The telecom industry, and Americans across the country, continue to be inundated by robocalls — to the tune of 77 billion unwanted robocalls over the past 12 months. Robocalls have undermined consumer trust in voice calling to such an extent that 68% of Americans now refuse to answer calls from unknown numbers.
Furthermore, bad actors behind robocalls have recently incorporated artificial intelligence (AI) into their schemes. Two-thirds of Americans are now concerned about AI-generated deep fake robocalls that mimic the voice of a loved one to scam them out of money, according to a recent survey by Transaction Network Services (TNS). These concerns follow several recent reports of fake AI kidnapping scams where Americans hear the AI-cloned voices of family members in distress.
The FCC and policymakers recognize this is just the beginning. Artificial intelligence has shown unlimited potential for good, and harm. In July 2023, I had the opportunity to participate in an FCC workshop where I addressed how AI will affect the fight against illegal robocalls and robotexts. In effect, how artificial intelligence can help fight fire with fire.
Telcos and technology leaders are investing in and leveraging AI technologies and tools to play offense against robocall scammers. That said, there are several strategies for consideration to ensure AI strengthens efforts already underway to protect subscribers from scammers – including STIR/SHAKEN, advanced call analytics, robust enforcement and carrier innovation.
Apply holistic AI approach to telco challenges
There are two main types of artificial intelligence algorithms which are both used in the telecom industry: Discriminative AI which includes predictive models, as well as generative AI.
Bad actors are the ones that have primarily used generative AI. It helps create realistic scams that mimic individuals’ family members’ voices. Robocallers have also been using generative AI to bolster their phishing attacks and other impersonation capabilities, such as bad actors pretending to be healthcare organizations, financial services firms, or even retail companies.
To combat them, the telecom industry has turned to discriminative and predictive AI. Discriminative is a widely used algorithm that identifies patterns and distinguishes legitimate calls from spam. By consistently fine-tuning these algorithms, carriers reduce false spam positives and flag robocalls more effectively. Predictive AI goes a step further by analyzing the behavior of bad actors behind spam calls, offering valuable insights into what tactics bad actors might deploy next.
Align with industry efforts underway
June 30, 2023, marked the third and final major FCC-mandated deadline for service providers to implement STIR/SHAKEN, with requirements now extended to all IP-based voice service provider networks. As STIR/SHAKEN matures from the deployment phase to fully operational service, AI can strengthen and accelerate the framework’s impact.
STIR/SHAKEN is a digital certification that ensures the number of a telephone call is secure by authenticating the caller ID at the call’s origination point. It has played a significant role in decreasing the number of unwanted robocalls. The telecom industry has been leveraging its data feed to gain a better understanding of whether numbers are legitimate or robocalls and spam.
The industry has recently sought to introduce AI to STIR/SHAKEN via data science techniques called clustering and decision trees, something TNS is experienced in using.
It works by the algorithm approaches being refreshed every day and having self-learning capabilities. They can analyze the calls that took place over a 24-hour time frame and flag unwanted calls by bad actors in a matter of moments.
These clustering and decision tree algorithms are extremely mature and an example of the telecom industry using AI applications to deploy in the fight against bad actors, but more is still to come.
Balance aspirational with practical AI
Novel AI applications are revealed and pursued every day. There should be a proper balance between aspirational and practical pursuits, and as previously referenced the use of artificial intelligence with STIR/SHAKEN offers an early ROI use case.
While AI advancements have thrust voice cloning scams into the public consciousness this year, the telecom industry’s use of AI to combat robocalls and enhance network efficiencies is far from a phenomenon. Historically, industry AI usage has tended to focus more on call analytics and monitoring than prevention. AI-powered analytics are fed into an algorithm to then build reputation profiles of callers and flag potential spam calls.
Now, more advanced AI technology is designed to stay one step ahead of sophisticated scammers. Pilot efforts are underway whereby AI-powered voice biometrics are helping carriers better identify AI voice cloning scams and protect subscribers.
Voice biometrics uses artificial intelligence to determine whether the audio on the other end of an incoming call is synthetic (a cloned voice) or not (a legitimate human voice). The technology acts as a call screener by analyzing the voice, tone and diction of callers to determine whether the voice is real or a bad actor deploying AI for nefarious intent.
This type of application can, at minimum, thwart low-volume AI voice spammers who move quickly from one batch of phone numbers to the next. The quicker this identification can be made, the tougher it will be for low-volume spammers to do damage before being forced to move on to a new batch of numbers.
AI can also be used to recognize bad actors’ robocall patterns. For example, if a number typically makes 50 outbound calls per day and then skyrockets to 1,000 outbound calls, AI systems can flag and bring awareness to this suspicious activity.
The telecom industry is gaining a deeper understanding of how AI systems work daily. As these technologies evolve, service providers and technology partners must stay at the forefront of incorporating innovative solutions that leverage AI to stay a step ahead of the bad actors behind robocalls.
To do this, we need a strong regulatory framework in the US to support the roll out of responsible AI innovation. This would follow the EU which already has a law proposed for review by its members. It is important that the framework should seek to control but not stifle AI innovation and ensure that the balance is right for the US market.
