A transcription company serving law enforcement agencies, government departments, and major technology firms fixed a security breach after it left a substantial volume of internal information exposed on the internet, including customer transcriptions and their original materials, according to TechCrunch. The security failure at San Francisco-based Tigerfish allowed public access to transcriptions of police interviews, body-camera recordings, and clients' raw audio and video files. The exposed information also contained invoices identifying the firm's customers, among them police departments and prosecutors' offices.

Security researcher Anurag Sen, who discovered the breach, found the files sitting on an Amazon-hosted S3 cloud server that was openly listed and reachable by anyone through an easy-to-guess web address, with no password protection required. Sen uncovered the exposed information in October and alerted Tigerfish to the problem but received no response. He later shared details with TechCrunch to help flag the incident for resolution. After TechCrunch contacted the company, Tigerfish secured the exposed cloud storage. Company president Adam Goldberg confirmed in an email that the firm has "blocked all public access to the bucket" and verified that all other S3 buckets are also blocked from public access. Tigerfish blamed a spam filter for catching Sen's initial disclosure email, and it remains unclear whether the company has informed its customers of the breach.

The incident highlights how straightforward errors or system misconfigurations can lead to sensitive information becoming publicly accessible online, particularly as AI makes it easier to find weaknesses that threaten companies and their data. Sen, a 30-year-old researcher working from Indonesia, has spent the past decade hunting for sensitive datasets left unprotected on the internet, describing his unpaid efforts as being like an "unpaid air-traffic controller." Despite growing attention to how AI speeds up vulnerability discovery, Sen said many of the data leaks he uncovers stem from basic human mistakes, such as an engineer setting up a test database on a Friday, turning off authentication to test an API, and forgetting to restore security before the weekend.

Sen's work over the years reveals the continuing challenges created by such security lapses, just as the new generation of vibe-coding and AI-generated development introduces fresh security risks from insecure and unreviewed code pushed to production. The modern web operates on absolute trust and assumes the world's data stays secure, but Sen's discoveries show how "fragile" the internet really is behind the scenes. Sen argued that while "everyone often fears shadowy hackers," the genuine crisis involves leaving sensitive systems and databases "wide open to the public internet without a password." Amazon has recently made it harder to create public S3 buckets without navigating explicit multistep warnings, and newer coding platforms like Lovable and Supabase have introduced similar safeguards to prevent user errors. The push for better defaults and clearer warnings reflects an industry reckoning with the reality that convenience and speed can't come at the expense of basic data protection, especially when police evidence and customer information hang in the balance. As companies lean more heavily on cloud infrastructure and automated tools to handle sensitive work, the gap between what security experts know is possible and what everyday engineers remember to do in production keeps widening, turning every misconfigured bucket into a potential headline.