A security researcher known as BobDaHacker disclosed in early August 2026 that tl;dv, an AI meeting assistant claiming more than two million users, had left a core database collection unprotected for at least seven months, allowing any authenticated account on the platform to enumerate and read meeting records belonging to every other customer on the service. The exposed collection held 181,874 meeting records across 84,312 users and 35,003 domains. The exposed.gov domains alone pointed to meetings convened by governments of 23 different countries, including the United States, Japan, Brazil, Ukraine, Malaysia, Qatar, and Israel. Other meeting records originated from the University of Berkeley and the University of Tokyo, along with corporate domains including HubSpot, Confluent, and Mekari.
The technical failure was precise and, in retrospect, preventable. Firestore requires an explicit security rule on every collection defining who can query what, and tl;dv had correctly written those rules for nearly every collection in its schema, including users, chats, transcripts, and recordings, all of which returned a 403 Forbidden when queried across tenants. The meetings collection was the exception. A normal tl;dv account could obtain a Firebase authentication token and query that collection, which included the meeting creator's email address, conferencing provider, timestamps, recording status, and a conference identifier usable to reach the underlying Google Meet or Microsoft Teams room. The issue did not require a stolen administrator account, malware, or a flaw in Google Cloud itself. The researcher went further than pulling metadata: using a conference ID retrieved from Firestore, BobDaHacker joined a live Google Meet belonging to the Malaysian Ministry of Education, where a presenter was addressing more than 157 participants. For private calls, impersonating an AI notetaker and requesting to join the meeting typically worked, with the researcher reporting admission roughly 80 percent of the time. A secondary finding compounded the picture: an internal employee-facing application at worldcup.tldv.io exposed an unauthenticated API endpoint that returned all 43 listed players' data, including 19 tl;dv employees' full names and corporate email addresses.
The disclosure timeline is as consequential as the flaw itself. BobDaHacker reported the issue on January 28, 2026. Dark Reading reported that the exposure remained active when its article was published on August 4, 2026. The researcher said the issue was first reported to tl;dv in January, followed by repeated messages through July. Dark Reading said its own attempts to reach the company received no reply before publication. tl;dv published a response on August 5, stating that a vulnerability affecting access to certain meeting metadata had been identified through both a routine assessment by its penetration testing vendor, Abicom, and the researcher's responsible disclosure, and that a remediation had been deployed and formally validated by Abicom. The company added, however, that it had recently become aware of a newly discovered, alternative exploitation path related to the same part of its technology stack, which observers read as an acknowledgment that the patch did not fully close the window BobDaHacker documented through late July. Critics also noted that tl;dv holds SOC 2 certification, a compliance standard that the incident called into sharp question.
The case has sharpened a broader industry debate. Legal analysts at Mayer Brown noted in June 2026 that the normalization of AI notetakers has largely outpaced legal and compliance scrutiny, and that for meetings involving privileged communications, organizations should evaluate whether the use of third-party AI notetakers creates an unacceptable risk of privilege waiver. Privacy law specialists have flagged that most organizations buy rather than build AI technologies, procuring them from vendors including notetaking tools whose systems depend on large-scale data ingestion, a supply-chain dependency that existing SaaS security frameworks were not designed to govern at the scale and sensitivity that bulk-conversation data represents. A recurring thread in public discussion of the disclosure was a broader loss of trust in cloud-hosted meeting recorders generally, with commenters pointing toward local-only transcription tools that never send recordings or transcripts to a third-party server as an emerging alternative category.
The tl;dv case is the clearest demonstration yet that an AI notetaker is not a productivity add-on but a bulk conversation database, and that one missing line of database configuration can hand that database to anyone with a free account.