A transcription company that caters to law enforcement, government departments, and Silicon Valley tech giants has resolved a security lapse after leaving a large amount of internal data, including customer transcriptions and their source materials, exposed to the web.
The data security lapse at San Francisco-based Tigerfish exposed transcriptions of police interviews, body-cam footage, and customers’ raw audio and video footage. The exposed data also included invoices of the company’s customers, which include police departments and prosecutors’ offices.
According to Anurag Sen , a good faith security researcher who found the data, the files were located on an Amazon-hosted (S3) cloud server that was openly listed and accessible to anyone via an easy-to-guess web address — no passwords needed.
The security lapse underscores how simple flaws or misconfigurations can result in sensitive information getting exposed online, especially as AI makes it easier to uncover weaknesses that can put companies and their data at risk.
Sen discovered the exposed data online in October and alerted the company to the issue, but did not hear back. Sen later provided details to TechCrunch in an effort to flag the incident with the company for resolution.
Sen, 30, is no stranger to TechCrunch. For the past decade, he has searched for sensitive datasets exposed online with the goal of notifying the companies and organizations that their information is open to the internet. Like many security researchers, much of his efforts are driven by a duty to secure the internet, and he describes his work as an “unpaid air-traffic controller.”
From his island home in Indonesia, Sen has uncovered sensitive U.S. military emails spilling from a compartmentalized Department of Defense cloud server; databases exposing reams of customer records ; the inner workings of a country’s national license plate surveillance system ; and public systems that inadvertently sent out streams of SMS codes used by people around the world to log into their online accounts.
Given Sen’s propensity for finding often highly confidential information online, we verified the source of the leak and the data, and contacted Tigerfish.
Soon after making contact, Tigerfish secured its exposed S3 cloud bucket.
“We’ve blocked all public access to the bucket you identified, including directory listings and individual files,” Tigerfish president, Adam Goldberg, said in an email. “We’ve also verified that all of our other S3 buckets are blocked to public access.”
“We’re continuing to review and strengthen our storage security more broadly,” added Goldberg.
The company said that Sen’s email disclosing the data leak was caught in its spam filter.
It’s unclear if the company has notified its customers. Tigerfish did not respond to a follow-up email.
Discoveries of exposed datasets have consumed a large part of Sen’s early cybersecurity career so far. He scours the web daily using automated scanners and specialist search engines to identify databases, cloud storage, or other sensitive information that likely shouldn’t be left unprotected online.
The modern web runs on absolute trust, he said, and operates under the notion that the world’s data is kept secure. But he described how his findings reveal the reality of how “fragile” the internet is behind the scenes.
Despite the recent focus on how AI is accelerating people’s ability to discover vulnerabilities, Sen said that many of the data leaks and exposures that he finds come down to basic human errors.
“An engineer spinning up a test cloud database on a Friday, disabling authentication to test an API, and forgetting to shut the door before heading into the weekend,” was just one example, he recounted.
Sen’s findings over the years highlight the ongoing problems caused by such security mistakes, just as the new wave of vibe-coding and AI-generated development is creating new security problems due to insecure and unchecked code deployed to the web.
Similar to Amazon’s recent efforts to make it difficult to make S3 buckets public without going through clear multistep warnings, newer vibe-coding platforms like Lovable and Supabase have also made changes to their platforms to prevent user mistakes.
“Everyone often fears shadowy hackers,” Sen said. But he argues that the real crisis is leaving sensitive systems and databases “wide open to the public internet without a password.”
When you purchase through links in our articles, we may earn a small commission . This doesn’t affect our editorial independence.
Zack Whittaker is the security editor at TechCrunch. He also authors the weekly cybersecurity newsletter, this week in security .
He can be reached via encrypted message at zackwhittaker.1337 on Signal. You can also contact him by email, or to verify outreach, at [email protected] .
2027 San Francisco Get an Early Peek at the Next TechCrunch Disrupt Disrupt, Startup Battlefield, and more will return next year, and you can be at the front of the line. Sign up for our waitlist
Microsoft CEO Satya Nadella explains why he asked 60K employees to quit Julie Bort
Amazon and Google will together own nearly 30% of Anthropic after its IPO Julie Bort
Google joins the tech outage trend by announcing it’s not having one Sarah Perez
FBI says Artex AI open source hacking tool used to breach networks in ‘under three hours’ Zack Whittaker
The tech products we enthusiastically recommend this holiday season Amanda Silberling
Where was Nvidia’s security team? Connie Loizos
Elon Musk, other tech execs sign petition calling on Nvidia to reinstate Jensen Huang Anthony Ha