When people hear the term “STEM Olympiad,” they instinctively picture brilliant, dedicated students solving some of the hardest math and physics problems in the world. And sure, that used to be a reasonably accurate picture. Olympiads were once considered nearly a one-way ticket to MIT.
There are many types of Olympiads, ranging from math to biology. They almost always follow the acronym USA (subject) Olympiad, which is then abbreviated into USA_O. For example, the USA Mathematical Olympiad is abbreviated into USAMO, the physics version into USAPhO, the coding version into USACO, etc. The most prestigious olympiads are as follows: USAMO, USAPhO, USACO, USABO (biology), USNCO (chemistry).
So, when did all the cheating begin?
Small-scale cheating has always existed. A teacher might assist students during an exam, or a dishonest testing center might give students access to answers. That problem is probably unavoidable, as there will always be malicious or negligent proctors.
But when did large-scale Olympiad cheating begin?
Let's examine the 2023-2024 AMC (mathematics olympiad) cycle. November 2023: The AMC 10/12A leaked days before test day. The test was literally everywhere you looked. One search on the AoPS forums, Discord servers, Chinese social media and you would see it. Thousands of students had access to the questions before they ever sat down to take the exam, and you could already guess what happened to the AIME cutoffs. The MAA found the source, banned the offending test center, and promised AI-assisted anomaly detection going forward. Spoiler: this was probably the lightest cheating that would happen in the subsequent years. Actually, if we look at some of the data, the cutoff inflation wasn't that bad. Before 2024, the USAMO qualification index generally hovered around the low-to-mid 220s, although it varied by year and test combination. In the 2023–24 cycle, some USAMO cutoffs rose to approximately 245–248. That is a substantial increase, roughly equivalent to another one or two AIME problems, but it pales in comparison with what followed.
We'll get back to more MAA cheating scandals later. Let's examine something that has drastically changed since AI became a thing: USACO. Now, it's pretty self-explanatory that an olympiad which is based off coding in C++, Java, Python, etc would be heavily affected by AI usage. So it's obvious that USACO might be further gone. Its contests are unproctored. Students can take them from their bedroom, phone in hand, browser open, with absolutely nobody watching. The University of Waterloo didn't even bother releasing results for its 2025 Canadian Computing Competition (CCC) as the organizers wrote that "it was clear that many students had submitted code they did not write themselves, relying instead on forbidden external help". USACO's response to the substantial amount of cheating was to gut its own Platinum division, demoting almost everyone back to Gold and sparing only the handful who'd proven themselves in person. This was deeply unfortunate for legitimate students who had worked for years to earn their rank. One of my childhood friends, for example, is an extraordinarily talented programmer who reached USACO Platinum during his sophomore year of high school in 2023, before powerful coding-focused LLMs became widely accessible. His accomplishment is now overshadowed by the growing number of students suspected of using outside assistance to reach the same division.
Modern LLMs are now capable of solving at least some USACO Platinum-level problems in a single attempt. As a result, merely reaching Platinum carries far less credibility than it once did. The clearest remaining marker of elite ability is selection for the USA IOI training camp, where students must demonstrate their skills in a much more controlled and interpersonal environment. It is far harder to fake expertise there, and even someone who cheated their way in would likely be exposed as soon as instructors began asking detailed technical questions.
Unfortunately, similar concerns have now reached the Physics Olympiad (USAPhO). I had qualified for USAPhO myself in 2025, so I already understood roughly what level of physics ability the competition required. That year, one student in my grade reportedly scored a 21 out of 25 on the F=ma exam, placing her around the top 30 students in the country. This was surprising because, from what I had observed in AP Physics C, she was not among the strongest physics students in the class. Even more strangely, her score was higher than those of two genuinely exceptional physics students - one who later won a gold medal at the International Physics Olympiad and another who represented the United States and won gold at the European Physics Olympiad. Yet despite apparently outperforming both of them on F=ma, she did not receive even the lowest level of recognition on the subsequent USAPhO exam.
This distinction matters because AAPT had previously stated on its website that it would introduce an additional measure to make the competition fair. That statement was later removed, but the measure that followed was the expanded “Honors and Distinction List.” In addition to the traditional medals and honorable mentions, AAPT published a category called “Distinction for Commendable Performance,” recognizing hundreds of additional students who demonstrated credible performance on the written, proctored USAPhO exam. In 2025, 394 students appeared somewhere on that list. Despite supposedly placing around the top 30 nationally on F=ma, this student was not among them. She did not receive a medal, an honorable mention, or even distinction for commendable performance. This was one of the first signs I personally encountered that F=ma scores were no longer reliably measuring genuine physics ability.
During the 2026 AAPT USAPhO cycle, two students who attended the 2026 USAPhO training camp told me that AAPT officials were aware of attempted cheating during the camp itself. According to them, organizers internally distinguished between “successful campers” and other non-senior campers, with students outside the successful group reportedly having attempted to cheat during camp assessments. They estimated that the issue involved roughly half of the camp.
The same campers also believed that cheating had severely inflated the 2026 USAPhO qualification cutoff. That year, the cutoff was 16, while I scored a 14. One camper told me that, based on historical standards and the difficulty of the exam, my score ordinarily should have been enough to qualify. That does not prove exactly where the cutoff would have fallen without cheating, but it illustrates the damage caused by cutoff inflation: legitimate students can miss qualification entirely because other contestants obtained scores they did not honestly earn. The most disturbing part is that the suspected cheating apparently did not end after qualification. If students who relied on improper assistance were invited to camp and then attempted to cheat again under more controlled conditions, the problem was not a handful of isolated rule violations. It had compromised the selection process itself. At that point, even legitimate qualifiers suffer, because outsiders can no longer tell which accomplishments reflect genuine ability and which ones do not. Nowdays, a medal on USAPhO is almost meaningless unless they can demonstrate real physics capability in front of you.
Now, let us return to the MAA. If you thought 2023 was bad, you'll be in for quite a surprise.
A months-long investigation by The Lakeside Tatler uncovered Discord servers where sellers openly advertised advance access to AMC exams and answer keys. Some sold only the answers, while others provided complete solutions written by former high-level competitors or generated with AI. One AMC exam reportedly cost between $200 and $400, while materials for later competitions could cost hundreds or thousands of dollars. The investigation reviewed more than 300 messages, identified at least 30 cheating servers, and recorded over 3,200 unique users across 27 of them. This was not a handful of students privately sharing answers. It was a marketplace, complete with advertisements, menus, preorders, reviews, competing sellers, and customer-support channels. Sellers reportedly obtained exams by bribing or working with competition officials, registering fake organizations to receive materials as supposed proctors, or obtaining photographs from earlier testing locations. Once an exam was acquired, experienced competitors and AI tools could create an answer key almost immediately. Investigators independently verified a file containing the 2025 AMC 12A problems and solutions that had appeared a full day before the exam. So it was really not a surprise when the 2026 AIME cutoffs dropped.
The 2025 AMC 12A cutoff for AIME qualification rose to 96 out of 150, nearly 20 points higher than the previous year and reportedly the largest single-year increase in the competition’s history. More than 250 students received perfect scores. Previously, the record was reportedly only 46. There were so many perfect scores that the Distinguished Honor Roll cutoff, intended to recognize approximately the top one percent, became a literal 150 out of 150. I am not accusing every full scorer of cheating. In fact, MAA gave every AMC 12A participant a 6-point score boost due to a mistake in problem 25. This resulted in scores of 144 and 145.5 to be boosted to 150/150.
Actually it's quite funny, because same girl I mentioned earlier reportedly scored above 140 on the 2025 AMC 12A, yet skipped the AIME because there were supposedly “too many cheaters.” That excuse was difficult to believe. I mean seriously, who the f*** would skip the AIME after scoring that high on the AMC. You are literally on track to qualify for the USAMO after that AMC score. Anyway, back to MAA.
However, the cheating didn't stop at the AMC this time. The AIME is supposed to be more secure. It consists of 15 difficult problems whose answers are integers from 000 to 999. The MAA stated that the 2026 AIME materials were released to competition managers only 30 minutes before the testing window opened. Despite this, investigators verified that the 2026 AIME II appeared online nearly five hours before the West Coast testing window began. Because every answer is only a three-digit number, a buyer would not even need to understand the solutions. They could simply memorize or copy the answer sequence and receive an extremely high score. The MAA eventually confirmed that the reports were not merely internet rumors. Its investigations identified rule violations connected to cheating reports, including activity on Discord. The organization issued disqualifications and, in some cases, lifetime bans. However, it did not disclose how many students were punished or how they were identified. For the 2025–26 cycle, the MAA also doubled the weight of the AIME in the Olympiad qualification index. The formula changed from the AMC score plus ten times the AIME score to the AMC score plus twenty times the AIME score. In theory, giving more weight to the harder exam should reduce the importance of one unusually high AMC result. In practice, that solution becomes meaningless if the AIME itself is also leaked.
The MAA then made its most dramatic structural change. The 2026 USAMO and USAJMO were administered at 31 designated regional sites instead of hundreds of ordinary schools and testing centers. These included universities such as MIT, Stanford, and Caltech. Some locations checked photo identification, used metal detectors, prohibited electronics from bathrooms, and maintained unusually high proctor-to-student ratios. Those measures may sound extreme for a high-school mathematics competition, but organizers had good reason to be concerned. Even then, people allegedly attempted to leak the Olympiad itself. Investigators verified that USAMO and USAJMO problems appeared on Discord around 20 minutes after testing began. Organizers also received reports of students hiding phones in bathrooms, concealing smartwatches, and carrying copied problems out of testing rooms to photograph them.
The MAA deserves some credit for finally taking significant action. Centralized sites, stronger verification, disqualifications, and lifetime bans are all steps in the right direction. But these measures arrived only after the credibility of the early rounds had already been severely damaged. The USAMO itself remains far more secure, since it is administered under stricter supervision and requires students to write complete proofs. At this point, a high USAMO score or medal is one of the only reliable ways to verify genuine mathematical ability within the Olympiad system. By the end of the 2025–26 cycle, an extraordinary AMC or AIME score no longer proved what it once had. It could represent years of dedication and genuine mathematical brilliance, or a few hundred dollars and access to the correct Discord server. That, my friends, is exactly how prestige dies.
To the discouraged reader: Olympiads were never meant to be trophies for the dishonest. They were meant to forge people into stronger alloys. They teach you to struggle without quitting. To face a problem that seems impossible and attack it again, and again, and again - until it finally breaks. They build discipline, intuition, and courage. Cheaters can steal a qualification. They can steal a medal. They can even destroy the prestige of the competition itself. But... the one thing they cannot steal is the mindset that you built. Years from now, the cutoffs will be forgotten. The rankings will disappear. The medals will gather dust. What will remain is the person you became while earning them honestly.