HTML Preview Flaw Let Attackers Bypass ChatGPT Lockdown Mode, Researchers Reveal

Cybersecurity researchers have disclosed details of a now-patched vulnerability that could be exploited to bypass Lockdown Mode protections in OpenAI ChatGPT, according to thehackernews.com. The flaw, discovered by Check Point Research, allowed an attacker to get around the optional setting OpenAI introduced in June 2026 to limit data exfiltration risks from prompt injections. Protect.Computer reported that Lockdown Mode restricts web access to cached content, blocks loading images from the internet, disables network requests from Canvas-generated code, and turns off features such as Deep Research, Agent Mode, live connectors and file downloads. Check Point’s researcher Alexey Bukhteyev found that a different feature, the HTML page preview that OpenAI added to ChatGPT code blocks in February 2026, offered a way around those limits, Protect.Computer noted. WebPulse reported that a convenience feature opened a path around a security setting, and OpenAI says it has fixed the flaw. IT Nerd explained that Check Point researchers didn’t beat the AI controls, but found a way to go around them, with a weakness in an ordinary app feature, HTML previews running in a sandbox, letting content escape the restrictions. "The flaw could enable silent chat data theft, phishing" Xploitwire reported that Check Point found nested sandboxed frames let crafted HTML escape ChatGPT's Lockdown Mode, and OpenAI patched it on September 18, 2026. Notebookcheck.net added that the attack required no user interaction in the ChatGPT Classic desktop app, and the report highlights risks such as data leakage, fake sign-in windows, and malware distribution. The vulnerability underscores how even security-focused features can be undermined by seemingly unrelated convenience tools, a lesson for both developers and users relying on AI platforms for sensitive work.

7 sources•7 NovRead the Baker's Take→