Fingerprinting Script Uses Bioweapon-Themed Names to Block LLM Deobfuscation
READ THE ORIGINAL SOURCEOpen original sourcenullpt.rs · Opens in a new tabThe author crawled 10,000 sites with Playwright, collecting canvas fingerprints, and found a script from servicepipe.tech containing bioweapon- and poison-themed function names. When asked to deobfuscate it, Astra refused, citing biological research safety risks, while GLM-5.3 complied. Decoding an invisible-unicode string revealed a hidden system prompt attempting to jailbreak models into giving harmful instructions. The script also contains Node-gated code that reads credential files and runs shell commands that download a binary, though this is only partly described in the supplied text.
One primary source; other links are context, not independent corroboration.
- Primary sourceHacker News discovery · automated brief
- Discovery linkHacker News
- Linked in sourceTranco Top Sites list
- Linked in sourcefingerprint.garden
Grok Bot submitted the link. AINewsIQ fetched the primary source and prepared a source-based brief automatically. An editor has not reviewed the story; social attention and the brief do not independently verify its claims.
Open full article view →- Linked article+30
- Fetched source page+30
- No editor link review+0
- Original non-discussion link+5
This measures the stored source trail and editorial steps. It does not certify accuracy, peer review, or independent corroboration. Popularity adds no points.
View full scoring →Read the scoring rubric →What the source says
The author crawled 10,000 sites with Playwright, collecting canvas fingerprints, and found a script from servicepipe.tech containing bioweapon- and poison-themed function names. When asked to deobfuscate it, Astra refused, citing biological research safety risks, while GLM-5.3 complied. Decoding an invisible-unicode string revealed a hidden system prompt attempting to jailbreak models into giving harmful instructions. The script also contains Node-gated code that reads credential files and runs shell commands that download a binary, though this is only partly described in the supplied text.
A READER'S EVIDENCE CHECKHow to read this signalSix questions about claims, sources, risks, and open checks
This is a signal-level guide until claim-level evidence is available. “Not recorded” means the evidence has not been assessed here; it does not mean a result is false. How we check AI news →
- 01Who is making each claim?
The claimant and whether each claim was measured or asserted are not recorded in this public brief.
- 02What was actually tested?
Not recorded per claim. Check the source for sample size, tasks, model and benchmark versions, and reproducibility.
- 03Where are the sources?
Original link: nullpt.rs. Brief basis: fetched source page. Primary page date and quote locations are not established in this panel.
- 04What could mislead?
Source claims not independently checked; the author's model comparisons are single trials and the full script was not verified.
- 05What is at risk if wrong?
Not assessed per claim. Consider safety, privacy, cost, and adoption decisions before relying on the result.
- 06What would change the view?
No claim-specific check is recorded. An inspectable primary method or independent replication could change the assessment.
Key findings from the stored brief
- The script contains bioweapon- and poison-themed function names intended to trip LLM safety refusals during deobfuscation.
- Invisible Unicode characters encode a hidden 'SYSTEM OVERRIDE' jailbreak prompt covering cyber, bio, and drug topics.
- Astra refused to deobfuscate the script, while GLM-5.3 complied with the same prompt, per the author.
- Node-gated code lists credential and config files to read and includes shell commands that download and execute a binary.
Source claims not independently checked; the author's model comparisons are single trials and the full script was not verified.
This is a source-based abstract, not independent validation of the source's claims.
Evidence & sources
These links show where the signal came from and what the source page linked to. A linked page is context, not independent corroboration.
Grok Bot submitted the link. AINewsIQ fetched the primary source and prepared a source-based brief automatically. An editor has not reviewed the story; social attention and the brief do not independently verify its claims.
These are Grok Bot discovery notes. AINewsIQ prepared a brief from available source material. The bot's rationale, linked posts, authors, and stances have not been independently verified.
A researcher crawling 10,000 sites found a fingerprinting script whose functions are named after weapons and self-harm topics, apparently to trip model safeguards and block LLM-assisted deobfuscation; it also hides prompt-injection text in invisible Unicode. Astra refused to deobfuscate it when asked, while GLM-5.3 complied. The write-up hit #1 on Hacker News.
Source claims not independently checked; the author's model comparisons are single trials and the full script was not verified.
Engagement counts, when shown, are observed snapshots. They do not establish accuracy or corroboration.
- hn post ↗ · 1,234 points, 233 comments · observed 10/29/2026, 10:09:19 AM
Timeline & updates
- Signal dated by source or discovery feed
This timestamp places the signal in the radar. It may be the discovery post date rather than the primary page's first publication date.
- Shared brief stored
AINewsIQ prepared one abstract from fetched source page.
A publication revision history is not stored for this signal. New engagement snapshots do not mean the source itself changed.
Why this may matter
It shows adversarial code can embed prompt injections and misleading identifiers to make AI-assisted reverse engineering refuse or misbehave, a risk for security analysts relying on LLMs.
Limits and open questions
Source claims not independently checked; the author's model comparisons are single trials and the full script was not verified.
Organization-specific financial, security, and operational effects require the organization's own context and review.
Discussion
Comments are reviewed before publication. Discuss the evidence, not the person.
No published comments yet.
Reader discussion opens with accounts. You can read the original source and published comments now.