By Miro Bocchino, Network Partner, Ground Truth Intelligence
What does the launch of Google’s Search Generative Experience mean for investigators? It’s a significant change and yet another leap forward in accessing data. But as an intelligence analyst, I’m naturally cautious – always looking for the most reliable and defensible way to help my clients make the best decisions. In this post, I take a look at the current drawbacks of using Google as a research tool and how this compares with the new functionality of generative AI.
To Google, or not to Google
Google Search was never designed for investigative due diligence. Still, it has become a commonly used search engine for desktop research both by in-house compliance teams and external service providers. It’s global, has no language restrictions, and is free to use for all despite covering only a fraction of the internet’s content. An expert investigator’s skills are needed to understand the workarounds and techniques required to drown out the noise and conduct a complex investigation. It’s essential to remember that modifications in Google’s functionality can significantly affect the investigation process. For instance, the recent removal of Google’s cached site links serves as a reminder that it is crucial to stay updated with Google’s changes to avoid missing critical information while conducting desktop research.
Furthermore, the lawsuit against Google’s “incognito mode” has raised new concerns about the reliability of Google search for investigative purposes, especially for seasoned investigators who generally resort to other search tools to avoid detection. But this should not deter investigators from the potential uses of this search tool. Instead, protecting privacy while utilising Google Search’s resources is essential. With careful consideration and responsible use, users can continue to benefit from this search engine while closely monitoring developments in generative AI.
Here Comes Generative AI
Automation has become increasingly prevalent in the field of investigative due diligence. By automating repetitive tasks, investigators can focus on more complex and high-level assignments, which improves productivity and significantly reduces turnaround times. This approach ensures that the quality of work remains high while making the process more cost-effective. Automating repetitive tasks is a smart way to achieve greater efficiency and productivity in research work. Yet, generative AI, an AI model that can create content, is predicted to transform the future of the search experience completely.
Despite technological changes, it is crucial to remember that privacy and security concerns should always be the top priority during investigations. Leaving a digital trail could compromise the confidentiality between companies and their clients. For instance, the latest version of ChatGPT can store users’ chats to facilitate future conversations. If generative AI saves search inputs during an inquiry, even briefly, to improve customer experience based on their preferences, it could pose confidentiality concerns. Therefore, it is critical for companies to carefully choose technology that aligns with their specific needs and make any necessary adjustments to meet their internal requirements.
Google is currently testing the new Search Generative Experience (SGE). It’s important to note that SGE is currently being tested for personal accounts in limited locations. For example, SGE testing is yet to be available in the EU and UK. Nonetheless, it’s expected that SGE will eventually be rolled out globally and become a significant part of the search experience.
What If I Am Not Famous?
SGE’s snapshot feature is practical when researching high-profile individuals and larger companies. Providing concise and easy-to-read summaries becomes an invaluable tool for quickly obtaining information.
SGE’s effectiveness diminishes when searching for a target with low media coverage. Unfortunately, AI-generated results tend to cast a wide net, resulting in a plethora of false positives and irrelevant information.
SGE can be helpful for investigators, allowing them to obtain snapshots of information quickly based on their search inputs. Nevertheless, conducting due diligence requires finding information that may need to be more readily available. The joke goes, “The best place to hide a dead body is page two of Google”. For instance, high-profile targets such as celebrities, billionaires, and top 100 large brands are rarely subjects of third-party enhanced due diligence processes. Therefore, SGE may have significant limitations when conducting desktop research on low- and medium-profile subjects, such as smaller firms and associates, as the snapshots provided by the tool can quickly become irrelevant or yield no results. If Google Search was not designed for risk management purposes, then SGE is unlikely to be tested to meet the requirements of investigators.
Chatbots function like predictive tools that build a model to anticipate the most likely words or sentences that would follow a user’s input. Relying on internet-derived data to train chatbots may lead to inaccuracies that can mislead an investigator. Even though chatbots can generate responses that appear convincing to users, they may not always be correct.
Moreover, learning algorithms can be biased, and they may even amplify the existing biases in the data, further affecting the accuracy of the chatbot’s responses. Providing accurate, reliable, and objective information is crucial in due diligence. However, the current state of generative AI and SGE could pose a significant challenge in achieving these objectives.
Asking About What I Do Not Know
Make an enquiry with SGE about accessing local court records in Finland and it will provide helpful information on the availability of relevant data and how to obtain it. Although this may appear significant, it is essential to remember that clients, especially multinational corporations with extensive supply and distribution networks operating worldwide, rely on investigators’ expertise to conduct due diligence work in high-risk jurisdictions where information is difficult to obtain. For example, in places like Russia or China, investigators rely on their local knowledge, human intelligence (so-called boots-on-the-ground), and lateral thinking to overcome the challenges of limited information access. It is improbable that SGE alone will overcome the obstacles of stringent censorship policies and secretive jurisdictions prevailing in specific global locations.
Keywords Matter! So What?
Investigators use specific search strings to narrow down their results and eliminate irrelevant information when searching online. However, the search experience on SGE can be inconsistent, even when using standard Boolean operators. The placement and choice of keywords and Boolean operators can significantly impact the search outcome. This is also a common issue with standard Google searches. The main difference is that the SGE option may or may not be accessible, depending on the structure of the search query. This limitation arises from the fact that SGE is not designed according to the methods utilised by investigators during the due diligence process. SGE offers rapid summaries and automated follow-up queries based on the text provided by users in the form of questions, statements, or request inputs.
Is SGE Trying to Tell Us Something? And What’s Next?
Ask SGE to search for information on directorship positions and court records in specific jurisdictions! The caption below is impossible to miss.
This is not legal advice. You may want to consult a lawyer about this question.
As with the traditional Google search, this new generative search experience may not suit seasoned investigators, as it indirectly suggests it’s not tailored for investigative purposes. It may benefit new joiners and in-house practitioners who require training or want to broaden their knowledge, but it is yet to be determined whether experienced investigators will encounter significant differences in their search process while carrying out investigative due diligence in high-risk jurisdictions where data is scarce. Ultimately, it is yet to be seen how practitioners in the field of due diligence will balance the need for confidentiality against the data retention policies that SGE and other generative AI search tools will implement.