The Press Council considered a complaint from Yoti Ltd concerning an article published by Crikey on 14 June 2024, headed “I tricked a selfie AI age-verification tool into letting a child ‘buy’ a knife” (Online).
The article’s subheading said that “Governments are looking at using AI to estimate people’s age using a selfie. I fooled it using a stock image and an old aging filter.” The article reported that “A tool used to estimate age using facial analysis that is being promoted as a way to stop underage children from accessing social media or online pornography can be fooled using an aging filter on a popular photo-editing app.” The article went on to report that “One of Yoti’s methods for age estimation, which Opposition Leader Peter Dutton spruiked just yesterday — is ‘facial analysis’, which can calculate someone’s age using a selfie.” It reported that “Critics argue that age estimation through facial scans is flawed and vulnerable to being tricked. Yoti offers some measures to stop people from fooling its system, but it doesn’t change the underlying technology’s reliance on easily falsifiable information.” The article went on to say, “I know because I tricked Yoti’s age estimation into letting me ‘buy’ a fixed-blade knife using the photograph of a 10- year-old that I had put through an aging filter on a photo-editing application.”
The complainant said that the article is entirely misrepresentative noting in particular the headline and the accompanying social media posts. The complainant said it had conversed with the publication and explained in detail how the “demo” the publication had used for the basis of the article, was a limited demo version of the age assurance technology and how it did not include Yoti’s liveness technology which would detect a manipulated image. The complainant said it explained how actual customers would always use liveness detection in a working model to ensure that manipulated images could not fool the system. The complainant said that the demo was designed to allow people to explore the technology in full to see what you could and could not do with this version. The complainant said that despite its explanation, the publication proceeded to report that it had “tricked the technology” when it had not and used an inflammatory and misleading headline. The complainant said the article damages Yoti’s reputation which is highly regarded in the age assurance sector and endorsed by numerous government and industry bodies.
In response, the publication said the article made clear from the beginning that the article was about the underlying technology utilised in Yoti’s demo. It said that it updated the article and included the word “demo” in the headline and the article and added a clarification out of an abundance of caution after it was contacted by Yoti. The publication said that while Yoti disputes that the underlying technology used in the demonstration was “tricked” or “fooled”, it is obvious that this article demonstrates a fundamental flaw in the age verification technology that allows it to be tricked or fooled. Its system relies on visual cues that are able to be manipulated irrespective of someone’s age. That is why Yoti has separate liveliness detection technologies that it sells. The technology has a vulnerability which the publication demonstrated, even if it’s one that can be mitigated.
Conclusion
The Council’s Standards of Practice require publications to take reasonable steps to ensure that factual material in news material is accurate and not misleading (General Principle 1), and is presented with reasonable fairness and balance, and that writers’ expressions of opinions are not based on significantly inaccurate factual material or omission of key facts (General Principle 3). They also require publications to take reasonable steps to provide a correction or other adequate remedial action if published material is significantly inaccurate or misleading (General Principle 2) and provide an opportunity for a response to be published by a person adversely referred to (General Principle 4).
The Council accepts that the publication has accurately reported how it used a used an image of a child to buy a knife when it used an online demo tool provided by Yoti on its website. The Council also accepts that age verification technology may be manipulated. However, on the information before it, the Council notes that the online demo used by the publication for the premise of the article, was limited in its purpose and designed only to demonstrate to potential purchasers, how age verification technology works in an online environment. The Council accepts that the demo was not designed to detect, as in this instance, the manipulation of an image. The Council also notes that it was clearly stated on Yoti’s website, that its ‘Anti spoofing software’ which is designed to detect the manipulation of an image, was not enabled on the demo. In this context, the Council considers it is misleading to state that technology used by Yoti to estimate age using facial analysis was “tricked” or “fooled”. The Council also considers that in context of an article that questions the fallibility of facial analysis technology, the article unfairly suggests that the technology used by Yoti is flawed. Accordingly, the Council concludes General Principles 1 and 3 were breached in these respects.
As to corrective or remedial action, the Council considers the article is significantly misleading. While the Council recognises the remedial steps taken by the publication, the Council does not consider that this constituted adequate remedial action for the implication that the complainant’s technology can be “tricked” or “fooled”. The Council notes that even with the inclusion of the word “demo” in the headline and article and the additional clarification in the article, the article’s premise that the technology used by Yoti can be fooled remains misleading and unfair. The Council again notes that the demo used in support of the article’s premise that age estimation software is flawed, was designed for a different purpose. The Council notes that on the information before it, the publication has not offered the complainant with an opportunity for a subsequent reply. Accordingly, the Council concludes that General Principles 2 and 4 were breached.
