Written by 2:34 am AI, Discussions

– Prohibiting a Friend from Conducting an AI STI Test: A Concern among Friends

You can use image-based AIs for plenty of things, but probably not for this.

Picture this scenario: Your meeting has concluded, and you and your partner are preparing to retire for the night. You anticipate a conversation about STI statuses and the importance of protection, as is customary in a safe and responsible encounter. Now, imagine your reaction if your partner were to request capturing an image of your genitalia to upload to an unfamiliar website. This is the vision of intimacy that Calmara, a novel service introduced by the “men’s health” company HeHealth, envisions for the future.

Banner image from the top of the HeHealth website describing Calmara.ai as as

HeHealth Website

As per its official statement, users are encouraged to photograph their partner’s male anatomy for analysis using a sophisticated machine learning algorithm designed to detect sexually transmitted infections. Despite the website’s banner labeling the app as “Your intimate bestie for unprotected sex,” it paradoxically advocates for safe sex practices. However, there are notable shortcomings in the proposal, including its focus solely on male genitalia and its limitation to visually detectable infections.

Even if the intended use is accurate, one may still harbor doubts about the reliability of the results upon closer inspection of the data. Calmara asserts that its image analysis boasts up to 90% accuracy, citing that its AI has been tested by over 40,000 individuals. The stated accuracy levels range from 65% to 96% across different conditions, aligning with the information provided in the press release claiming a 94.4% accuracy rate (as referenced in a recent NSFW preprint paper). Our attempts to seek clarification from the company regarding this significant disparity remain unanswered.

Image of the Calmara website showing its claim of


While models can interpret visual data, akin to how they analyze cellular images for drug discovery, visual information may not be as reliable for STI testing due to various factors. Many infections lack visible symptoms, and carriers can remain asymptomatic for extended periods post-infection. The company acknowledges this limitation in its FAQ, acknowledging that the tool serves as a “first line of defense rather than a comprehensive solution.” Moreover, external factors such as lighting, specific health conditions being assessed, and diverse skin tones can influence the accuracy of the results. The unpublished paper, marred by errors, discloses that 40% of its training dataset comprises “augmented” images, where disease patterns are superimposed on images of healthy anatomy, raising additional concerns.

Image from the Calmara FAQ illustrating the variability of its tests. Image from the Calmara FAQ highlighting the variability of its tests.


The official stance on the Calmara website emphasizes that the devices are meant to “encourage and bolster public health initiatives and a healthy lifestyle” and are not intended for diagnosing, treating, managing, or preventing any diseases or conditions. However, if the tool were truly a general wellness resource, it would likely not adopt the moniker “Your intimate bestie for unprotected sex.”

The ethical implications surrounding consent and, as noted by writer Ella Dawson on Bluesky, age verification, seem to have been overlooked. The company provides no assurances regarding the handling of the collected data, despite claiming to safeguard it within a “digital fortress.” Nonetheless, this has not deterred the company from hinting at potential integration with dating apps in the future.

In essence, the myriad of potential red flags for misuse and ethical dilemmas associated with this tool render it unsuitable for use, as it fosters a false sense of security among users.

Visited 4 times, 1 visit(s) today
Tags: , Last modified: March 21, 2024
Close Search Window