UNAM retest will decide 22,000 admissions places | Magica
UNAM’s retest will reallocate 22,000 places, but the remote-score anomaly remains unexplained
Editorial Team
••📖6 min read
UNAM will require about 58,000 applicants to take an in-person control exam after anomalous results in its first fully remote undergraduate admissions test. The new score will determine roughly 22,000 places, while the university has not publicly established what caused the distorted results or which applicants, if any, acted improperly.
About 58,000 candidates will be called to an in-person control exam; its result will allocate roughly 22,000 undergraduate places.
UNAM says the remote-score distribution changed sharply, but its review cannot identify individual wrongdoing or how any score was obtained.
The remedy restores a contest for scarce places while leaving the failure of the remote process—and the burden on honest applicants—unresolved.
Mexico’s National Autonomous University of Mexico, or UNAM, will use a fresh in-person exam to decide roughly 22,000 undergraduate places after its first fully remote admissions test produced an anomalous score distribution. The new examen de control will cover about 58,000 applicants: people initially selected and others whose remote score met the lowest admission threshold for the same programme, campus and mode in at least one cycle from 2021 through 2026.
That means the measure is not a simple check on offers already made. The control-exam score will determine admission for a pool that is more than twice the number of available places. UNAM’s technical commission designed the broader pool to give another chance to people who may have been displaced by the irregularities, while its own recommendation says the invitation is not an accusation against any individual candidate.
Share of test-takers scoring at least 100 or 110 correct answers before and during UNAM’s first fully remote undergraduate admissions exam. Source: Ars Technica.
A distorted ranking, not an individual verdict
More than 158,000 applicants took the 120-question exam over three weekends in May and June. It was UNAM’s first fully online, at-home undergraduate entrance test. In the preceding 2021–25 cycles, 3.5% of test-takers scored at least 100; in 2026 that share was 16.3%. The share scoring at least 110 rose from 0.9% to 5.5%, according to the account of the commission’s findings.
The numbers establish an abnormal result distribution, not a list of people who cheated. Eduardo Bárzana García, president of the 16-member technical commission created to review the admission process, said the statistical work identified substantial changes in the general distribution of correct answers but could not by itself determine individual irregular conduct or how each result was obtained, as records.
SoftBank has drawn $20 billion of a $40 billion bridge facility for its OpenAI investment and expects another $10 billion draw in October. Its Aug. 6 earnings release should show whether the group can turn a stated menu of asset-backed financing, bonds and asset sales into longer-term funding.
The distinction changes what a proportionate remedy looks like. Roughly 36% of the more than 158,000 original candidates are being called back, rather than only those who received offers. Yet fewer than half of that group can be admitted. The original test identifies who can compete again; it no longer fixes the final ranking.
Remote safeguards identified conduct, but did not settle the cause
UNAM adopted the remote format after decades of in-person admissions testing, presenting it as a way to make participation easier for people outside Mexico City and abroad, as contemporary reporting described. It hired Mexican online-education company Territorium Life because of its experience running online evaluations for other national public universities and its technical capacity to handle many tests at once, Bárzana said in the presentation report. A separate report put the contract at 36 million pesos.
The process combined identity validation, software that restricted the test computer, webcam and microphone monitoring with artificial-intelligence alerts, and one human supervisor for every 150 candidates. It detected prohibited conduct including phones, outside assistance and identity substitution; nearly 2% of the exams were cancelled during the process. Those safeguards show that some violations were detected. They do not establish whether the anomalous scores came from an evasion of monitoring, answer leaks, external assistance, prohibited materials, some combination, or another cause.
UNAM had already said it filed a criminal complaint on July 3 over conduct that could constitute illegal practices by people and companies selling fraudulent services connected with taking the exam, in its statement. That is a request for investigation, not a public finding about the remote scores or the company that ran the platform.
The institution has therefore rejected the original ranking without publicly assigning a technical fault. That is more precise than saying artificial intelligence caused the episode: AI-assisted monitoring was part of the safeguards, but the retained record does not establish that an AI tool failed, that candidates used generative AI, or that Territorium Life was responsible.
Leonardo Lomelí Vanegas, UNAM’s rector for the 2023–27 term, was appointed by the university’s governing board after a community consultation and a review of candidates’ records and plans. He previously served as UNAM’s secretary general from 2015, according to the appointment record. In his current role, he ordered the review and accepted the commission’s recommendation—authority that makes the control exam the operative admissions decision rather than an advisory proposal.
UNAM says the option was contemplated in the 2026–27 admission call. In its announcement, Lomelí apologized to applicants who had been admitted honestly but now must sit the control exam, arguing it was necessary for certainty and equitable access.
For those candidates, the second test transfers some of the cost of an institutional failure back to people who may have done nothing wrong. A group of initially selected applicants protested, asking that the remote results be respected. Marco Antonio Sánchez, a 19-year-old admitted to mechatronics engineering with 113 correct answers, said he would take the retest but wanted university leaders to accept their own share of responsibility, in a report from the demonstration.
The logistical reversal also cuts against the remote format’s inclusion rationale. UNAM plans multiple versions of the 120-question exam and in-person sites beyond Mexico City so the group can test simultaneously. It delayed the start of classes for new entrants to August 31; returning students were due back August 17, according to the updated schedule.
Applicants called to take UNAM’s control exam and the places they can obtain. Source: EL PAÍS.
The next decision is about the next admissions cycle
The immediate test of the remedy is operational: whether UNAM can stage a new, comparable exam for 58,000 people quickly enough to fill places and start the delayed term. Its design—different versions, in-person supervision and a selection pool based on historic cutoffs—can produce a new ranking. It cannot, on its own, explain the original one.
For the next cycle, the unresolved evidence is more consequential than the rhetoric around a single technology. UNAM’s commission says it will continue developing recommendations. The university will need to establish whether the weakness lay in system design, deployment, monitoring, leaked material, outside services, or a mix of factors; explain what testing and oversight a remote process received; and show how any revised format preserves both integrity and the access gains that prompted the move online.
A newly published review of Phoenix Ikner’s yearlong ChatGPT record describes crisis-support replies alongside detailed answers in the hours before the 2025 Florida State University shooting. The record raises hard questions about OpenAI’s safeguards and account-level risk detection, but it does not establish causation or reveal what the company’s internal systems knew.
Patrick Steven Yaroch, a former FBI supervisory special agent who worked counterintelligence, is accused of using bureau-held wallet information to transfer cryptocurrency into accounts he controlled. The charging records make ChatGPT conversations part of the government’s evidence, but not the alleged means of moving the money.
Caterpillar raised its 2026 sales outlook after quarterly revenue first topped $20 billion. A surge in power-generation demand points to the data-center buildout, but stronger construction sales, pricing and a growing dealer inventory make the company’s record backlog an incomplete measure of that boom.
Cloudflare is taking reservations for cloudflare.pay handles and has outlined stablecoin wallets that would let account owners constrain AI-agent spending. The wallet is not yet available, and Mastercard’s broader machine-payments program shows that identity, limits and settlement are already a competitive infrastructure race.
Apple is seeking expedited discovery before a ruling on its preliminary-injunction request against OpenAI and former Apple hardware employees. The filing adds allegations involving other former employees, while OpenAI calls the case factually wrong; the court must decide whether Apple’s proposed evidence-gathering is justified and proportionate.
HappyRobot's $150 million Series C values the enterprise AI-agent company at $1.2 billion and funds expansion beyond logistics. The company says its deployment model can automate complex operational work, but the disclosures leave its reliability, pricing and cross-sector repeatability unmeasured.
Bending Spoons has agreed to acquire Airtable in an all-cash transaction with a $1.285 billion enterprise value. Against Airtable’s company-reported $480 million in annual recurring revenue, that is about 2.7 times ARR—a more useful, though still limited, view of the transaction than a simple comparison with the company’s 2021 funding valuation.
Wisconsin Assembly member and gubernatorial candidate Francesca Hong proposes a statewide pause on new AI data-center construction while lawmakers write rules on utilities, environmental protections, subsidies, labor and local approval. The proposal puts a clear choice before the state: set those terms before more projects advance, or keep local dealmaking open while regulations are debated.
GPT-Live lets ChatGPT listen and speak at once, and OpenAI says it can bring in GPT-5.5 for harder work without stopping the conversation. The launch’s real test is whether that experience holds up across long sessions, regional capacity and uneven networks—and whether its related WebRTC proposals become broadly deployed infrastructure rather than a company-specific advantage.
Five states enacted AI mental-health restrictions in 2026, but the laws regulate different conduct: Tennessee targets representations that a system is a qualified professional, while Colorado and Maine restrict how licensed clinicians can use AI in therapy.
Xero's survey of 1,081 New Zealand SME customers finds 61% are proactively using AI, but most users are learning by experimentation. Its new 12-week bootcamp with ASB is a response to that confidence gap; it is not yet evidence of a productivity gain.
Apple alleges that OpenAI and former Apple employees sought confidential hardware information through recruiting, supplier contacts and a former engineer’s post-employment system access. The complaint has not been tested in court, and OpenAI says it is unaware of evidence that it has merit.
Amazon’s market value passed $3 trillion after AWS revenue accelerated and the company raised its 2026 capital-spending plan to $220 billion. The rally reflects confidence in Amazon’s cloud demand, but the company’s own capacity forecast, the broad uses of its budget and strong rival cloud growth leave the return on that spending unproven.
The Trump administration says it has completed a voluntary framework for testing the cyber capabilities of the most advanced U.S. AI models and will discuss it with major developers Tuesday. The key scope, testing and disclosure rules have not been made public.
MediaTek now says its second custom cloud AI accelerator is on track for high-volume production in 2028, later than its April goal of entering production by the end of 2027. The company is betting that supply-chain coordination and advanced packaging can support a much larger cloud-chip business, but it has not disclosed a customer, package supplier, cost comparison or 2028 revenue guide.
Firecrawl has merged a pdf-inspector dependency update to lopdf 0.42.0, the release identified by RustSec as fixing unbounded recursion on deeply nested PDF objects. The code change protects new builds of this repository from that specific failure mode, but the retained evidence does not establish a downstream release or deployment rollout.
OpenAI has published emails and messages challenging parts of Apple’s trade-secrets case. The material confirms a misdirected lawyer email, but the redacted messages also document efforts to transfer and retrieve Apple work around a former employee’s departure, leaving the core allegation unresolved.
SK hynix and Sandisk say they have published the first Open Compute Project specification for High Bandwidth Flash, a NAND-based layer intended to sit alongside HBM and SSDs in AI inference systems. The document fixes a technical starting point, but not the performance, cost or hardware support that would make HBF a deployed memory tier.
Chinese state media says an AI-enabled system helps air-force planners prioritize targets, coordinate attack waves and allocate tasks in large exercises. The developer-reported test results suggest a push to speed military planning, but they do not establish combat deployment or independently verified performance.
