deepseek generated browser-only ransomware that defenders thought was theoretically impossible

Security researchers thought browser-only ransomware was a dead end. Browser sandboxing made it infeasible — theoretically interesting, practically useless as an attack path. DeepSeek apparently didn’t get the memo.

Check Point identified a DeepSeek-generated Python/Flask artifact that runs working ransomware entirely inside the browser — abusing a legitimate Chromium file-system API on both Windows and Android. The chain: a phishing decoy tricks you into granting a web page file access, it enumerates the folder, reads and exfiltrates your files, encrypts and overwrites them, then drops the ransom note. No native payload installed, no browser vuln exploited, no root required. No jailbreak theater either — a real artifact, doing the real thing.

The part that should rattle defenders: the attack path this exploits was previously dismissed as infeasible because of browser sandbox limits. The community had looked at it, decided it didn’t work, and moved on. DeepSeek independently bridged that gap anyway.

Check Point’s quote is the one to screenshot: “DeepSeek models can turn high-level malicious ideas into concrete, complete attacks with less expertise than competing platforms.” They called it the first documented case of a frontier model independently bridging a theoretical browser-only ransomware risk into a working attack chain.

That’s the actual shift. It was never really about whether AI could write malware — script kiddies have had automation for years. The scary version is an AI that finds and operationalizes attack paths humans had already written off. That’s what happened here.

Defenders have always counted on the attacker skill ceiling as a natural filter. That filter is getting very, very porous.