Why Is This Fingerprinting Script Creating Bioweapons?

A browser-fingerprinting researcher crawled 10,000 popular websites with Playwright, recording canvas-rendering activity and other signals used by fingerprinting and anti-bot systems. The resulting collection, fingerprint.garden, contained 1,033 distinctive canvases, including examples from familiar vendors and a striking design reminiscent of Tame Impala’s Currents album cover. Tracing that image led to a JavaScript file served by ServicePipe. Its obfuscated code initially looked routine, but contained function names and output strings about ricin doses, suicide, explosives, and other dangerous topics - suggesting deliberate attempts to make AI tools refuse to analyze the script.

The researcher decoded hidden Unicode text that posed as a system override, demanding instructions for cyberattacks, tax evasion, bioweapons, and illicit drugs, alongside political censorship tripwires. The script also contains Node.js-only commands behind an environment-variable check: they appear to inspect configuration and system data, download and run a remote “metrics agent,” package and transmit collected evidence, and establish a recurring connection. Yet the researcher questions whether this is genuine malware or another layer of anti-AI deception: some file reads use empty callbacks, and commands write jailbreak and self-harm prompts into marker files. The investigation’s central puzzle is whether the alarming code represents an actual intrusion tool, or intentionally fabricated behavior designed to confuse automated analysis.