A San Francisco-based transcription vendor that handles audio and video for law enforcement agencies, prosecutors’ offices and major technology companies left some of its most sensitive customer material sitting on the open internet, protected by nothing more than an easily guessed web address. The company, Tigerfish, only locked down the data after a security researcher’s warning went unanswered and reporters stepped in.
- What was exposed: Transcripts of police interviews, body-camera footage, the raw audio and video files behind those transcripts, and invoices belonging to police departments and prosecutors’ offices.
- How it happened: The files sat in an Amazon S3 storage bucket that was publicly listed and reachable without a password.
- Who found it: Security researcher Anurag Sen, who spotted the data in October 2026 and alerted the company, but received no reply.
- How it ended: Tigerfish blocked all public access to the bucket after being contacted by the press, and said its other S3 buckets were already closed to the public.
A vendor-layer failure with real-world stakes
The material involved is not routine commercial data. It covers recordings tied to criminal investigations, including the voices and faces of people caught up in legal proceedings. The distinction between a finished transcript and its source recording matters: the original audio and video can preserve tone, phrasing and details that never make it into written text.
According to reporting on the incident, the bucket was openly listed and accessible through a simple web address, with no authentication required. The available information does not specify how many files or clients were affected, how long the bucket remained public, or whether anyone besides the researcher and journalists accessed the contents. Exposure alone does not prove that files were downloaded or misused.
The case fits a familiar pattern of cloud storage misconfigurations, in which a single unchecked setting turns a private repository into a public one. Sen, who has spent roughly a decade hunting exposed datasets from his home in Indonesia and describes the work as unpaid air-traffic control, argues that most of these incidents stem from basic human error rather than sophisticated intrusion. His example: an engineer spinning up a test database on a Friday, disabling authentication to test an API, and forgetting to restore it before the weekend.
How the disclosure unfolded
Sen discovered the exposed files in October and notified Tigerfish, but his message never reached a human inbox. The company later said the email had been caught in its spam filter. After failing to get a response, Sen passed the details to journalists, who verified both the source of the leak and the nature of the data before contacting the firm.
Tigerfish moved quickly once reporters were involved. Company president Adam Goldberg confirmed the remediation in writing:
We’ve blocked all public access to the bucket you identified, including directory listings and individual files. We’ve also verified that all of our other S3 buckets are blocked to public access.
Goldberg added that the company is continuing to review and strengthen its storage security more broadly. What remains unresolved is customer notification. It is unclear whether Tigerfish has informed the police departments, prosecutors’ offices and corporate clients whose material sat in the bucket, and the company did not respond to a follow-up inquiry.
The uncomfortable lesson about invisible intermediaries
Ask who holds the recordings of a police interrogation and most people picture the department or the prosecutor’s office. In practice, a copy often lives with an outside transcription contractor, which may in turn store it in a cloud bucket configured by an engineer under deadline pressure. That is the structural blind spot this incident lays bare: the most sensitive material in the judicial system gets replicated across third-party infrastructure that neither the public nor, frequently, the client organization can see or audit.
The technical fix here took minutes. The harder repair is procedural. A disclosure channel that routes a good-faith warning into a spam folder, and a remediation timeline that begins only when a newsroom calls, suggest that security monitoring and inbound vulnerability reporting are not yet treated as operational priorities at firms handling law-enforcement data. Cloud providers have already responded to this class of error by adding multi-step warnings before a bucket can be made public, and newer development platforms are introducing similar guardrails. Those measures help, but they do not substitute for contractual requirements that vendors prove their storage is locked down, disclose incidents to clients promptly, and maintain a monitored path for outside researchers to report problems. Until procurement processes at police departments and prosecutors’ offices demand that level of accountability, the weakest link in the privacy chain will keep being the vendor nobody thought to ask about.
Sources & References
- A transcription firm used by police, prosecutors, and tech giants exposed customer data online
- Tigerfish expuso transcripciones policiales y datos de clientes por una falla en la nube
- Tigerfish Exposed Police Transcripts in Open S3 Bucket
- A transcription firm used by police, prosecutors and tech giants exposed customer data online
- A transcription firm used by police, prosecutors and tech giants exposed customer data online | daily.dev