A new clinical case report describes how a 60-year-old man developed bromism—an archaic psychiatric syndrome rarely seen since the early 20th century—after replacing table salt with sodium bromide based on information he said he gleaned from a chatbot. The case, published this week in Annals of Internal Medicine: Clinical Cases, underscores the dangers of relying on unvetted artificial intelligence (AI) advice for health decisions and arrives as Thailand accelerates efforts to reduce population salt intake to curb hypertension and heart disease. Investigators said the man mistakenly treated a chemical substitution used in cleaning and pool treatment as if it were a safe dietary swap, leading to psychosis, hospitalization, and weeks-long treatment for bromide toxicity. The report has triggered global debate over AI safety guardrails in consumer health and the practical, safer paths Thais can take to cut sodium without risking harm (acpjournals.org; 404media.co; arstechnica.com).
The case is startling both for its medical rarity and its modern origin story. According to the clinicians, the patient arrived at an emergency department complaining of hallucinations and paranoid thoughts that he was being poisoned. Once stabilized with fluids for dehydration, he revealed that he had attempted to “eliminate chloride” from his diet and, for three months, had replaced sodium chloride (table salt) with sodium bromide purchased online. The case report notes that this decision followed online queries and “consultation with ChatGPT,” in which he had read that chloride could be swapped with bromide—likely in an industrial or cleaning context, not for human consumption (404media.co; acpjournals.org). Over his first day in hospital, his paranoia and visual and auditory hallucinations worsened, culminating in an attempted escape and a brief involuntary psychiatric hold. Blood tests later revealed a bromide concentration around 1,700 mg/L—orders of magnitude higher than typical reference values—and doctors initiated “aggressive saline diuresis,” flooding him with fluids and electrolytes to speed the excretion of bromide. He required about three weeks of inpatient care before recovering (arstechnica.com).
For Thai readers, the story lands at a critical juncture. Thailand’s average daily sodium intake is roughly 3,636 mg—about 9.1 grams of salt—almost double the World Health Organization’s recommended limit of less than 2,000 mg sodium (under 5 grams of salt) per day. Reducing salt is a national priority to tackle hypertension and noncommunicable diseases, and Thai agencies, professional societies, and civil society partners have spent a decade building a “low salt” movement—publishing sodium data, engaging food companies, and pushing labeling and procurement policies. The Thai Low Salt Network, supported by the Thai Health Promotion Foundation and partners, catalyzed many of these efforts, including a nationally representative 24-hour urinary sodium survey and ongoing advocacy for mandatory targets in packaged foods (thelancet.com/regionalse-asia; who.int). Against this backdrop, a case of salt substitution gone catastrophically wrong is a pointed reminder: cutting sodium is good; substituting with industrial chemicals or veterinary drugs is not.
Bromism—the syndrome caused by excessive bromide accumulation—was once so common that it accounted for an estimated 5–10% of psychiatric hospital admissions in the United States before mid-20th century reforms saw bromide sedatives pulled from use. Bromide salts (such as potassium bromide) were popular sedatives in the late 1800s and early 1900s, but prolonged use could lead to neurological impairment, grotesque skin rashes (“bromoderma”), and psychiatric disturbances ranging from confusion to frank psychosis. As most countries withdrew bromides from routine human medical use, the condition all but vanished from modern psychiatry textbooks, surviving mainly in toxicology case reports and veterinary practice—where bromide still has a role, under veterinary supervision, in managing canine epilepsy (wikipedia.org; pmc.ncbi.nlm.nih.gov). The US Food and Drug Administration, for instance, has not allowed bromide sedatives in human medicine for decades and in 2024 also revoked authorization for brominated vegetable oil in foods, reflecting evolving evidence on brominated compounds’ risks (fda.gov).
Clinically, bromism can be deceptive. Because bromide is a negatively charged halide, it can trick common laboratory ion-selective electrodes into over-reading chloride, producing spurious hyperchloremia and even a “negative anion gap” that confuses diagnosis. Several recent reports—some involving patients who purchased bromide-containing supplements online—highlighted this diagnostic pitfall: standard panels may show very high “chloride” with normal sodium and pH, when in fact bromide is interfering with the assay. Confirmatory bromide levels can arrive days later, so toxicologists advise a high index of suspicion when neuropsychiatric symptoms accompany these lab anomalies, especially if there is any hint of bromide exposure. Treatment hinges on accelerating bromide elimination: vigorous IV fluids and electrolytes (saline diuresis) and, in severe cases, hemodialysis (pmc.ncbi.nlm.nih.gov; pmc.ncbi.nlm.nih.gov).
The new case is not simply a tale of chemical toxicity; it is also a case study in AI-era risk. The authors note they did not have the patient’s actual chatbot logs and therefore cannot say definitively what wording or context the AI provided. In their own testing of an earlier model, they found bromide featured in the response to questions about substituting chloride, but without a conspicuous health warning or the kind of clarifying triage a clinician would provide (for example, “Are you asking about cleaning, disinfection, or diet?”). When later versions were queried, the chatbot reportedly performed better: it asked clarifying questions and placed bromide only in the cleaning/disinfection context, not as a food ingredient or dietary swap (arstechnica.com). In other words, a combination of user misunderstanding, industrial chemistry context, and AI’s uneven safety behavior likely contributed to this extreme, but preventable, outcome. Coverage by independent outlets has echoed the case authors’ cautious framing: that the patient “did his own research,” misapplied information, and suffered a serious, outdated syndrome once common only when bromides were medicines (404media.co).
Global health authorities have been warning for over a year that large AI models, while powerful, are not medical devices and require careful governance. The World Health Organization has urged “caution to be exercised” in using AI chatbots in health settings and has issued more than 40 recommendations for safe, ethical, and effective deployment of large multimodal models in healthcare, including strong transparency, risk management, post-market monitoring, and mechanisms to mitigate misinformation and bias (who.int; who.int). Within ASEAN, regional guidance released this year emphasizes risk identification and human oversight for generative AI across sectors. Thailand’s digital agencies are meanwhile exploring local-language AI models to improve public services, making this a timely moment to bake “safe-by-design” features into any health-adjacent chatbot deployed for the Thai public (asean.org; govinsider.asia).
For households in Thailand trying to cook with less salt, the lesson is emphatically not to abandon sodium reduction—it is to do it safely and realistically. The Thai Low Salt Network and partners have documented how much salt creeps into the national diet: packaged foods, condiments like fish sauce and soy sauce, and street food dishes can all pack high sodium loads. Their research estimated average sodium intake at 3,636 mg/day, and they have pressed for procurement standards in public canteens, clearer front-of-pack labels, and reformulation targets for staples like instant noodles and sauces. Progress is visible in some product categories, but experts stress that voluntary approaches have limits and mandatory standards will likely be needed to meet the national 30% reduction goal aligned with WHO targets (pmc.ncbi.nlm.nih.gov; who.int).
Cultural context matters here. Thai cuisine balances heat, sourness, sweetness, and umami—often with a generous hand on condiments and stocks. Asking for “wan-sai noy” (less salty) at a khao rad gang stall or favoring fresh herbs, lime, and chilies for flavor can meaningfully reduce sodium without sacrificing taste. Swapping out high-sodium sauces in home cooking for lower-sodium versions, tasting before adding fish sauce at the table, and cutting back on processed snacks are all practical, safe steps that add up. Critically, so-called “salt substitutes” sold to consumers are typically potassium chloride (KCl) blends—never bromide—and even KCl requires caution for those with kidney disease or those on certain medications, which is why pharmacists and doctors should be consulted before use. No credible dietary guidance endorses sodium bromide as a food ingredient; it is not a culinary material. Bromide compounds today are encountered mainly in specialized industrial applications, pool and hot tub treatments, and in veterinary medicine under prescription for canine epilepsy—not in household cooking (pmc.ncbi.nlm.nih.gov).
The medical details in this week’s case highlight the consequences when chemistry leaves the lab and enters the kitchen. Bromide accumulates in the body over weeks due to its long half-life. Neurotoxicity can present as confusion, ataxia, memory impairment, or psychosis; dermatologic reactions can be severe; and lab artifacts can hinder diagnosis, as bromide fools chloride electrodes. Physicians now train for rare toxicology scenarios, but cases remain uncommon enough that patients and families may be blindsided. In this case, fluids and time helped—saline diuresis flushes bromide—but the hospital stay was prolonged and frightening. The patient’s reported bromide concentration—about 1,700 mg/L, versus a reference window under 10 mg/L—reflects sustained, daily intake over months and would not arise from incidental exposure. It is a red flag for anyone tempted by chemical “solutions” to lifestyle problems (arstechnica.com; pmc.ncbi.nlm.nih.gov).
The AI dimension demands nuance. Large language models can be genuinely helpful at explaining general nutrition concepts, generating meal ideas within constraints, or pointing to authoritative sources. At the same time, they are known to produce confident mistakes, omit crucial warnings, or collapse industrial and dietary contexts unless prompted very specifically. The clinicians who authored the case report observed that earlier models did not ask why bromide was being considered and did not issue a direct health warning; later models performed better in their ad hoc testing but still placed bromide only in non-dietary contexts. That is a microcosm of the broader policy challenge: as AI improves, people will ask it more health questions. Without clear guardrails and consumer education, some will misinterpret the output—especially when seeking extreme diet hacks. WHO and other bodies have been explicit that AI systems used in health should be subject to robust evaluation, transparent labeling when chatbots are not medical devices, and clear pathways for users to escalate to qualified professionals (who.int; who.int).
In Thailand, authorities have begun moving against online health misinformation more broadly, and regulators are partnering with e-commerce platforms to tighten oversight of health products sold online. Those efforts should extend to high-risk advice in consumer AI apps and to the sale of compounds like bromide salts to the general public without clear, lawful use cases. Public health messaging can go further, too: when agencies encourage salt reduction—and they should—they can pair the message with simple, Thai-specific “do this, not that” guidance to preempt dangerous improvisation. For instance: “Use herbs, citrus, and chili instead of adding extra fish sauce; choose low-sodium soy sauce; check front-of-pack sodium scores; ask your pharmacist before buying any salt substitute” (tilleke.com; nationthailand.com).
Historical perspective helps explain the headline’s “19th-century” framing. Bromide sedatives were a mainstay in the Victorian and Edwardian eras. People used them for anxiety and sleep, sometimes daily, sometimes in large doses. As toxic effects accumulated—neurological and psychiatric—“bromism” became a recognized syndrome. Modern psychiatry consigned bromism to the dustbin as safer drugs and regulations emerged, and it persists today mostly in toxicology teaching cases and veterinary textbooks. That is why this week’s report is jarring: a 21st-century consumer, misled by a digital mash-up of chemistry and wellness ideas, reconstructed a Victorian-side-effect profile in his own body (wikipedia.org). The irony is that bromide remains a carefully used tool in veterinary neurology; in dogs, potassium bromide can be an effective add-on antiseizure drug, with dosing tailored by veterinarians and serum levels monitored. That clinical use does not translate to human diet or cooking—ever—and confusing those domains is precisely the kind of category error that AI prompts, unless systems are designed to aggressively disambiguate context (pmc.ncbi.nlm.nih.gov).
What happens next? Expect three developments. First, more scrutiny of AI guardrails in consumer apps, especially those that touch diet and wellness. Thailand’s planned local-language models for public services will need clear exclusions, escalation pathways to licensed professionals, and warnings in Thai that certain queries cannot be safely answered by a chatbot. Second, renewed emphasis on sodium reduction through food environment changes rather than household chemistry experiments: mandatory sodium targets for packaged foods, lower-sodium reformulations of sauces and seasonings, clear front-of-pack labeling, and public procurement standards for schools, hospitals, and canteens are all evidence-based strategies Thailand has already scoped and should accelerate. Third, better public education on what “salt reduction” means in practice for Thai kitchens and street food culture, with pragmatic swaps and ordering tips affirmed by dietitians and public health experts. If those three strands advance in tandem, Thailand can make real progress on sodium without courting exotic toxicities (pmc.ncbi.nlm.nih.gov; who.int).
For readers looking for concrete steps, here is a practical, Thai-focused plan to cut sodium safely—no AI hacks required:
Aim for less than 2,000 mg of sodium per day (under 5 g of salt), as recommended by WHO. Learn to translate labels: 1 g sodium equals about 2.5 g of salt. Many condiments list sodium per serving; measure once to see what your usual splash really adds up to (who.int).
In restaurants and street stalls, ask for “less salty” versions and taste first before reaching for fish sauce or soy sauce. Favor clear soups over rich stocks, and choose dishes naturally lower in sodium (stir-fried vegetables with garlic and chili, grilled fish with lime and herbs, som tam with less fish sauce).
At home, build flavor with lemongrass, kaffir lime leaf, galangal, basil, coriander root, garlic, and chilies; add acidity with lime or tamarind. Use low-sodium soy and fish sauces, and limit MSG and seasoning powders that may add sodium.
Read labels on packaged foods and pick lower-sodium options for instant noodles, snacks, and processed meats. Where front-of-pack labeling exists, use it.
If you are considering a “salt substitute,” speak with a pharmacist or doctor first—especially if you have kidney disease, diabetes, heart failure, or are on blood pressure medicines. Consumer salt substitutes generally contain potassium chloride, which can be unsafe for some people. Never use bromide compounds for cooking or diet.
For personalized diet advice, consult a licensed dietitian or physician. If you ask a chatbot for general cooking ideas, verify any health claims against authoritative sources like WHO or Thailand’s Ministry of Public Health sites, and avoid acting on chemical substitution advice.
If you or a family member develop unexplained confusion, rashes, or psychiatric symptoms while using any supplement or “salt alternative,” seek medical care and tell clinicians exactly what you’ve been taking; it can speed diagnosis.
The week’s case report is a sobering story, not a reason to fear technology or abandon salt reduction. It reminds us that public health progress is about systems and habits, not shortcuts. Thailand has spent years building the backbone for safer, lower-sodium diets; that work can continue as the country also builds trustworthy, Thai-language digital tools that know when to say, “This is a question for your doctor.” As global authorities keep emphasizing, AI can support health—but it cannot replace human clinical judgment, regulatory oversight, or common-sense kitchen wisdom (who.int).
Sources referenced in this report include the new case description in Annals of Internal Medicine: Clinical Cases; news coverage that reviewed the paper and spoke with clinical toxicology experts; WHO guidance on sodium and on AI in health; recent toxicology reviews of bromide interference in lab testing; and Thai public health research on sodium intake and policy. Readers can explore the original materials here: the case report’s DOI page (acpjournals.org), independent coverage by 404 Media and Ars Technica (404media.co; arstechnica.com), WHO sodium guidance and AI governance advisories (who.int; who.int), Thailand’s sodium-reduction roadmap and data (pmc.ncbi.nlm.nih.gov), and toxicology literature on bromism and lab pitfalls (pmc.ncbi.nlm.nih.gov; pmc.ncbi.nlm.nih.gov). Consumers should note that brominated vegetable oil has also now been revoked from US foods, reflecting a broader regulatory posture toward brominated compounds (fda.gov).
Thailand’s salt story is, at heart, about balancing tradition and health: keeping the delicious complexity of tom yum and som tam while protecting hearts and kidneys. The bromide case, bizarre as it sounds, is a teachable moment. Keep the goal, lose the gimmicks.