“We’ve hooked up some machinery,” said Tommy, “that amounts to a mechanical translator… When you’re ready to talk to the skipper of the other ship, sir, I think we’re ready.”

In First Contact, Murray Leinster described science fiction’s very first universal translator. Eighty years later, the artificial intelligence (AI) revolution promises to realize his vision—albeit for human, not alien, languages.

Whilst the commercial translation sector will naturally bristle at AI, military planners facing a waning supply of linguists may welcome it. Recent figures are hard to come by, but as of 2016, fewer than 500 individuals across the United Kingdom’s three military branches combined were classified at Level 3 or higher, denoting professional working proficiency. Across the pond, guidelines mandate at least 80 percent of US Special Operations Forces possess a minimum level of foreign language proficiency, yet only “three of the eight active-duty Army SOF formations” fulfilled this target as of 2023.

Two Militaries, One Slow-Burn Crisis

The severity of this vacuum betrays poor decision-making by both countries’ governments and militaries, over decades. For the United Kingdom, the first miscalculation entailed Prime Minister Tony Blair making language GCSEs (national exams taken by 16-year-olds in England, Wales, and Northern Ireland) voluntary from 2004. Within 20 years, modern foreign languages amounted to less than three percent of all A-Levels (comparable to Advanced Placements). Military linguist courses do not require language qualifications for entry, but prior exposure is necessary if students are to seriously contemplate a related career. By 2023, most pupils surveyed did not consider languages as useful for future employment, and military reports have warned about “social stigma” surrounding languages hampering recruitment. With almost 40 percent of British Army recruits in 2013 reading at an 11-year-old’s level, one further ponders whether candidates even possess sufficient literacy skills to master a foreign tongue. Upon exiting office in 2010, Labour left behind a “record-breaking” deficit and an underfinanced Ministry of Defence; the subsequent Strategic Defence and Security Review under the Conservative-Liberal Democrat coalition only compounded the crisis by curtailing spending for military linguist programs.

Regrettably, the British defense establishment exhibits a similar pattern of treating military linguists as an afterthought. The United Kingdom’s original linguist training program was hastily devised during World War II when the military “suddenly realised” that Japanese and German speakers were perilously absent. Calls to action by successive parliaments in the last two decades suggest little has changed. In 2007, a House of Commons report highlighted “severe shortfalls” in military linguists undermining operational capability; in 2016, a House of Lords debate slammed the understaffing as “putting the UK at a competitive disadvantage”; and in 2025, Baroness Coussins, Honorary Fellow of the Chartered Institute of Linguists and former member of the Lords Defence Select Committee, voiced concerns about the Strategic Defence Review potentially outsourcing language training to cut costs, thereby diminishing the quality and variety of its offering.

America’s conundrum draws noteworthy parallels. High school language requirements only exist in ten states, a policy which “may reflect Americans’ perceptions of what skills are necessary for workers today,” with merely 35 percent regarding languages as “extremely or very important” for professional outcomes. Obligating Advanced Placements in world languages admittedly exceeds practicability in a country of 349 million and near-plenary state control over curriculums; accordingly, tertiary language education assumes increased importance. Yet, the University of North Carolina at Chapel Hill terminated all of its six area-studies centers in 2026, the very programs credited with producing “invaluable” language experts during World War II. Beyond academia, a summer 2025 US Marine Corps International Affairs Program statement announced potential reforms to the Defense Language Institute (DLI), including teaching being dispersed outside Monterey and “significant” budget cuts. Defense Secretary Robert McNamara inaugurated the DLI in 1963 to consolidate and advance language teaching across all four services; these changes will undermine an institution refined over years to best serve the military’s specific needs.

The domestic linguist crisis renders recruitment of native speakers a tempting policy option, yet this talent pool remains under-exploited with good reason. The War on Terror proved that physical fitness demands alongside danger of death or persecution at the hands of terrorists made warzones intolerable for most civilian linguists. For others, their travel history and family links present security risks which vetting often cannot refute, as Ukrainians applying to translate American military manuals have discovered. Native linguists may also lack the total loyalty, unfettered by personal beliefs, that military service instills—thus jeopardizing missions. In the Vietnam War, locals altered translations to avoid upsetting their American employers; in more recent memory, American officers were cautioned about Iraqi interpreters possibly harboring a detrimental “agenda.” As a retired US Army major bluntly put it: “But what if your translator is also out to get you?” Beyond security-minded preoccupations, using indigenous personnel incurs both a strategic and ethical cost: The military misses an opportunity to develop in-house expertise and educated locals are removed from the labor market where they could better complement military efforts by contributing to the country’s long-term development. Given these limitations, the aforementioned actions limiting homegrown linguist numbers prove baffling.

Linguists in 21st-Century Defense

Writing in 1941, American linguist Mario A. Pie asserted, “The story of the German parachutists, who came down in Holland equipped not only with Dutch uniforms, but also with a command of the Dutch language, teaches one such a lesson which we cannot afford to disregard. […] Military conquest in the present war has often been preceded, accompanied, and aided by linguistic mastery of the tongue of the conquered by the conquerors.” Despite being overshadowed by their civilian counterparts in intelligence organizations, military linguists continue playing a vital role in national security.

In 21st-century warfare, military personnel with appropriate language skills teach allies about operations, tactics, and strategy, as the United States has demonstrated in Ukraine. Without them, exchanges become mired by communication bottlenecks. Linguists in Anglophone militaries also remain a must-have despite English’s status as a military lingua franca. For instance, most South Vietnamese Air Force members sent for instruction in the United States possessed such rudimentary English skills that they were incapable of passing on training to fellow countrymen as intended.

Furthermore, 20 years of counterinsurgency in the Middle East and Central Asia have spotlighted military linguists’ contributions in irregular warfare: being first to hear life-or-death intelligence; monitoring enemy communications to gauge positions, maneuvers and intentions; advancing ahead of standard combat units to act as “barometers” of local sentiment; and leading stability or peace-building operations. 

The AI Hype Fuels a Growing Bandwagon

Tracing earlier machine translation tools’ functionality helps explain current enthusiasm surrounding AI as the answer to militaries’ linguist woes. The IBM701 broke headlines in 1952 as the first incarnation of Rule-Based Machine Translation (RBMT); relying on manually created grammatical rulesets and a word-bank, it nonetheless struggled with idioms and cultural nuances. Emerging 30 years later, Statistical Machine Translation (SMT) instead used probability to determine the most optimal translation based on a vast database of previously rendered segments; apt for scientific and technical writing but wanting for informal or literary styles. In 2016, Google Translate shifted to Neural Machine Translation (NMT), a subcategory of AI, in doing so heralding the AI translation paradigm. By assessing entire sentences rather than individual vocabulary as well as learning from human corrections, NMT translations were the most fluent and faithful to the source text yet. Nowadays, talk of generative AI translation generally refers to large language models (LLMs), another AI subcategory. Although NMT and LLMs each possess unique strengths, the latter stand out for maintaining a natural voice even across long texts.

Many have been keen to reap the benefits, with the European Union “doubling down” on AI for press releases without human supervision. Slowly but surely, the military domain has dipped its toe into the AI language pool. NATO implemented limited AI translation for English and French in September 2023, with other languages in the pipelines. British Chancellor Rachel Reeves’ 2025 Spring Statement aligned with these trends, promising that “at least ten percent” of defense cashflow will be allocated to technologies including “AI capabilities” from 2025-26. With the British military cash-strapped to the tune of £28 billion, the present AI push is framed as part of improving the United Kingdom’s “debilitating” tooth-to-tail ratio, thus achieving more effective defense outcomes overall.

The US Department of Defense (DoD) has similarly hailed AI “reach[ing] all desktops in the Pentagon and in American military installations around the world,” evoking the technological faddism of the post-Cold-War era. Already under the Biden administration, the Army Small Business Innovation Research Program offered firms a contract to engineer AI-based translation software for obscure Indo-Pacific languages. Considering these developments, we can expect military linguists on both sides of the Atlantic to undergo their own AI makeover sooner rather than later.

Language Is Not a Software Problem

Contrary to these elevated perceptions, AI falls short of its reputation as a silver bullet. Sufficient for low-stakes work such as synthesizing documents and translating their structural elements, that’s where the good news ends.

First amongst the myriad flaws, LLMs perpetuate their training corpus’ biases. Whilst presuming a “nurse” is female represents a minor discrepancy in a hospital, defaulting to a male translation of “combatant” risks opening an entirely wrong line of inquiry. Worse still, if AI encounters any unknowns, generating nonsensical or factually incorrect outputs (described as hallucinations) is inevitable, “no matter which one these (or yet undiscovered) [mitigation] techniques one employs.”

Second, only extensive datasets build effective models, a privilege which low resource languages do not enjoy. Take Swahili—designated as a “strategic language” by the DoD—which, despite being spoken by 200 million people and having the most online content of all African languages, lacks sufficient training data. Certain languages wholly lack military terminology, such as Uzbek, rendering AI useless without information from which to build an algorithm.

Third, accepting that no dataset can ever be complete or bias-free, the European Union has opted for humans to post-edit AI translations. But, with machine translation necessitating as many as three corrections per sentence, the delaying effect on turnaround times makes it highly inadequate for military campaigns’ blinding pace.

Fourth, the above problems arise prior to the added complexity introduced by auditory rather than written input. In one study, automatic speech recognition systems for non-American accents fared “considerably worse” than General American accents, with absolute performance differences ranging between 2 and 12 percent while relative ones varied from 16 and 49 percent. Notably, the accuracy demonstrated by laboratory tests is not replicated in real-world conditions. For tonal languages, specifically Mandarin Chinese, recognition rates dropped from around 95 percent for “isolated tone[s]” to 80 percent for “continuous speech.”

Fifth, certain languages are inherently less compatible with machine translation. AI struggles to accurately translate Korean’s “subject+object+verb” syntax when compared to English’s “subject+verb+object.” Turkic languages’ flexible syntax similarly “overwhelm” AI. Moreover, one must distinguish between high-context Asian and African cultures compared to the explicit verbal communication style of European languages. Culture-specific body cues are inherent to high-context languages, only detectable by a sharp-eyed military linguist. As one Israel Security Agency interrogator recounted of a particular mission, a suspect’s use of a certain Arabic hand gesture, rather than any verbal proclamation, convinced him of the man’s sincerity and thus facilitated more time for catching the real culprit—the suspect’s mother. Separately, examine the following: in Arabic, the word “landing” is different for planes (هبوط) and boats (إنزال); the distinction evidences how context determines vocabulary in Arabic, forcing a choice that English leaves implicit. Easy as it looks, the Massachusetts Institute of Technology Review concluded that “even the most sophisticated AI models had a 34 percent failure rate in maintaining context across multilingual translations.” NATO’s newly formed AI translation department likewise conceded that AI can’t handle the “political or historical context of a certain text.” Nearly 75 years after linguist and mathematician Yehoshua Bar-Hillel reflected on RBMT’s struggles with words that possess several related but distinct meanings (polysemy), today’s AI appears equally stumped. On the battlefield, the cumulative effect of such deceptively small mistakes can be exponentially more consequential. For example, machine translation in the Bosnian War was not only “unreliable,” but “in the worst cases fatally flawed.”

However, where military linguists truly outshine AI is in their ability to navigate languages’ culture and biases. Consider Persians’ bemusing “taarof” etiquette, according to which sellers declare their wares undeserving of payment («قابل نداره» meaning “it’s worthless”) in the name of honor. In cultures where human interaction is prized as a sign of respect, will AI translation win hearts and minds, even if it adopts the correct register? Unlikely. Studies agree: the Phraselator, used by US forces in Iraq, was shown in Nepali trials to impede interactions by stripping interlocutors of human connection. Additionally, military linguists smooth over cultural barriers and bridge the mismatch between operating styles; the past 70 years of American history are littered with attempted transformations of foreign militaries failing precisely because American advisers were not grounded in local warfighting traditions and culture.

The Coming War Needs Linguists

Despite unprecedented cross-Atlantic tension over Greenland leading to doomsday talk about the “end of NATO”, threats targeting the United States remain decidedly international and, as such, coalition warfare is here to stay. Against this backdrop, the security implications of neglecting military linguists are manifold and severe.

British politicians must address their own linguist-undermining policies—namely, unseriousness vis-à-vis defense spending. The military linguists upon which a nation relies for its defense are the last place where financial savings should factor into the equation. Yet, half-baked solutions abound; for example, the Liberal Democrats have recently proposed selling the public war bonds to fill a £20 billion defense funding hole.

It would be wrong to suggest that the US military’s linguist problems start and end with recruitment. First, Theater Special Operations Commands, responsible for selecting languages for prioritization, “don’t know what languages [they] need to learn,” even proving unable to inform the Government Accountability Office “how they arrived at their recommendations.” Second, linguists are misemployed after training. A US Army Arabic linguist posted in Iraq recalled how his team was composed of equal parts Arabic and Korean-language professionals and found themselves doing work “in any capacity but their own.” Last, recruitment gains vanish when linguists are not supported to maintain proficiency. The same US veteran explained that “many” do not pass annual recertification tests. But without linguists to train, these issues become secondary.

A Call to Action

Other militaries have already climbed out of the technology quagmire and forged a new path. In a 2025 interview, Israel’s head of the Intelligence Corps’ school of languages emphasized: “The war in Gaza has demonstrated clearly how wrong it was to believe that technology can replace human beings when dealing with language. In the coming years, the IDF intends to enlist more people who deal with language.” The United Kingdom and the United States should not need to repeat the same mistakes to learn these valuable lessons.

Undoubtedly, AI outperforms humans in quantity, speed, and cost any day—but not in quality. Instead of investing millions in stop-gap AI solutions, which at present are risk- rather than force-multipliers, the British and American militaries should tackle the issue at its core: a personnel shortage that only personnel can solve. As technology continues to dominate the battlefield, military linguists must remain the exception.

Image: Adobe Stock