opposite.”
The genesis of BEOS can be traced back to the United States, where, in the mid-1980s, neuroscientist Lawrence Farwell developed a technology called brain fingerprinting, which he claims has had a 100% success rate in filtering criminals from innocents. The technology is based on a pattern of electrical activity in the brain called the P300, which shows up as a spike in an electroencephalogram (EEG) recording about 300 milliseconds after something unexpected or meaningful appears—for example a “boop” sound after a string of “beep” noises.
Farwell’s idea was to leverage this signal for counterterrorism or crime fighting by exposing suspects to images and words, some of which are related to a crime. If a suspect shows a P300 peak (particularly if it is followed by a subsequent dip in the signal) when they read or see crime details that only the perpetrator would know, that suggests they might have been involved, he says.
Multiple experts consulted by Science.org say there’s little evidence it works, but Farwell cites publications in journals including Psychophysiology, the Journal of Forensic Sciences, and Cognitive Neurodynamics. “The objective fact is that the scientific community has found my research to be worthy of publication in the top scientific journals,” he told Science.
“I would not be in favor of this being used in an actual case,” says Narayanan Srinivasan Indian Institute of Technology Kanpur
A 2001 review by the U.S. Government Accountability Office concluded that government agencies such as CIA and FBI “do not foresee using the brain fingerprinting technique,” mainly because of its “limited use.” But Farwell says the agencies did not try to evaluate the validity of his technology and that brain fingerprinting had since been sold to the government of Pakistan, where he conducted “extensive training and guidance in real-world applications.”
In India, Champadi Raman Mukundan, a well-known clinical psychologist, was intrigued by the concept and signed an agreement with the Directorate of Forensic Science (DFS) in Gandhinagar—a state government lab—to develop the technology. Rather than focusing on the P300 signal, Mukundan set out to incorporate multiple other brain signals to reveal not only whether someone knows about a crime, but also whether they have what he called “experiential knowledge” of the crime—meaning they actively participated in it.
Mukundan set up a company, Axxonet, and around 2005, the team created a BEOS prototype. The technology is based on the idea that the emotions and physical sensations associated with an experience leave imprints in the brain, says Priyanka Kacker, who researches behavioral forensics at NFSU and worked with Mukundan. Recalling an experience involves reactivating those imprints—and the brain areas responsible for encoding them. Mukundan claims the BEOS system can assess whether a person has firsthand experience of an event by tracking changes in these regions.
In practice, the first step is for forensic scientists to interview both the police and the suspect to get their versions of events. These are then broken up into two scripts consisting of short phrases, or “probes,” which are played to the suspect, who wears an EEG headset to measure electrical brain activity. For each phrase, software monitors the activity for signs that the suspect has experiential knowledge of that event.
A forensic scientist then evaluates the system’s findings, looking for sequential instances of experiential knowledge during phrases that describe key aspects of the crime. “If you get experiential knowledge on a probe, like, ‘I took a knife,’ that obviously cannot be interpreted as that person has killed him,” says Deepti Puranik, a forensic scientist at the Narsee Monjee Institute of Management Studies who says she has done 200 to 300 BEOS tests over the past 2 decades. A response to a second phrase like, “I held it to his neck,” however, would raise the chances that the person had taken part. “You have to look at the sequence.”
Axxonet’s website says BEOS was “adapted and developed” based on a flurry of EEG studies dating back to the 1980s, most of which investigate brain function in patients with conditions such as schizophrenia, alcoholism, and obsessive compulsive disorder. More recent studies allegedly supported the efficacy of BEOS. But researchers say those studies were poorly conducted and many were published in little-known journals. “I have not heard of a single one of these journals,” said Sridhar Devarajan, a neuroscientist at the Indian Institute of Science Bengaluru. “The hypotheses and the methods are, to put it mildly, frivolous and sloppy.”
The study Axxonet cites most often was conducted by DFS between 2006 and 2008, after BEOS had already been applied in hundreds of real-world cases, and hasn’t been published in a peer-reviewed journal. It included 110 participants, about half of whom were told to conduct a mock crime in which they break into a room and steal from a piggy bank. The other half were told about the facts of the crime but did not actually perform it. The researchers then performed BEOS tests on all participants.
The results suggest the technology could correctly identify people who had committed the crime about 90% of the time and misclassified about 5% of those who didn’t. The study concludes that “BEOS profiling can be used as a valid scientific test for forensic purposes.”
That evidence isn’t enough to support the real-world use of a device that could have a major impact on a person’s life, skeptics say. For a start, the sample size is far too small, Devarajan says, adding that such research should include “tens of thousands” of participants. Others point out that the results haven’t been validated elsewhere. “With anything like this, I would be doubly sure, triply sure, before I would say anything about [a suspect],” Srinivasan says.
Chetan Mukundan says further research has been conducted in the field, including one “robust independent field trial” by a non-Indian law enforcement agency that yielded excellent results. “Unfortunately for us the report is confidential.”
Others say so little is known about the algorithms, methods, and data analyses underlying BEOS that it is impossible to scrutinize it in any kind of meaningful way. “They wouldn’t reveal it to me. It was totally proprietary,” says Emily Murphy, associate professor at the University of California College of the Law, San Francisco, who met with the elder Mukundan in 2009 to study BEOS and its application in law.
Meghana Srivatsav, a psychologist at Flame University, interned at DFS in 2011 before working for a year at Axxonet. She recalls that BEOS was a work in progress, constantly being tweaked and tested, even while it was used in high-stakes, real-world cases. “That’s not how it should work,” she says. “When there’s uncertainty in the lab, and we are testing and retesting and that is put into practice, as a scientist, I see that as a bit of a problem.”
Though the workings of BEOS are proprietary, “the experts in the system who need to know about the internals of the technology have tested and validated it,” says Chetan Mukundan. The technology relies on advances in neural processing, he adds, and is constantly being updated to “accommodate new releases and security patches” and user feedback.
Yet the science underpinning BEOS doesn’t square with what we know about how the brain works, neuroscientists say. “I’m very critical of the approach and the results,” says Rafael Yuste, a neuroscientist at Columbia University and chair of the Neurorights Foundation, which works on protecting people’s brain data. Our memories are highly malleable, and it’s not easy to disentangle their neurobiological underpinnings from those of similar phenomena, such as false memories or imagination, adds Anthony Wagner, a psychologist at Stanford University and director of the Stanford Memory Lab. A neural signal detected by BEOS could in fact reflect a memory generated when a suspect was interviewed, rather than a real memory of the crime, he suggests.
If BEOS were able to reliably tease out truly experienced memories, that would constitute a major scientific development. “It’d be great to see the evidence that they’ve had this breakthrough,” Wagner says. “I would be very skeptical.”
Despite these concerns, BEOS has become an important component of the arsenal of forensic tests used by the Indian police. Initially, tests were only conducted at DFS in Gandhinagar, but at least five other state forensic laboratories in the country have either carried out BEOS tests or purchased the system. NFSU, set up in 2008 by India’s prime minister, Narendra Modi, while he was still chief minister of Gujarat, also carries out tests.
As in Surjaram’s case, BEOS is often used in concert with two other questionable techniques: lie detection—also known as polygraph—and narcoanalysis, in which a “truth serum” injected into suspects puts them into a stupor and supposedly renders them more apt to speak. “These technologies have a history which has been discredited,” says Jinee Lokaneeta, a political scientist at Drew University whose 2020 book The Truth Machines: Policing, Violence, and Scientific Interrogations in India explores the use of these technologies. “And yet they sort of remain.”
One rationale for adopting those technologies has been to modernize the Indian criminal justice system and ease the rampant use of physical torture by police, Lokaneeta says. But in some cases, they have simply led to new forms of harm. In 2006, Abdul Wahid Shaikh and 12 other suspects were accused of being involved in the Mumbai train bombings, a series of seven blasts on a suburban railway that killed more than 200 people. Shaikh and the others were incarcerated and, during interrogation, underwent multiple rounds of narcoanalysis, BEOS, and polygraph tests. Some were conducted with the permission of courts and in controlled settings, such as in a hospital. But Shaikh says he was also sometimes brought into police custody where authorities “forcefully conducted these tests.” In 2015, 9 years after he was first arrested, the courts ruled there wasn’t enough evidence to incriminate Shaikh and acquitted him.
BEOS also drew criticism after a high-profile 2008 case, in which a woman called Aditi Sharma and her lover Pravin were accused of poisoning Sharma’s former fiancé with a prasad—a traditional Indian sweet—laced with arsenic. Sharma underwent a series of tests, including BEOS, which suggested she was guilty. She and Pravin were initially sentenced to life imprisonment, only to be released on bail months later because of a lack of direct incriminating evidence. U.S. scientists quoted in a story in The New York Times condemned the use of BEOS, which they said was not credible.
Meanwhile the Indian government had assembled a small team of leading neuroscientists to independently review the research behind the technology and visit forensic science laboratories where BEOS was being used. The committee identified a number of concerns, particularly regarding “experiential knowledge”: Among other things, it found that there was no supporting peer-reviewed research validating the concept and that irrelevant probes were just as likely to elicit a telltale signal as relevant ones.
Their 2008 report concluded that the scientific basis for BEOS was “suboptimal” and recommended it not be used as evidence in court. The government disagreed. Four months after the committee finished its review, DFS disbanded it, arguing that not all members had visited the laboratories, which were “doing good work.”
“We will get to a point where there will be an unbeatable lie detector test.”
Jared Genser Neurorights Foundation
Two years later, however, India’s Supreme Court stepped in with a judgment that restricted the technology, although because of legal rather than scientific concerns. Its ruling stated that forced use of BEOS and other forensic technologies such as lie detection and narcoanalysis violate a person’s right against self-incrimination. Among other safeguards, investigators must first obtain a person’s consent and grant them access to a lawyer, the court said. The judgment also ruled that the test results themselves cannot be brought into courtrooms but that any other evidence investigators obtain through such tests can be.
The Supreme Court’s decision has had an impact. Several courts have reversed earlier decisions in cases where police didn’t get proper consent to conduct BEOS and other forensic tests, for instance. In some cases, suspects who were initially sentenced to life imprisonment were later deemed innocent.
However, Science.org has found that BEOS still shapes high-stakes court decisions, particularly on bail applications. The Supreme Court ruling might seem to ban courts from considering BEOS results in those hearings. But the rules for bail applications are generally less stringent than for other court procedures, in part because final judgments are made later during the trial. “You’re not bound by the rules of evidence very strictly when you’re looking at bail,” says Chinmay Kanojia, a criminal defense lawyer in India.
In one judgment from 2021, a man accused of raping and murdering a girl underwent a BEOS test 7 years after the crime during a court-ordered reinvestigation—a delay that compounds skepticism about the technique. “It’s very likely that there is no empirical evidence that would support claims that the BEOS method is capable of accurate memory detection at long delays,” Wagner says. The BEOS evidence suggested the man was guilty, and the court rejected his bail application.
In another case from 2018, a headmaster from a school in Ahmedabad was accused of raping one of his students, a 7-year-old girl. The judge initially leaned toward believing the girl’s story, but changed his mind when he saw that BEOS, lie detection, and narcoanalysis test results pointed toward the man’s innocence. “It is only with the aid of the results of the three scientific tests that I am in a position to reach an appropriate conclusion,” said Jamshed Burjor Pardiwala, the high court judge on the case.
Pardiwala’s decision to release the man on bail was later overturned by the Supreme Court, citing procedural violations. But he has since become a Supreme Court judge and in May 2028 is slated to become chief justice of India, the highest ranking position in the Indian judiciary.
Science.org contacted the Supreme Court of India with questions for Pardiwala but did not get a response after multiple attempts.
More recently, even the Supreme Court itself has deferred to evidence from BEOS and the other forensic technologies it had sought to regulate. In 2023, the court intervened in a case where murder suspects had been discharged, saying the order had “failed to consider” that these technologies pointed toward their guilt. It ordered the trial to go forward.
As India’s criminal justice system grapples with how to use BEOS, the government has made forensic sciences a top national priority. “In the next 10 years, India’s criminal justice system will be the most modern, scientific, and speedy in the world,” according to a November 2024 press release from the Ministry of Home Affairs. To do this, the country aims to have at least 150,000 forensic experts by 2028.
The institution spearheading this effort is NFSU, which trains students in the use of BEOS and emphasizes its value. The tool is “99.99% accurate and reliable,” says Pravesh Charan, who is doing a doctoral degree at the university. Professors at the university teach that any lapses during tests are attributable to the experimenter themselves, rather than the technology. “If that person is not trained in using the BEOS system, especially how to prepare probes, then that could be a major limitation,” Kacker says. “Otherwise, systemwise, there is no such limitation.”
NFSU did not respond to multiple requests for comment.
Meanwhile police and forensic science laboratories around India have continued to acquire the technology. In 2022, police in the state of Karnataka added BEOS to their investigative toolkit. And last year, a forensic science laboratory in Raipur, a city in central India, bought a BEOS system for the equivalent of about $118,000.
Axxonet says they “do not actively pursue the sale of BEOS-based tools.” But forensic scientists at NFSU have been promoting the technology abroad. Last year Charan conducted a training session on forensic science techniques, including BEOS, for police officials in Guyana, and says they are in the process of buying the technology.
In September 2024, Charan also conducted a training session for the Bangladesh Institute of Forensic Psychology and Sciences, a company that says it consults for various organizations, including Bangladesh’s military and intelligence and law enforcement agencies. In October 2022, Kacker showcased various forensic techniques, including BEOS, to delegates from Rwanda, according to her university’s annual report. A few months ago, she gave a lecture on these technologies to Tanzanian police. Trade data also show that a private security company in Singapore called PCS Security purchased BEOS-related equipment in 2018.
In April 2023, NFSU inaugurated its first-ever campus abroad in Jinja, Uganda, where it will likely teach students the same forensic science techniques they learn in India.
BEOS’s expanding sphere of influence “doesn’t surprise me, but it does concern me,” says Marcello Ienca, professor of ethics of artificial intelligence (AI) and neuroscience at the Technical University of Munich. To him, the quick proliferation of unvalidated brain-reading technologies underscores the importance of developing international regulations.
Ironically, those concerns may sharpen if BEOS or some other neurotechnology ultimately does live up to its billing. “I think there is every reason to believe that we will get to a point where there will be an unbeatable lie detector test,” says Jared Genser, general counsel of the Neurorights Foundation, citing advancements in AI and recent breakthroughs in neurotechnologies. That, he and others say, could pose a serious threat to privacy and autonomy. “We do need to be thinking now about the guardrails for that kind of technology.”
Reprinted with permission from Science, Vol 388, Issue 6748, 2025
The story was supported by the Science Fund for Investigative Reporting and an Investigative Reporters and Editors Freelance Investigative Fellowship Award.
Jonathan Moens is a science journalist in Rome. He tweets @jonathan_moens.