Could AI make alcohol age-verification easier?
In the UK, the tech is already being used to identify people who may be under 25 (and should be challenged to show ID to prove they are old enough to buy alcohol) and identify those that are clearly older than 25 (so that 60-year-olds don't need to show ID every time).
Where wrinkles can be a good thing...
Innovative Technology (ITL) has developed biometric age verification technology, which is being used in real life to control access to age-restricted products and venues.
The AI-powered age verification tech works by analysing the customer’s facial features to establish their age, offering retailers an added layer of security when selling products such as alcohol.
“Our algorithms look at a person’s facial features and estimate an age based on what the algorithms have learnt,” explains Dr Andrew O’Brien, Biometrics Product Manager, Innovative Technology.
“The technology operates completely anonymously, it does not store any data and does not require any previous knowledge or personal details about any person. Local processing is performed on the device and it can be used completely offline to further ensure privacy. This is all achieved in under one second!”
In the UK, the main purpose of the tech is to identify consumers that are under 25, in line with the country's Challenge25 policy (staff are instructed to ID anyone who looks U25, even though the country's legal drinking age is 18 years old).
What’s important to understand, says O’Brien, is that biometric verification in the context of alcohol sales is not the same as biometric verification in passports or border control.
“Border checks are completely different and require personal data to be stored for identity verification,” he said. “We are simply analyzing facial features to perform age checks with no data stored.”
On average, the technology overestimates by just 0.39 years, with a mean absolute error of 0.94 years. That means it’s suitably accurate for being used in Challenge25 areas.
And the tech, says O’Brien, is affordable and easy to use.
“MyCheckr can be used straight out of the box and comes complete with a screen and camera for manned checkouts to assist staff with the age check process, whereas ICU Lite can be easily integrated with POS systems for self-service checkouts to fully automate age checks. Available at a one-time purchase cost with no additional fees per age check, and no internet connection required, we offer a high tech and valuable tool that is both affordable and simple to use.”
No more 'assistance required' at self-service tills
Ed Heaver heads up UK age compliance and verification service Serve Legal, which runs third-party audits for local and national retailers to assess their compliance with age restrictions. That’s done via 4,000 young auditors, who (while themselves are 18+) should be challenged by retailers when buying alcohol.
AI and biometric verification could offer ‘vital support’ for staff and a frictionless experience for customers making in-store purchases, said Heaver. The most obvious example here of the tech’s potential is with self-service tills.
What’s important is that such verification will not eliminate the need for ID checks but help sort out who needs to show ID.
“Developments in recent technology are looking to bring automatic age estimation technology into supermarket self-scan machines to verify if someone needs to be challenged for ID for an age verified product," he said.
“Currently, at self-scan machines, all customers are stopped until a staff member is available to verify their age.
“Implementing biometric verification will not eliminate the need for ID checks but will identify customers who appear under 25 and need to be challenged. This allows other customers to complete their purchases without interruption.
"Additionally, this technology will offer reassurance and support to staff who must challenge customers, particularly those who may be reluctant to show their ID.”
Cost, effectiveness and data… how to overcome the challenges for AI?
It’s easy to guess where concerns about AI are going to come from.
There's the cost of the systems, and the question of whether it is actually going to be effective (self-service tools, after all, still have their detractors). Add into that, a host of complicated privacy concerns and regulations.
O’Brien agrees these are key concerns among their clients… but adds these are also key misconceptions.
Addressing the question of effectiveness, the company has just had its tech re-certified by independent body ACCS (Age Check Certification Scheme): which tests both online and offline systems for age and identity checks (and what’s more, this year’s certification showed the tech has improved since the last certification test two years ago).
“Our biometric devices have been rigorously tested by Age Check Certification Scheme (ACCS) and this independent certification concludes that our technology is suitably accurate and fit for deployment in a Challenge 25 policy area," he said.
"It is proven to be more accurate than human age estimations too, even when implementing Challenge25, staff could vastly misinterpret someone's age without knowing it.”
The tech creates a clear-cut decision on when ID needs to be requested: easing the responsibility from staff, as well as creating clear rules for consumers.
“Many customers using our technology have reported less confrontation towards staff from customers attempting to buy age-restricted goods, as well as giving a higher level of assurance for servers when asking for ID.
"Additionally, when fully integrated at the point-of-sale, customers over the age restriction threshold no longer have to wait for age approvals, speeding up transactions and ensuring increased customer throughput.”
In some cases, the tech does its job without even being activated.
"What we also hear from our customers is a noticeable reduction in underage people even attempting to purchase age restricted items, as they see our technology as a clear and obvious deterrent," said O'Brien.
How big a challenge is cost and privacy?
High-level tech equals high-level costs? O'Brien disagrees this is the case.
"Using our devices, customers only need to pay for the hardware," he said. "No matter how many age checks are performed, the cost remains the same, making this an affordable solution with a quick return on investment, especially when compared to a subscription service."
Privacy infringements and - in Europe - GDPR compliance is another key concern that O'Brien believes can be addressed.
"We never store any data; all processing is performed locally on the device so we never send data over any network; and finally, any data that is processed is automatically deleted afterwards.
"The data is used to estimate age, cannot be used to uniquely identify a specific natural individual and so complies with GDPR regulations.
"We have been advised by legal experts that we do not infringe on Biometric Information Privacy Act (BIPA), UK GDPR or EU GDPR, as such a Data Protection Impact Assessment is not required."
Address challenges now
Heaver is also trying to think ahead - and think like his (underage) foe. He says the key to AI will be to create systems that work and inspire faith from their users from the word go.
“Whilst the development of this technology has the potential to be ground-breaking, the need to address challenges early on is imperative,” he said.
“Liveness detection is a vital component, ensuring that the biometric system interacts with a live human rather than a photo, prosthetic, digitally injected video, or another artificial artifact. This prevents potential abuses of the system, such as using someone else's digital ID or injecting a deepfake video to deceive biometric analysis.
“Additionally, deficiencies in the testing landscape must be addressed to ensure systems are developed according to ethical best practices. Testing should involve sufficient sample sizes with appropriate diversity to identify any significant discrepancies between demographic groups. This fairness metric is essential for deployers, such as banks and supermarkets, to ensure compliance with equality legislation and to avoid deploying biased liveness detection systems.”
What about age checks online?
Age-restricted products purchased online often rely on a simple checkbox for age verification, such as, "Are you over 18+?". It's easy for an under-age consumer to say 'yes'.
In this scenario, the responsibility tends to shift to couriers or delivery providers to complete the sale with compliant age verification protocols at the doorstep.
"This scenario presents several challenges," said Heaver.
"The online customer may be of legal age, but the recipient might not be, or vice versa. Ensuring both the online customer and recipient are of legal age is imperative to comply with age verification laws.
"Additionally, while supermarket staff are supported by posters, automatic till prompts, and colleagues when challenging a customer's age, delivery drivers often face these situations alone. Should they be unable, or refuse, to show ID – the delivery driver faces an awkward and potentially difficult situation.
"It is vital for retailers and employers to support delivery drivers in enforcing these checks to ensure compliance with age verification regulations."
What’s most important, says Heaver, is for retailers to carefully evaluate how AI might help in their own situation.
“With government agencies looking for evidence of proactive due diligence through independent testing, businesses and retailers should take this opportunity to invest in well-tested, responsible models and expert compliance testing.
“Implementing this technology requires retailers to balance accuracy and user experience, ensuring the benefits outweigh potential drawbacks.
“We believe it is crucial for retailers and businesses to thoroughly evaluate their investment in these technologies before implementation.”
