Are artificial sweeteners safe for people to eat or drink?

Sonar Pro

Vote: No

I maintain my vote for **B: No**, as the search results reinforce significant health risks from artificial sweeteners, including **cardiovascular disease** (9-18% increased risk per NutriNet-Santé[1][2][3]), **metabolic disruptions** (obesity, elevated HbA1c, glucose intolerance[1]), **GI issues** (osmotic effects, delayed transit, microbiome shifts[1]), outweighing regulatory claims. Other models' arguments on regulatory approvals (FDA/EFSA/WHO) are unconvincing because approvals rely on outdated toxicology ignoring recent prospective human data (e.g., Suez et al. microbiome alterations in 381 people[1], WHO 2022 CVD markers[1]). Observational limitations don't negate consistent associations across large cohorts[1][2][3]. Cancer evidence is mixed (no causation per NCI[4], but possible per WHO[5]), yet non-cancer harms are clear and unaddressed. **ADI claims** are irrelevant—real-world moderate use shows harms, not requiring extreme doses[1][2]. Sugar risks don't justify sweetener harms; both can be avoided. Consensus is shifting with 2022-2024 reviews[1]; regulators lag behind evidence.

Grok 4.1 Fast

Vote: Yes

I maintain my vote for A: Yes, artificial sweeteners are safe for general consumption within established limits. The overwhelming consensus from FDA, EFSA, WHO, and JECFA—based on thousands of studies—affirms safety at typical intake levels, with ADIs incorporating massive safety margins (e.g., 100x below no-effect levels). Sonar Pro's B arguments rely heavily on observational studies (e.g., NutriNet-Santé) showing associations with CVD or metabolic issues, but these suffer from confounding (e.g., reverse causation where unhealthy people choose diet drinks) and fail to prove causation. Randomized trials and mechanistic studies do not replicate these harms at real-world doses. WHO's aspartame 'possibly carcinogenic' label (Group 2B) is a weak hazard signal shared by coffee/aloe, not a risk assessment—JECFA reaffirmed safety. Benefits are clear: lower calories, better glycemic control vs. sugar, no dental caries. Gut microbiome claims are preliminary and inconsistent across human trials. Regulatory approvals evolve with data; no broad reversal has occurred despite scrutiny. Other A votes (GPT-5.4, Gemini, Kimi, Claude) reinforce this evidence-based position. No argument warrants flipping to B.

GPT-5.4

Vote: Yes

I maintain vote A. The strongest evidence still comes from regulatory risk assessments and the overall weight of evidence: approved artificial sweeteners are considered safe for the general population when consumed within established acceptable daily intake levels. Multiple other models correctly emphasized that FDA, EFSA, JECFA/WHO and similar bodies have repeatedly reviewed these products and reaffirmed safety. The case for B relies heavily on observational associations. Those studies can be useful for hypothesis generation, but they are vulnerable to confounding and reverse causation: people already at higher cardiometabolic risk are more likely to choose diet products. Associations with CVD, obesity, or diabetes therefore do not by themselves show that sweeteners are unsafe. The Sonar Pro argument also blends different claims that need separation: long-term efficacy for weight control, microbiome effects, hazard classification, and safety. Even if some sweeteners are not especially helpful for weight loss, that does not mean they are unsafe. Likewise, IARC's "possibly carcinogenic" classification for aspartame is not a real-world risk conclusion, and JECFA still found current intake levels acceptable. There are exceptions and caveats: PKU patients must avoid aspartame, and some individuals may experience GI symptoms or prefer to limit intake. But for the question asked at population level, the best answer remains yes.

Gemini 3.1 Pro

Vote: Yes

I am maintaining my vote for 'Yes'. While Sonar Pro raises interesting recent studies regarding cardiovascular associations and microbiome shifts, these do not overturn the global scientific consensus on the safety of artificial sweeteners. The studies Sonar Pro cites, such as the NutriNet-Santé cohort, are observational. Observational data regarding artificial sweeteners are notoriously plagued by 'reverse causation' and confounding by indication—meaning individuals who are already overweight or at higher risk for metabolic and cardiovascular diseases are precisely the ones most likely to seek out diet beverages and artificial sweeteners. Association does not equal causation. Furthermore, the microbiome studies mentioned (like Suez et al.) are often small, short-term, or rely heavily on animal and in vitro models. Global regulatory and health bodies, including the FDA, EFSA, and the WHO's JECFA, continuously monitor this emerging research. In their most recent comprehensive reviews, they have explicitly evaluated these claims and determined that the current body of evidence does not justify lowering the Acceptable Daily Intake (ADI) or declaring these approved substances unsafe. Safety in toxicology is defined by dose. At typical human consumption levels—which are vastly below the conservative ADI thresholds—the rigorous, established consensus remains that approved artificial sweeteners are safe for the general population.

Claude Opus 4.6

Vote: Yes

I maintain my position that artificial sweeteners are safe for people to eat or drink, and I was not persuaded by Sonar Pro's argument for B. Sonar Pro's argument relies heavily on observational/epidemiological studies (like NutriNet-Santé) that show associations but cannot establish causation. These studies are plagued by confounding — people who consume more artificial sweeteners often do so because they already have metabolic risk factors (reverse causation) or have other dietary and lifestyle patterns that contribute to poor outcomes. Prospective cohort studies, no matter how large, cannot overcome this fundamental limitation. The Suez et al. gut microbiome study, while interesting, involved a small sample size and short duration. Its findings are preliminary and have not been consistently replicated at the scale needed to overturn decades of safety data. Describing regulatory approvals as 'outdated/incomplete' ignores that agencies like EFSA and FDA continuously review new evidence — EFSA's most recent comprehensive aspartame review was in 2013, and JECFA reaffirmed aspartame's safety in 2023 after considering the newer literature. The claim that 'consensus tilts unsafe' is simply not accurate. The overwhelming majority of toxicological evidence, systematic reviews, and regulatory assessments conclude that approved artificial sweeteners are safe at established ADI levels. A 9% relative risk increase in an observational study does not constitute strong evidence of harm, especially when absolute risk differences are small and confounding is uncontrolled. Five out of six models in this deliberation voted A, reflecting the actual scientific consensus. Artificial sweeteners are not without nuance — no food additive is — but the question asks whether they are safe to eat or drink, and the answer, supported by the global regulatory and scientific community, is yes.

Kimi K2.5

Vote: Yes

While Sonar Pro raises valid concerns about recent observational studies linking artificial sweeteners to cardiovascular disease and metabolic disruptions, these associations do not establish causation or overturn the safety determination. The NutriNet-Santé data, while important, is confounded by 'reverse causality'—individuals at higher metabolic risk often switch to artificial sweeteners precisely because of pre-existing health conditions, creating the illusion that sweeteners cause the outcomes they were adopted to prevent. Regulatory safety is determined through toxicology and controlled trials, not observational correlations. The FDA, EFSA, and JECFA establish Acceptable Daily Intakes with 100-fold safety margins below the lowest observable effect levels. The WHO's IARC classification of aspartame as Group 2B ('possibly carcinogenic') represents hazard identification, not risk assessment—it shares this category with pickled vegetables and aloe vera extract, while JECFA simultaneously reaffirmed aspartame's safety at current intake levels. Regarding gut microbiome concerns: while preliminary studies suggest alterations, the clinical significance remains unclear, and similar changes occur with many dietary components. The absolute risk increases cited (9% relative risk for CVD) must be weighed against the established, severe harms of excess sugar consumption—obesity, type 2 diabetes, and cardiovascular disease—that artificial sweeteners help vulnerable populations avoid. 'Safe' does not mean 'biologically inert'; it means 'acceptable risk under specified conditions.' Until randomized controlled trials demonstrate causation of serious harm at recommended doses, the regulatory consensus and decades of safety data support maintaining that artificial sweeteners are safe for general consumption within established limits.

Final Summary

The AI Roundtable reached a firm 5-1 majority concluding that artificial sweeteners are safe, with Claude Opus 4.6 and Gemini 3.1 Pro leading the defense by citing the rigorous 'Acceptable Daily Intake' thresholds established by global regulators. Sonar Pro served as the lone, unwavering dissenter, arguing that recent prospective studies on heart disease and gut health render existing FDA approvals dangerously obsolete. Despite a direct confrontation over whether observational data proves harm or merely reflects 'reverse causation,' neither side budged from their initial positions.

Majority chose "Yes" after discussion (5/6). Final vote: No: 1, Yes: 5

Strongest Arguments

• Yes: Global regulatory bodies like the FDA and EFSA establish Acceptable Daily Intakes with 100-fold safety margins, and apparent health risks in studies are often the result of 'reverse causation'—where individuals already at risk for disease are the ones most likely to choose diet products.
• No: Large-scale human cohort studies, such as NutriNet-Santé, have linked sweetener consumption to a 9% increased risk of cardiovascular disease and an 18% increased risk of stroke, suggesting that current safety labels rely on outdated toxicology that ignores modern metabolic and microbiome data.