False positives and false negatives

How best to leverage adverse media searches in the battle against financial crime

Over the past ten years, the level of regulatory scrutiny of financial services firms’ financial-crimes compliance has expanded significantly, with regulators around the globe taking scores of enforcement actions and levying $36 billion in fines.

Many financial institutions have scrambled to implement effective remediation efforts and as a result, staffing levels in financial crime investigation and analysis have significantly increased. Innovative techniques and processes to combat financial crime have also evolved; in particular, adverse media searches have become an even more critical screening practice.

However, the problem of accumulating too much information has plagued the early phase of technology and skilled analysis rollouts. In an effort to capture every possible impropriety, banks are collecting too much data and now must consider the problem of managing both false positives and false negatives; they often need help focusing on what is most relevant and valuable to their searches.

In this white paper, we will look at this problem, why it has become so important to address it, and how new technology and skilled researchers can help solve it. 

The growing importance of adverse media searches

Adverse media searching and screening techniques — sometimes called negative news search — reflect a desire on the part of financial institutions to use a systematic and efficient approach to heading off potential alliances with criminal organizations and illicit actors, including those who intend to engage in money laundering or other financial crimes. Adverse media search is a part of due diligence that U.S. regulators often look for. Financial institutions must decide how to implement it with skill and rigor while considering their clients’ — and their own — risk appetite and utilize the right tools that can sift through a high volume of available data. 

Although some financial institutions had already been using an adverse media approach to alert them to potential risks from clients — as the availability of digital search tools has increased — an acceleration of this activity in recent years only occurred after a new compliance requirement on customer due diligence went into effect in 2018, implemented by the U.S. Treasury’s Financial Crimes Enforcement Network (FinCEN). 

While financial institutions’ obligation to know with whom they were doing business existed previously, the new rule underscores their affirmative obligation to understand and know their clients and report on possible attempts at financial crimes on the part of clients. Specific obligations within this category of due diligence include rules around anti-money laundering (AML), know your customer (KYC), and countering the financing of terrorism standards.

In addition, the European Union’s 6th AML Directive (6AMLD), which took effect in 2021, requires clients from high-risk regions to undergo adverse media searches when seeking banking services in Europe. The EU directive specifies 22 predicate offenses commonly followed by attempts to launder money as a guide for financial institutions’ searches.

Financial institutions across the globe now seek ways to comply and assess risk in an effective and efficient manner. The intergovernmental Financial Action Task Force (FATF) has set guidelines for the kinds of offenses that financial institutions should prioritize when implementing or improving adverse media searches, including human trafficking, organized crime, terrorism, trafficking in stolen goods, and drug trafficking. These requirements are unlikely to ease up anytime soon — for instance, FATF has now issued newer guidance about how to monitor cryptocurrency and other virtual assets.

Increased adoption of adverse media searches

Financial institutions are not the only businesses that can benefit from more streamlined tools to conduct adverse media research. Suppliers and vendors increasingly look to such techniques to avoid illegal or unethical actors in their supply chain, including those committing labor abuses or those committing narcotics trafficking or terrorism financing. Any firm or organization concerned about future reputational risks may benefit from using these tools to head off problems with investors, management, and other stakeholders. Indeed, firms can find more of what they need faster by using more advanced search tools and analytic methods to look for all criminal activity, previous conflicts, or misconduct allegations. Unfortunately, these more focused searches are normally done ad hoc and may be costly and time consuming to complete.

However, the financial sector has a specified set of regulatory obligations with which to comply and a much wider area to monitor and protect — its risk surface — as nearly all firms and organizations seek banking services. Not surprisingly, some customers and potential customers require more intense screening than others, with the need for enhanced due diligence practices in these cases. 

Adverse media searches — whether manually driven through open-source search engines or aided by systems based on artificial intelligence (AI) — carry the inherent problem of raising both false positives and false negatives, especially if using older or legacy AI-based tools. False positives are delivered search engine matches that, once scrutinized by analysts, are irrelevant to the query. As it stands now, most searches return a large percentage of false positives. On the other hand, false negatives are potentially relevant bits of information not caught by an institution’s current search methods. 

The problem of producing a considerable percentage of false positives in adverse media searches has many ripple effects. For example, an increasing number of analysts must be brought on to comb through search results and databases, evaluating them for relevance to the query. Further, a query may relate to seeking information about an individual, a corporation — with beneficial owners — or an organization with many involved principals, which also may increase the chances of more false positives getting delivered. 

Add to this that financial institutions often need to understand each client’s overall network of vendors, suppliers, donors, or customers, and you may quickly find yourself in a world swimming in data and information — in which the problem is not too little data but too much. Finally, given the multitude of false positives, analysts will become conditioned on them and treat all potential adverse media hits as false positives, possibly missing the one true positive and increasing the risks to the institution. 

Until the recent availability of AI tools, banks and investment firms had to rely on the most popular search engines and separate searches in various public databases. To avoid the resulting unmanageable tidal wave of information, users would have to manually prepare intricate Boolean searches, combining keywords and operators like “and” or “or” to limit or widen the results. Again, however, this is time consuming and still generates a very high level of false positives or ultimately irrelevant results. To the extent that such manually refined searches are still dependent on fuzzy matches — a problem that we are all familiar with from our own experience with search engines — information overload will reign. 

In coping with too much information and too many search results, skilled analysts often need to shrink the search parameters to make a careful inspection of results plausible. The alternative is simply not being able to onboard or maintain timely services for customers. 

Technology drives greater precision in due diligence

AI-driven search tools can reduce the volume of irrelevant hits in a search because they can interpret the context and use of keywords or phrases more readily than search engines. These machine-learning or large-language model tools can differentiate, for instance, between a genuinely concerning match about a client’s activities on the one hand and an article about money laundering or other malfeasance that mentions a similar name or simply appears on the same news brief page as an unrelated one about the client. The Association of Certified Anti-Money Laundering Specialists (ACAMS), an organization that certifies and trains professionals to detect money laundering, places the false-positive problem in the banking industry as a crucial current obstacle to conducting quality screening of clients. 

Regarding large client companies, these AI-driven tools can screen out hundreds of news articles concerning negative publicity that may be unrelated to financial crimes risk. AI can also help wade through and exclude volumes of press releases, marketing materials, and other publicity items that often bog down traditional searches. These same tools may also unearth materials from unconventional sources that even the most rigorous search protocols would miss, thus potentially reducing false negatives in searches — for example, small foreign-language news sources that use local idioms or colloquialisms. 

Indeed, to further increase efficiency within your workflow, Thomson Reuters and WorkFusion have created a joint solution that integrates relevant, high-quality adverse media and sanctions data into first-of-its-kind technology that automates screening alert review through an AI digital worker.

That means institutions can gain significant, unrealized efficiencies and economies by taking a more precise, AI-driven approach. Importantly, there is a not-insignificant reputational win that financial institutions gain, especially in the eyes of their customers, by weeding out high-risk or potentially damaging customers. Customers and the general public look at these proactive processes positively — people want their banks and financial partners to do business with good customers.

As AI-driven tools continue to enable the industry to adopt a more proactive, risk-based approach, some financial institutions can use these efficiencies to move beyond a simple regulatory compliance mindset. Firms could instead routinely screen all clients in a category of concern rather than just those identified as high risk, thus averting a problem earlier rather than having to respond to it later. In other words, allowing the screening of low- and medium-risk clients — who make up the large majority of clients — and where adverse media may materially change their risk profile, it is better to be proactive and manage this risk appropriately before regulators or an audit identify it. 

Thomson Reuters CLEAR Adverse Media provides real-time connections to adverse media sources and extensive international coverage for sanctions, politically exposed persons (PEPs), and state-owned enterprises (SOEs). Using the platform, customers have the flexibility to start with an ad hoc or batch operating model suitable for overcoming backlogs, consent orders, and more.

However, many see any innovation and advancement in these areas stymied by an overly narrow focus on regulatory compliance rather than an investigative framework for impeding financial crime. This means that firms will have to deploy the tools that enable broader screening carefully; otherwise, they will only add to the costly and inefficient false positive problem.

How can institutions improve their search prowess?

So, what should a financial institution’s AI-driven search system contain to improve upon current practices? For one, it should be able to organize information for analysts along several parameters, including ratings on potential relevance, labeling of event or item type, chronology, region, source type, and more. Thomson Reuters CLEAR Adverse Media combines the best data with the best automation, which means that analysts only have to get involved when their expertise is necessary, minimizing false positives to review.

A dashboard-style set of results with these added guides — such as noting why the item was flagged — or editable filters marks a big advance from the kinds of delivery that currently characterize open-source search engine results. These more streamlined results, coupled with skilled human analysis, can help institutions address both the false-positive and the false-negative problem. 

Yet, professional organizations such as ACAMS are skeptical about whether the current regulate-and-fine structure is working well enough as an incentive to improve the quality of search and screen efforts on its own. Estimates of current adverse media screening system false positives are 90% and above; while “better safe than sorry” internal directives may drive the over-cultivation of adverse media hits, specialists in the AML area regard high false positives as a sign of weakness rather than strength. In fact, this over-cultivating strategy is seen as one that is likely to fail, often because these same imprecise processes also miss potentially vital false negatives and true positives.

Even those institutions that have tried to keep up by deploying newer screening technology should periodically review whether their current system is working for them. One of the best tests is whether analysts and compliance officers can explain to regulators and third-party reviewers how the search and matching algorithms work. “Regulators must no longer treat market-leading tools that create dangerous inefficiencies with deference,” ACAMS stated. “If a financial institution cannot explain how their tool’s matching algorithms work, then that tool is not safer than a more innovative alternative just because it is already deployed.”


As many financial institutions seek to implement quality screening tools efforts, especially in cases of adverse media searches, it is critical that they use the most up-to-date technology to aid analysts and compliance officers. 

By improving detection and reducing financial crime, banks and other financial services firms will be helping reduce instances of money laundering, drug smuggling, human trafficking, corruption, and embezzlement while exhibiting a more effective program. Furthermore, customers and society will see the results of these investigations as highly worthwhile. Research has shown that companies with improved corporate governance profiles enjoy higher shareholder value, higher equity returns, and a reduction in downside risk. 

Trust our best-in-class tools

Automate false-positive reviews, increase operational capacity, and manage risk by leveraging AI technology from Thomson Reuters and WorkFusion

Thomson Reuters and WorkFusion have partnered to bring together premium adverse media and sanctions data with leading AI and automation technology that simplifies many of the time-consuming, error-prone tasks related to screening, reviewing, and disposition of false-positive alerts. By utilizing this joint offering, your team of analysts will be able to automate reviews; make more confident decisions; increase operational capacity; and stay compliant with federal, state, and local regulations.

Thomson Reuters is not a consumer reporting agency and none of its services or the data contained therein constitute a “consumer report” as such term is defined in the Federal Fair Credit Reporting Act (FCRA), 15 U.S.C. sec. 1681 et seq. The data provided to you may not be used as a factor in consumer debt collection decisioning; establishing a consumer’s eligibility for credit, insurance, employment, government benefits, or housing; or for any other purpose authorized under the FCRA. By accessing one of our services, you agree not to use the service or data for any purpose authorized under the FCRA or in relation to taking an adverse action relating to a consumer application.