AI has undeniably transformed the world, but as Maya Shabi and Anna Pogreb from EverC point out, its effectiveness depends on the quality of the data you feed it.
In a short time, Generative AI has graduated from bleeding-edge tech to a go-to tool. This rapid rise has been applied to online crime, making it even harder to identify and disrupt. Bad actors and criminal networks are using AI-enabled tactics to exploit the increasingly interconnected nature of our digital world.
But it’s not all bad news. With these challenges comes an unprecedented opportunity to leverage AI technology for good. At EverC, we design solutions and services to detect and disrupt illicit activity online. Our teams and technology use AI to fight AI-fuelled fraud.
There is no doubt that AI technology has changed the world – but to use it effectively, we must keep one thing in mind:
The output is only as good as the data you feed it.
At EverC, we scan and classify millions of different websites and products, which gives us access to a vast amount of data. We use this data to train our systems to be smarter and faster than the fraudsters.
Our technology and teams incorporate machine learning (ML) and large language models (LLMs) into our solutions, and in both cases, data is key to using them effectively.
Regular ML models can be trained on data, but how you input the data is the key. To properly train your algorithms, you must provide labelled data. For example, you tell it what you want to catch and what a false positive looks like, and perform several rounds of training. Our team has built expertise over time from working in this industry, surveying clients, and watching trends, so we can tag the data very accurately. Better data means better results. But that is only one piece of the puzzle.
We also use LLMs. LLMs can take in a lot of data. The more the merrier! But instead of providing this structure through labelling or tagging, you feed it lots of data and structure your queries in a way that will return more precise results. Here, AI can discover patterns and outliers. With the right prompt, it will return the pattern you need.
But whereas you train a regular ML model by structuring the data, with LLMs, you are training yourself to present the right questions in the right way, which in turn trains the model to find the right pattern and deliver the results you need.
Our experts work to create prompts that will return the most precise findings, often using multiple models and queries, then testing to continuously improve results. This concatenation of models is fed with data based on model type, to gather the results needed. Properly orchestrated algorithms and LLMs provide you with an array of possibilities for results – much like sitting in a well-stocked kitchen with your grandmother’s cookbook. The ingredients are the same, but the recipes make all the difference.
Payment providers and platforms are experiencing explosive growth. As growth increases, so does the risk involved. Managing risk is a particularly daunting task, given the highly regulated nature of payments and the rapidly changing ecommerce landscape. Tech-forward solutions are necessary for safe growth at scale, and this is where AI models can truly shine, providing the benefit of continuous learning.
We train our models and learn from patterns in our ecosystem to ensure our partners maintain compliance with anti-money laundering and counter-terrorist financing regulations. We detect high-risk or potentially violative activity so that bad merchants or sellers can be eliminated from these platforms. This helps to make the Internet a more transparent, trustworthy place to conduct business.
The world is full of illicit actors using AI for nefarious purposes. In the payments industry, we have an opportunity to use it for good. And by doing so, we create and train the technology that can make ecommerce safer for all.
About Maya Shabi
Drawing on her expertise in counterterrorism, geopolitical risk assessment, and mitigation in the intelligence sector, Maya Shabi tracks emerging online threats to leverage technology to beat bad actors at their own game. Maya is deeply fascinated by the evolution of society's engagement with technology and drives thought leadership and product development to disrupt some of the world’s worst actors. She is passionate about standing up for vulnerable and marginalised communities that are impacted by threats across the industry landscape.
About Anna Pogreb (CFCS)
With over 15 years of experience in payments, compliance, and technology, Anna Pogreb is fervently dedicated to fighting fraud with integrity and intellect, as well as restoring dignity to those most impacted by it on a global scale. Her ability to analyse risk is built on a lifetime of diverse experiences that have given her unique knowledge of cultures, languages, and tactics used by fraudsters. Anna is a certified financial crime specialist and an active member of the Israeli Fraud Fighters community group.
About EverC
EverC is focused on powering growth for the ecommerce ecosystem. Our automated AI-driven, cross-channel risk management solution rapidly detects high-risk merchants, transaction laundering, and illicit products, and provides ongoing monitoring to uncover evolving risks. Our team comprises domain experts in risk intelligence, open-source, deep, and dark web, and online fraud detection.
Every day we send out a free e-mail with the most important headlines of the last 24 hours.
Subscribe now