OpenAI has a goblin problem.

Instructions designed to guide the behavior of the company’s latest model as it writes code have been revealed to include a line, repeated several times, that specifically forbids it from randomly mentioning an assortment of mythical and real creatures.

“Never talk about goblins, gremlins, raccoons, trolls, ogres, pigeons, or other animals or creatures unless it is absolutely and unambiguously relevant to the user’s query,” read instructions in Codex CLI, a command-line tool for using AI to generate code.

It is unclear why OpenAI felt compelled to spell this out for Codex—or indeed why its models might want to discuss goblins or pigeons in the first place. The company did not immediately respond to a request for comment.

OpenAI’s newest model, GPT-5.5, was released with enhanced coding skills earlier this month. The company is in a fierce race with rivals, especially Anthropic, to deliver cutting-edge AI, and coding has emerged as a killer capability.

In response to a post on X that highlighted the lines, however, some users claimed that OpenAI’s models occasionally become obsessed with goblins and other creatures when used to power OpenClaw, a tool that lets AI take control of a computer and apps running on it in order to do useful things for users.

“I was wondering why my claw suddenly became a goblin with codex 5.5,” one user wrote on X.

“Been using it a lot lately and it actually can't stop speaking of bugs as ‘gremlins’ and ‘goblins’ it's hilarious,” posted another.

The discovery quickly became its own meme, inspiring AI-generated scenes of goblins in data centers, and plug-ins for Codex that put it in a playful “goblin mode.”

AI models like GPT-5.5 are trained to predict the word—or code—that should follow a given prompt. These models have become so good at doing this that they appear to exhibit genuine intelligence. But their probabilistic nature means that they can sometimes behave in surprising ways. A model might become more prone to misbehavior when used with an “agentic harness” like OpenClaw that puts lots of additional instructions into prompts, such as facts stored in long-term memory.

OpenAI acquired OpenClaw in February not long after the tool became a viral hit among AI enthusiasts. OpenClaw can use any AI model to automate useful tasks like answering emails or buying things on the web. Users can select any of various personae for their helper, which shapes its behavior and responses.

OpenAI staffers appeared to acknowledge the prohibition. In response to a post highlighting OpenClaw’s goblin tendencies, Nik Pash, who works on Codex, wrote, “This is indeed one of the reasons.”

Even Sam Altman, OpenAI’s CEO, joined in with the memes, posting a screenshot of a prompt for ChatGPT. It read: “Start training GPT-6, you can have the whole cluster. Extra goblins.”

In the thick of the 1979 New York City winter, Isaac Asimov sat at his typewriter to reflect on what he considered America’s “cult of ignorance.” In the “My Turn” section of Newsweek, where Asimov eventually published the essay, he was credited as a biochemistry professor at Boston University School of Medicine, though the title was honorary by that point; he’d been a full-time writer for years, and with more than 200 books to his name, I imagine the words flowed from his mind like an open faucet. When the essay appeared on January 21, 1980, the world learned what one of its sharpest minds thought about America’s “growing strain of anti-intellectualism.” Asimov predicted, with eerie precision, how “expertise” would become shorthand for “elitism” and distilled democracy’s epistemic failure into what would become one of our most-quoted idioms: “My ignorance is just as good as your knowledge” (Asimov 1980).

In the nearly five decades since, the internet and social media have allowed the subjects of Asimov’s fears to multiply and speciate; misinformation has become the norm, society has celebrated the death of expertise, and the anti-intellectual strain he described has outgrown its American roots to become a worldwide burden. And when those pathologies extended into science and medicine, pseudoscience moved swiftly to capitalize.

Pseudoscience isn’t just fake science. An Armaani handbag and a three-dollar bill are both fake, but we don’t speak of pseudofashion or pseudomoney. Pseudoscience is both the imitation of science and its subversion. It borrows all of science’s aesthetics—the jargon, the studies, and the white coats—but rejects the safeguards and protocols that validate the whole institution. Asimov knew the con hinged on public uncertainty and insecurity: inspect any piece of pseudoscience closely, he wrote, and you’ll find a crutch, “a security blanket, a thumb to suck, a skirt to hold” (Asimov 1987).

Just like wealth, education, and healthcare, pseudoscience isn’t evenly distributed; some cultures tolerate it, some generate it, and others distribute it with astounding efficiency. And the world considers the United States a major exporter. In a pre-COVID-19 survey of respondents from twenty-five economies, the United States ranked highest among countries blamed for the “disruptive effect” of misinformation, receiving 35 percent of the vote; Russia received 12 percent and China 9 percent (Ipsos Public Affairs 2019). Public health experts voice similar concerns about misinformation propagated by the United States, warning of a “domino effect” that could spill across borders into hospitals, clinics, and labs worldwide (Mahase 2025). Environmental science shows the same pattern, with reports of American-style climate denial being exported overseas and taken up by far-right politicians seeking to weaken or derail green legislation in the European Union (Horton et al. 2025).

To the rest of the world, then, America is a place where pseudoscience levels up: where anti-vaccine stories become movements, fringe wellness interventions become brands, and anti-science rhetoric becomes ideology. The hypothesis isn’t simply that America is more predisposed to pseudoscience but that America may be better at generating and broadcasting it.

Stress Testing the Perception

But it’s just that: a hypothesis. If we invoke Feynman’s “first rule” and set aside perceptions and optics, what do the data say about the United States being especially prone to pseudoscience? A country with a strong predilection for pseudoscience would presumably have a disproportionately large associated workforce. Yet the labor statistics tell a different story. Consider chiropractic, a vocation unequivocally associated with pseudoscience by any skeptical definition, with an evidence-to-popularity discrepancy as large as I’ve seen for any intervention. There are at least 57,000 chiropractors in the country as of 2024 (Bureau of Labor Statistics 2025)—more than all the McDonald’s and Starbucks stores combined. It’s a lot, to be sure, but it’s only thirty-four chiropractors per 100,000 jobs, the same as Australia and lower than Canada with forty-eight. It’s the same story for homeopathy (Dossett et al. 2016), for which the United States expresses a relative interest similar to that of other developed nations. As for psychic services, Pew Research from 2024 shows 9 percent of Americans “regularly and seriously” use psychics, astrologers, and tarot-card readers, compared to 8 percent in the United Kingdom, 11 percent in Australia, 13 percent in France, and 19–20 percent in Japan, Singapore, and South Korea (Evans et al. 2025; Le 2025). Perhaps supply and demand aren’t clean measures of national predisposition because both are distorted by price, access, insurance coverage, and stigma. Or perhaps the United States’ appetite for pseudoscience merely looks modest because these are just three items in the country’s vast bullshit buffet.

What about interest? Google Trends provides an “interest by region” metric—how prominent a term is within a country’s searches. Several pseudoscientific labels register strongly in the United States but not in a way that makes it appear exceptional. America ranks first in detox searches, meaning the term accounts for a larger share of searches there than anywhere else. But it ranks fourth for toxins,1 eighth for chakras, and fifth for healing crystals (for which the United Kingdom, oddly enough, is top).

Even broader measures of “trust in science” undermine public perceptions. In an international survey with nearly 72,000 responses, Viktoria Cologna and colleagues (2025) constructed a trust-in-science index, where 1 indicates very low trust and 5 indicates very high trust. The United States scored 3.86, above the global average of 3.62 and in the same neighborhood as the United Kingdom (3.82), Canada (3.81), and Australia (3.91). Based on these metrics, it would be disingenuous to describe America as uniquely pseudoscientific or uniquely distrustful of mainstream science.

The Divided States of America

My doctoral supervisor was a committed proponent of the scientific method. He was also patient, exacting, and impossibly attentive to small things. He had a line for moments like this: “The devil is in the detail, Nick.” On first pass, the numbers above are unremarkable, other than contradicting a stubborn public intuition. But the devil is in the detail; averages are seductively misleading because they conceal inconvenient truths. Politicians and many researchers love them for this reason; you can hoist an average in the air and parade it around like a champion gladiator, hoping nobody will examine what’s going on behind the pageantry. Two countries can be nothing alike—one packed with moderates, the other carved into multiple extremes—and still share the same average.

This “bimodality” is precisely what we see in the United States. The country isn’t uniformly more pseudoscientific, as the data above will attest. But among developed nations, it’s uniquely divided. These political, religious, and socioeconomic divisions shape values, beliefs, and the ways people interpret expertise (Cologna et al. 2025).

Take political division, for example. According to the V-Dem political polarization index—an expert-coded scale from -4 (low polarization) to +4 (high polarization) that captures how far a society is fractured into hostile camps—the United States exhibits more intense political division than most high-income nations. The United States ranks twenty-eighth worldwide, with a polarization score of +1.79. That’s higher than Canada (−0.39), the United Kingdom (−0.24), China (-0.1), Russia (+1.44), and Brazil (+1.69), and it soars over the European average of +0.23. What’s striking is the recency of the shift: 2008 was the hinge point, when the United States moved from below zero (predominantly neutral) to above zero (divided).

While trust in science is skewed liberal (Kennedy and Kikuchi 2026), the problem is bipartisan. Neither side holds a monopoly on science denial or conspiracy thinking; it’s just that the left and right favor different falsehoods (Enders et al. 2023). The right tends to oppose climate policies (favoring fossil fuels), public health measures, and vaccines (especially post-COVID-19), and they have lower trust in scientists and experts. The left, by contrast, exhibits greater moralized “naturalness,” encompassing detox culture, moral panic surrounding chemicals, and a preference for “natural” remedies that can extend to vaccine hesitancy and GMO opposition, where beliefs converge with the right. Move far enough in either direction, and science and reason become the enemy.

Adding complexity, it’s not simply a left-versus-right issue. Compared with citizens of peer high-income countries, Americans remain unusually religious; more Americans are likely to say religion is “very important” in their lives (Fahmy 2018). But it’s also increasingly divided on religious opinion, with a growing number of young people religiously unaffiliated, precipitating an age gap in religiosity that’s never been wider (Smith 2025). Economically, too, the United States is an outlier. On the Gini coefficient of income inequality, America ranks seventy-fifth among 206 countries—higher (worse) than the United Kingdom, Germany, France, Sweden, Canada, Australia, Japan, Switzerland, the Netherlands, Norway, and Denmark, among others (World Population Review 2026). These political, religious, and economic divides do more than sort people into distinct voting blocs; they independently predict conspiracy beliefs and trust in scientists (Kennedy and Kikuchi 2026; Frenken et al. 2023; Wellcome Global Monitor 2018).

So, while Americans aren’t uniquely credulous, they are uniquely divided—into ever finer shards of a shattered Venn diagram, each with its own information ecosystem, values, authorities, identity labels, and institutional trust. Division is an ideal growth medium for bad ideas. Pseudoscience thrives in the United States because of its multiple isolated subgroups, each reinforced by shared identity, insulated from corrective messaging, and easy to target and manipulate for profit or power. This isn’t the United States of Pseudoscience; it’s the Divided States of America.

Markets, Memes, and Misinformation

Is it plausible that a few overlapping subgroups can propagate a country’s worth of pseudoscience? They may be numerically small, but in a system designed to monetize attention, they can exert disproportionate influence. Capitalism is the first ingredient. In spring 1792, after a financial scare spooked the public into pulling money from the banks, twenty-four New York stockbrokers met, supposedly under a buttonwood tree on Wall Street, and drafted the Buttonwood Agreement: a pact designed to bring order to a noisy, reputation-driven trade. By establishing common terms and restricting dealings to known parties, it helped secure the market and restore public trust.

That modest agreement foreshadowed America’s modern capitalist machine—one that, by many measures, outperforms its peers. That system rewards competition and productivity, strengthens the economy, and raises living standards for many. It provides consumers with greater choice and stigmatizes them less for indulging in it. But there’s no such thing as a free lunch; capitalism weakens guardrails against fraud, unsafe products, and misleading advertising. It financially rewards persuasive claims rather than true ones. And wellness markets, for instance, are free to specialize in credence goods—products whose value is hard to verify even after purchase. The best storytellers and the boldest claims tend to prosper in industries where sellers can stack the deck (like Tarot reading). Capitalism doesn’t create pseudoscience, but it reliably creates the conditions that allow it to thrive.

Just as important is what capitalism did for export. Rules, contracts, and commissions lower the costs (and risks) of doing business with strangers, allowing markets to expand beyond face-to-face interactions. Over time, American commerce moved abroad with newfound confidence. The New York Stock Exchange now lists hundreds of international firms across dozens of countries, and capital crosses borders in the time it takes for a screen to refresh. We live in what Kenichi Ohmae called a “borderless world”—a global economy with the free flow of goods. And a system as powerful as America’s, which can scale and export markets, can do the same with values and ideas, including bad ones.

This brings us to the second ingredient: bad ideas spread fast. Pseudoscience succeeds because it satisfies human appetites—certainty, outrage, hope, and tribal loyalty. It compresses the complexity of vaccines, climate change, and the pursuit of wellness into a series of simple, shareable stories and memes. And in today’s attention economy, those ideas spread like cultural contagions: replicating fast and moving at the speed of light through a frictionless ether that rewards novelty over nuance. It’s why fake news reaches up to 100 times as many people as the truth in every category (Vosoughi et al. 2018). Reality, by contrast, tends to be cautious, technical, and slow; a lie will travel twice around the world before the truth has tied its shoes. In our borderless world, digital networks have done for pseudoscience what shipping lanes and air travel once did for bacon, bread, and beaver skins.

The Counter Export

The good news is that the United States isn’t just spreading pseudoscientific narratives internationally; it’s also deployed the single most influential effort to combat them. In 1976, amid rising public fascination with mysticism, UFOs, conspiracies, and the paranormal, philosopher Paul Kurtz convened a Suicide Squad of scientists and thinkers to apply the tools of inquiry to extraordinary claims. Alongside Asimov were the often imitated, never duplicated Carl Sagan, James “The Amazing” Randi, and the formidable skeptic trio of Skinner, Gardner, and Hyman. They stood together against a deluge of nonsense—the “growing strain” that threatened America’s place in the global knowledge economy. Over the decades, as spoon bending and séances gave way to wellness grifts and algorithmic misinformation, the Committee for the Scientific Investigation of Claims of the Paranormal (CSICOP), evolved into the Committee for Skeptical Inquiry (CSI) to meet society’s growing need for skepticism in a digital age. And if America exports pseudoscience, it also exports some of its loudest correctives: a long roster of tireless skeptics and science communicators.

As we mark the organization’s fiftieth anniversary, it’s easy to feel dejected at the current state of pseudoscience in America and beyond. But the more useful question is the counterfactual: What would the landscape look like without the past fifty years of organized skepticism—without the scientists, writers, investigators, and educators doing the often thankless work; without the founding visionaries and the generations of skeptics they inspired? Because while misinformation is emboldened by every baseless claim, every quack remedy, and every piece of divisive public-health legislation, so too is the countermovement; we grow stronger with every seed of doubt, every myth debunked, and every new skeptic willing to stare down the cult of ignorance.

Note

Zambia tops the list. Their interest isn’t driven by a thriving detox economy but rather by catastrophe; almost a century of unregulated mining and smelting has left Kabwe, the capital of Zambia’s Central Province, with some of the world’s worst lead contamination.

References

Asimov, Isaac. 1980. A cult of ignorance. Newsweek (January 21): 19.

———. 1987. Past, Present, and Future. Buffalo, NY: Prometheus Books.

Bureau of Labor Statistics. 2025. Occupational projections and worker characteristics. Online at https://www.bls.gov/emp/tables/occupational-projections-and-characteristics.htm.

Cologna, Viktoria, Niels G. Mede, Sebastian Berger, et al. 2025. Trust in scientists and their role in society across 68 countries. Nature Human Behaviour 9(4): 713–730.

Dossett, Michelle L., Roger B. Davis, Ted J. Kaptchuk, et al. 2016. Homeopathy use by US adults: Results of a national survey. American Journal of Public Health 106(4): 743–745.

Enders, Adam, Christina Farhart, Joanne Miller, et al. 2023. Are republicans and conservatives more likely to believe conspiracy theories? Political Behavior 45(4): 2001–2024.

Evans, Jonathan, Kirsten Lesage, William Miner, et al. 2025. Believing in spirits and life after death is common around the world. Pew Research Center (May 6). Online at https://www.pewresearch.org/religion/2025/05/06/believing-in-spirits-and-life-after-death-is-common-around-the-world/.

Fahmy, Dalia. 2018. Americans are far more religious than adults in other wealthy nations. Pew Research Center (July 31). Online at https://www.pewresearch.org/short-reads/2018/07/31/americans-are-far-more-religious-than-adults-in-other-wealthy-nations/.

Frenken, Marius, Michał Bilewicz, and Roland Imhoff. 2023. On the relation between religiosity and the endorsement of conspiracy theories: The role of political orientation. Political Psychology 44(1): 139–156.

Horton, Helena, Sam Bright, and Clare Carlile. 2025. Revealed: US climate denial group working with European far-right parties. The Guardian (January 22).

Ipsos Public Affairs. 2019. 2019 CIGI-Ipsos Global Survey – Part 3 Social Media, Fake News & Algorithms. Online at https://www.cigionline.org/sites/default/files/documents/2019%20CIGI-Ipsos%20Global%20Survey%20-%20Part%203%20Social%20Media%2C%20Fake%20News%20%26%20Algorithms.pdf.

Kennedy, Brian, and Emma Kikuchi. 2026. Do Americans think the country is losing or gaining ground in science? Pew Research (January 15). Online at https://www.pewresearch.org/science/2026/01/15/do-americans-think-the-country-is-losing-or-gaining-ground-in-science/.

Le, Valerie. 2025. Psychic Services in the US Industry Analysis. Online at https://www.ibisworld.com/united-states/industry/psychic-services/4413/.

Mahase, Elisabeth. 2025. US vaccine misinformation is having “frightening” ripple effect in Europe, leaders warn. British Medical Journal 391: r2167.

Our World in Data. 2025. Data: Political polarization score—V-Dem. Online at https://ourworldindata.org/grapher/political-polarization-score.

Smith, Gregory A. 2025. Religion holds steady in America. Pew Research Center (December 8). Online at https://www.pewresearch.org/religion/2025/12/08/religion-holds-steady-in-america/.

Vosoughi, S., D. Roy, and S. Aral. 2018. The spread of true and false news online. Science 359(6380): 1146–1151.

Wellcome Global Monitor. 2018. Chapter 3: Trust in science and health professionals. Online at https://wellcome.org/insights/reports/wellcome-global-monitor/2018/chapter-3-trust-science-and-health-professionals.

World Population Review. 2026. Wealth Inequality by Country 2026. Online at https://worldpopulationreview.com/country-rankings/wealth-inequality-by-country.

Getty Images

Dubrovnik, Croatia

Europe’s data center markets need more energy. Demand for energy to power the data centers is projected to grow from 10 gigawatts in 2024 to 35 gigawatts by 2030. The only place on the Continent with the power and political will to meet that demand is one most people haven’t thought of—and the war in the Gulf has reinforced the urgency to secure artificial-intelligence infrastructure for the U.S. and our allies.

With Iran threatening “complete and utter annihilation” of data centers in Abu Dhabi and calling American tech companies legitimate military targets, the question is no longer whether the West should rethink where to build its crucial AI infrastructure, such as data centers. The question is where they should go.

Croatia is geographically removed from every active conflict zone. As part of the North Atlantic Treaty Organization, the European Union, the Schengen Area and the eurozone, Croatia delivers regulatory predictability and political durability that institutional capital requires. A data center in Croatia is infrastructure protected by our sacrosanct collective defense guarantee.

Croatia’s grid is among the cleanest in Europe: Renewables supplied more than 52% of all electricity consumed in 2025, with a further 15% from nuclear. Electricity prices run significantly below the European average. The climate and access to natural water resources provide thermal management advantages that reduce energy consumption and operating costs, alongside high-voltage grid capacity that congested Western European hubs can’t offer.

European digital infrastructure investment is projected to reach €100 billion by 2030. EU data sovereignty rules are compelling major U.S. technology companies to store European data within EU borders—creating demand for U.S.-owned, EU-based capacity that doesn’t yet exist.

I have spent my career betting on overlooked infrastructure. The underlying logic is always the same: Find the structural gap, understand why it exists, and build first.

This is why I am in Croatia to announce Project Pantheon: a next-generation 1-gigawatt data-center campus on 310 acres. The campus will host 800 megawatts of usable IT load—enough to power the city of Baltimore. The facility is designed to be more resilient than most data centers in operation today, reaching 99.99999% uptime just as hyperscalers need facilities that can run massive workloads continuously and draw huge amounts of power without interruption.

Croatia sits between key areas of the EU, opening corridors to regions in Central and Eastern Europe as well as Greece and Turkey. Pantheon Atlas LLC’s total initial investment in the project is $13.5 billion with the potential to grow to beyond $50 billion.

This is the largest single private investment in Croatian history. No U.S.-led gigawatt-scale AI-optimized facility currently exists in Central or Eastern Europe. Pantheon will be accessible to European defense, intelligence and commercial users, thereby helping the EU reach its goal of tripling European data center capacity within five to seven years.

The war in the Gulf has revealed that concentrating AI infrastructure in jurisdictions where a single strike can take a $30 billion investment offline is a strategic liability. Croatia offers the advantage of a collective defense guarantee that no insurance premium can match.

Mr. Rich is managing partner of Pantheon AI and founder of 24 Ventures, a venture firm headquartered in Buffalo, N.Y.

Free Expression: A Daily Newsletter From WSJ Opinion

Get the Newsletter

‘Microlooting’ Is a Luxury Belief

AI Freedom Can Be Creative Slavery

A Hot Take Served Cold

About Free Expression

Copyright ©2026 Dow Jones & Company, Inc. All Rights Reserved. 87990cbe856818d5eddac44c7b1cdeb8

Appeared in the April 28, 2026, print edition as 'The Best Place for a Data Center? Croatia'.

“We stand alongside Lebanon in disarming Hezbollah,” Syrian President Ahmed al-Sharaa declared last month. No other Arab head of state has called for taking away Hezbollah’s weapons. Until Sharaa overthrew Bashar al-Assad’s regime in December 2024, Syria was helping to arm Hezbollah. Now, Syria finds itself unexpectedly sharing an adversary with Israel.

On April 19, Syria’s Interior Ministry announced that it had thwarted a sabotage plot in Quneitra province orchestrated by a cell linked to Hezbollah. According to Syrian authorities, the operatives had disguised a civilian transport vehicle to conceal rocket-launching equipment for a surprise attack. The rockets reportedly bore the slogan “Victory for our brothers in Lebanon and Palestine.” Days earlier, Syrian authorities announced that they had also disrupted a plot targeting a religious figure in Damascus, arresting suspects who the Interior Ministry said were linked to Hezbollah. The reported target, Rabbi Michael Khoury, is one of numerous Jewish communal leaders to have visited Syria following the fall of Assad and in December had been part of a delegation of Syrian American Jews attending the reopening of the Elfrange Synagogue in the Syrian capital.

Syria’s leadership, despite its constraints, is beginning to show a measurable willingness to curb Hezbollah’s footprint on its territory. The state has interdicted hundreds of weapons and rockets destined for Hezbollah in Lebanon since Shaara took power. The shift may be gratifying to Israel, but it is less about any sort of alignment with Israel than it is about converging interests. On the contrary, Syria and Israel have clashed repeatedly, especially over the status of the former’s Druze minority. Yet Shaara understands that allowing Hezbollah to exploit Syrian territory as a smuggling route would entrench the group’s presence and perpetuate instability inside a state struggling to reassert control. Sharaa himself has framed his actions as an attempt to “save the region” by preventing Syrian territory from becoming a launchpad for Hezbollah attacks.

This convergence of interests hardly erases the deep mistrust between the two countries. But it does create an opening for cooperation based on concern about a common threat. Sharaa has avoided naming Hezbollah explicitly as an adversary, yet Syria’s own security apparatus has warned that the group’s presence on the border with Syria “has become a threat.” It has also linked Hezbollah’s local networks to “sabotage cells” aimed at “undermining stability.” In practice, that cooperation could take the shape of deconfliction channels and intelligence sharing through intermediaries, particularly when it comes to the smuggling routes that Hezbollah uses in Syria to rearm.

For many Syrians, hostility toward Hezbollah goes much deeper than a concern about weapons trafficking. Syrians consider the group inseparable from the late Assad regime, sharing its complicity in massacres carried out during the civil war. That anger is spilling into public life. At a recent basketball game in Damascus between the Syrian and Lebanese national teams, what began as a show of reconciliation turned into a venting of enmity. Sharaa himself was in attendance and spoke of “putting an end to the tragedies” of both countries, yet thousands in the stands erupted into a chant of “God curse your soul, Nasrallah,” a direct rebuke of former Hezbollah Secretary-General Hassan Nasrallah and a reminder that, for many Syrians, the wounds of the group’s role in the conflict remain far from healed.

Under Assad, Hezbollah embedded itself deeply inside the Syrian state. At its peak in 2017, the group maintained an estimated 7,000 to 10,000 fighters in Syria. Hezbollah also recruited Syrians and cultivated local auxiliaries, including formations such as the so-called Golan File in southern Syria, positioned near the Golan Heights to threaten Israel.

After Assad fled to Moscow in December 2024, Hezbollah shifted from overt military entrenchment to a shadowy model built on covert local cells such as the Islamic Resistance Front in Syria, which have since claimed attacks against Israel launched from southern Syria and framed Sharaa as a “puppet for the Turkish, American, and Israeli security apparatuses who wants to bury the resistance identity in Syria.”

In one account carried by the London-based Asharq al-Awsat newspaper, an Iraqi official described a quiet exchange with Iranians over a shipment of weapons. The Iraqi was told that the Iranians and Hezbollah “have smuggling networks they rely on. … There are those who can deliver the shipments all the way to Damascus.” The official described a network stitched together from the remnants of the Assad regime and veteran traffickers—“individuals from different sectarian backgrounds and nationalities … some with long experience along smuggling lines.” If pockets of resistance to Sharaa’s government persist, this ecosystem has room to breathe and to aid Hezbollah’s rearmament and regeneration.

Hezbollah appears to recognize that restoring the old order may be out of reach. Its objective now is more pragmatic, focused on preserving access and ensuring that Syria remains usable as a conduit—even in a hostile political environment.

The new Syrian state holds territory in fragments—present in many places but in firm control of only a few. Units cycle from one flashpoint to the next, chasing Islamic State cells in the center, maintaining tense lines with the Syrian Democratic Forces in the northeast, and navigating friction with Druze factions in the south. These thinly spread forces are not sufficient to fully control Syria’s porous borders.

What complicates matters for Syria, and by extension Israel, is Lebanon’s inaction against Hezbollah. Syrian outlets aligned with Sharaa have continuously reiterated that “the Lebanese government must be firm and serious” in disarming Hezbollah. On social media, the language is less diplomatic. Syrian analysts have pointed to Hezbollah’s rearmament to argue that cooperation with Beirut is futile and that the “Syrian army needs to advance into Lebanon and disarm Hezbollah itself.”

This is where Israeli and Syrian interests most clearly converge, but so far convergence has not translated into cooperation. Over the past year, efforts to bridge the gap between the two countries “have stalled,” according to Syrian Foreign Minister Asaad al-Shaibani. This tension has eroded what had been a tentative opening, particularly during the period when Sharaa publicly hinted at normalization with Israel.

Israel’s distrust stems from Sharaa’s militant past and fears that foreign jihadis still operate within his ranks, all compounded by his backing from Turkey. Israel, for its part, has waged a sweeping campaign across Syria—launching hundreds of strikes and stepping in to back Druze factions in Suwayda against the government. Damascus has cast these moves as an attempt to fracture the state through armed proxies. That divide has fueled deepening animosity where even Syrian soldiers have been filmed chanting slogans tied to Hamas, revealing how, for both sides, the situation remains fragile and combustible.

Despite this, Israeli media and defense reporting suggest that at least some Israeli officials view Syria’s leadership favorably on the issue of interdicting Hezbollah’s weapons smuggling. Reportedly, the Israeli military had given Sharaa and his forces “very high marks” for preventing Iran and Hezbollah from smuggling arms into Lebanon.

Moving forward, Isreal would like to see Damascus pursue Hezbollah more aggressively, moving beyond episodic seizures toward a sustained effort to dismantle the group’s smuggling networks. This would involve targeting the facilitators, financial channels, and local intermediaries that continue to enable Hezbollah operations. Israel, for its part, will continue to act against high-value threats in Syria. But it should avoid steps that could prove destabilizing to the Syrian state. The issue of the Druze, for example—while critical to Israel—should be addressed through the September 2025 U.S.-backed arrangement in which Israel plays a supportive role, rather than through independent Druze armed factions.

If there is to be further progress, it depends on Washington. Over the past year, U.S.-mediated mechanisms have emerged to manage friction and prevent escalation. Now, the United States could help pressure both sides to work together more effectively. Indirect coordination or tacit understandings about red lines could significantly reduce friction while tightening constraints on Hezbollah’s movement in Syria. Crucially, Israel could provide Damascus with intelligence that would help it crack down on Hezbollah-linked networks, particularly those tied to weapons transfers and cross-border operations.

Even limited alignment between Israel and Syria can begin to narrow the space that Hezbollah has exploited. Over time, sustained coordination and small demonstrations of good faith could lay the groundwork for something more durable. Normalization may be distant, but incremental progress can gradually expand the constituency willing to consider it.