Customer records exposed: 24 million (Source: Infosecurity Magazine, citing Gyazo disclosure (September 11, 2026))
What happened
Gyazo, a screenshot-sharing service based in Japan, disclosed on September 11 that intruders had exploited a flaw in one of its upload servers, pulling roughly 24 million customer records out of the system, Infosecurity Magazine reported. Developer Helpfeel, which builds Gyazo, published its own account of the incident five days later, on September 16.
The metadata went further than the accounts
Helpfeel's disclosure also covered a second, larger dataset: 490 million records describing the images themselves rather than the accounts that uploaded them. The fields varied widely — unique image identifiers, the IP address and browser signature attached to each upload, GPS coordinates pulled from photo metadata, text the service's OCR engine had lifted off screenshots, titles, source URLs, and passphrases stored as hashes rather than plain text. Helpfeel said the exposed metadata could let a third party construct image URLs and view images without authorization, and it temporarily disabled viewing of some images as a result.
Why this matters beyond consumer accounts
Suzu Labs founder and CEO Michael Bell pointed to how developers actually use the tool — capturing terminal sessions, API keys, credentials sitting in configuration files, and screenshots of internal applications — as the reason the exposure carries weight beyond individual consumers. The service's OCR feature, which makes captures searchable, extracted and stored that text alongside the images, turning what was once pixels into indexed, searchable data. Seemant Sehgal, CEO at BreachLock, said the metadata layer carries the broadest reach of the incident, arguing that EXIF coordinates, OCR-extracted text, and session and URL-construction data give an attacker material to reconstruct behavior and location history for people who uploaded a screenshot and did not expect it to resurface.
A mitigating factor: the age of the data
Not every assessment of the incident was as alarmed. Filigran senior solutions engineer manager Damian Skeeles noted that the leaked metadata traces back to images uploaded no later than January 2019, putting a rough age on how stale the exposed data actually is. He said that age reduces the practical impact of any credentials captured in old screenshots, provided organizations cycle credentials on a regular schedule rather than treating a one-time rotation as sufficient.
Response and the broader lesson
Helpfeel said it has remediated the vulnerability used in the attack and is urging affected users to change passwords on Gyazo and any other service where credentials were reused, noting the stored passwords were hashed. CyberSmart CEO Jamie Akhtar said the incident reflects a broader operational pattern for internet-facing services: patch and test rigorously, restrict what upload systems can access, monitor continuously for suspicious behavior, and after a breach, invalidate sessions and tokens and force password resets. Separately, Comparitech consumer privacy advocate Paul Bischoff warned that the leaked email addresses and other identifying details give scammers raw material for phishing attempts that pose as Gyazo or a partner company.
What to ask your team
For a budget-holder, the relevant question is not whether Gyazo specifically is in use, but whether any screenshot, annotation, or OCR-enabled SaaS tool is embedded in developer workflows without a data-retention or credential-hygiene policy attached to it. Ask your team which cloud-based capture or annotation tools are approved for use, whether OCR-searchable text from screenshots is treated as sensitive data, and how quickly credentials exposed in shared images get rotated.
Produced by the WebPulse Newsroom with AI assistance from the original reporting credited below, and checked against that source by our editorial review. How we use AI.
Original reporting: Infosecurity Magazine.





