Transcript
GBT for Business, introducing new connectors for cloud apps, and most importantly, unveiling how Chad GBT retrieves data from those connected apps. We're looking at a seismic shift in how Chad GBT operates in the commercial marketplace. Chad GBT's new connectors are designed to bridge the gap between your organization's internal knowledge sources and the Chad GBT platform, enabling you to harness the full potential of your data while maintaining baseline privacy and security that you've come to know. However, the data security and governance implications are somewhat concerning if these connectors and their use are left unchecked. So what's different about these connectors versus what we've had in the past? It's ironic and somewhat indicative of the future to talk about the old days when Chad GBT Enterprise was just introduced and announced in late 2023, less than two years ago. And the connected apps feature was announced in May 2024, yet here we are. Last year, OpenAI announced the ability to select files to upload directly from Google Drive and Microsoft OneDrive and SharePoint through connected apps. After establishing the connection, users could simply select a file from a dropdown or copy a shared link into the prompt chat box. This ability is still available and likely will continue in the future, despite the introduction of all these new connectors and a new connected experience. Nevertheless, security teams still can't see when connectors are enabled and by which admin unless they also go into the admin center and look at the settings themselves. They also can't natively see when sensitive data has been selected and uploaded within the Chad GBT interactions of their users with full context, like who the user was, what was the file name, what's the file type, classification, and more. There is a compliance API, which Varonis uses, but nevertheless, you can't natively see those things. This visibility gap persists in the new connectors and honestly gets a little bit bigger without a unified platform to monitor Chad GBT Enterprise holistically across all the connected experiences. So what about these new connectors? The prior method of connected apps is a little bit of a manual user experience, which is why Chad GBT and OpenAI is looking to expand how the user accesses internal knowledge and data. Last month, or June 2025, for whenever you're watching this, OpenAI announced the beta release of connectors as a new way to integrate with the third-party applications that you have in your organization. These connectors allow Chad GBT Enterprise to act a little bit more like a retrieval augmented generation or RAG application or AI agent. Rather than users manually selecting files, Chad GBT Enterprise connectors will enable the agent to basically retrieve the relevant data or information for the user. So you don't have to go find the right file in a dropdown. I simply can just ask, what were those latest blockers for our engineering team in the latest development spread for the new Shark LaserBeam program? It will then go find the relevant information from, let's say, Teams, Teams discussions from my meetings about the Shark LaserBeam program. It will then go pull files from maybe SharePoint or Box, where I'm sharing some of these files with my colleagues about the Shark LaserBeam program. It's going to go find all of that information, synthesize it for me, and answer my prompt, which again was, what were the latest blockers for an engineering team? This is a similar experience that you probably have come to know with Microsoft 365 Copilot and your users. Copilot searches all your sources in the Microsoft 365 tenant, like Outlook, Teams, SharePoint, OneDrive, et cetera, and infers what is accessible, what's relevant to you as the user from the Microsoft Graph. It retrieves all that and then services the right information in response to the prompts. The connectors can integrate with way more data sources beyond Google Drive and Microsoft OneDrive, like you could in the past with JADGBT Enterprise, and even beyond what you can do in the Microsoft 365 tenant, because like I said, I didn't list just Microsoft things. So, what are the connectors? Connectors will now include apps like Box, Dropbox, GitHub, so for developers, Gmail, Google Calendar, Google Drive, HubSpot, Linear, Microsoft OneDrive, Microsoft Outlook, Microsoft SharePoint, Microsoft Teams, and Custom Connections through an MCP server. Ultimately, JADGBT will be able to pull data from multiple sources simultaneously in the future to answer complex requests, like my example about the Shark LaserBeam program. The speed and efficiency with which users can retrieve the data they need and create new data will be unmatched, yet the risk of users accessing data they shouldn't compounds exponentially with each connector that's greenlit. Organizations will be looking for a holistic approach to ensure each connector does not open a massive hole in their data security strategy. So, what are the top challenges to consider? There are many risks associated with enabling connectors that your organization needs to weigh with each resource they greenlit. There are many risks associated with enabling connectors that your organization needs to weigh with each resource. Top five on the surface for me would be unauthorized access, data leakage, insider threats, compliance, and regulatory risk. And lastly, the sprawl of these connectors. Let's start with unauthorized access. The integration of multiple internal data sources increases the risk of unauthorized access. If the data associated with that connector is not properly secured, unauthorized users might gain access to sensitive information stored in the various platforms you have connectors for, like HubSpot, SharePoint, Box, and Teams. This could lead to data breaches and exposure of confidential information. Next is data leakage. With the ability to query and synthesize information from various sources, there's a risk of data leakage always. Sensitive data might be inadvertently shared or exposed through the connectors, especially if proper data handling and sharing protocols are not in place. This could result in the unintentional dissemination of proprietary or confidential information. Somewhat related is insider threats. Employees with access to the connectors might misuse their privileges to extract and share sensitive information for personal gain or to harm the organization. This risk is heightened if there are no robust monitoring and auditing mechanisms in place. Now let's talk about compliance. Organizations must ensure that the use of connectors complies with the relevant data protection and privacy regulations, such as GDPR, HIPAA, CMMC, or CCPA. Failure to do so could result in legal and financial penalties. The connectors must be documented and their risks assessed and their usage logged and monitored. The other challenge for AI security teams will come as more and more cloud solution providers and third-party apps race to be added. I call this sprawl. For instance, the HubSpot connector was built by the HubSpot developer team and was the first MCP custom connector published in the chat GPT registry from the June announcement. More and more data resources will be added and organizations should be prepared for a future where all major resources in their organization are available in the connectors admin settings. By addressing these potential risks through robust security measures, organizations can leverage the benefits of the new connectors while safeguarding their data and maintaining compliance with regulatory requirements. Let's go ahead and jump right in to what do we do about some of those challenges. OpenAI provides several security and governance features for admins out of the box as an initial foundation. For starters, connectors are disabled by default and can only be enabled by the workspace owner and admin. Also the same standards for data encryption provided for chat GPT enterprise session data in transit and at rest applies to chat GPT interactions with connectors. Data accessed via connectors will not be used for training any of OpenAI's models as well. Arguably the most important security feature of all is that chat GPT will only access and reference content based upon that user's permissions. Therefore, if you have implemented least privilege and are automatically remediating risky permissions across the data estate, chat GPT enterprise should not unintentionally surface sensitive data to the wrong users. Nevertheless, let me go over some best practices that I think are key to maintaining a data security program for chat GPT enterprise and functionality needed within your data security platform or DSP to operationalize them. Let's first discuss access control. The first access point is the admin configuring connectors in the first place with workplace settings and things of that nature. OpenAI mentioned that native RBAC is on the way, but you need a way to know when admin roles are assigned or changed along with when connectors are enabled. Each connector also requires additional configuration that your organization should control and monitor. Microsoft services, for example, like SharePoint teams require delegated application permissions through the Microsoft Graph. When those permissions are assigned, security teams should be aware. Second, let's talk about automated risk remediation. Chat GPT's blast radius will largely be determined by how well an organization can find and remove excessive permissions on files, sites, lists, mailboxes, and storage locations for each connected account. Managing permissions in multiple cloud locations and apps can be very daunting, if not impossible. Mature organizations rely on an AI-powered DSP or data security platform to classify all the data within each resource and automatically right-size permissions based on its sensitivity. Now that we've talked about remediating risk automatically in each of the different cloud resources and cloud apps, let's talk about monitoring the actual files and how they're being used. Knowing what files are being referenced and their metadata can be critical to understanding where risk or threats exist. If a single user atypically begins accessing multiple sensitive files in their sessions, security teams need to be notified. Additionally, spikes in sensitive data access can be an indicator of a major exposure at a site or maybe at a parent resource level. Fourth, prompt and response monitoring. Not just looking at the files and how they're being used, but what's the actual content of the interactions? Some instances of insider threats or compromised identities can be detected by the context of the prompts. Though a connector may not produce the results a bad actor wants, we can understand a tint of the user from the questions asked in chat GPT conversations. Also, data flows can produce sensitive results unexpectedly. A user may not have permissions to access certain customer data in one resource, like HubSpot, but still be shown that data because the information was exported to an accessible SharePoint site or shared in an email. That's a bad day. So we need to be able to monitor those prompts and responses to see some of that activity that may not necessarily be malicious, like an insider threat, but still dangerous to the organization. Lastly, in all of this, whether it be preparatory, like remediating policies and permissions on particular assets and resources, doing things proactively, and also looking at things from a monitoring perspective, we also need to alert on misuse. OpenAI provides a compliance API that is the only approved pathway for log data. It is the same API used by Varonis. It fuels intelligent alerting to notify security teams of any AI misuse and abuse. The ideal platform will also filter activity through evergreen AI models to basically reduce noise and false positives. Because alerting can be great, but if it's too much alerting or inappropriate alerting on things that just really don't matter, that can be a bad day for security teams. So we've talked about the connectors, some of the challenges, and we've also talked about ways that we could be prepared for enrolling and enabling some of these connectors. So right now they're in beta, but continue to think through and walk through these various steps that I've discussed and some of the mispractices in how you approach connectors in Chad GPT Enterprise. By doing so, hopefully you'll be able to enable Chad GPT Enterprise connectors safely and securely for your organization.
