Fingerpoint Detector

Disclaimer

This site is the author's personal research project. Results are for research and reference only, and the author takes no responsibility for them.

LM Fingerpoint Detector (the website, the command line tool, the SDKs, and these docs) is a personal research project that the author built for their own use. It is public for research and reference only. The website shows a summary of this disclaimer on your first visit. By continuing to use it, you confirm that you have read and accept the terms below.

Website acknowledgment

The Simplified Chinese interface asks three comprehension questions about identity claims, use as evidence, and the author's responsibility. No answers are selected by default. All three must be correct before the user can explicitly confirm and continue. Escape and backdrop clicks do not bypass confirmation; the full disclaimer remains available. Mouse, touch, and keyboard input are supported.

Acknowledgment is stored only in the current browser. Simplified Chinese uses a separate, versioned record: acknowledging the old notice or another language does not waive these questions when switching to Simplified Chinese. The notice stays dismissed after confirmation, but returns if browser storage is cleared or unavailable. Other languages retain the original one-time notice.

This interaction reduces casual dismissal without reviewing the notice. It cannot prove that a person has read or understood it, and it does not guarantee the disclaimer's legal enforceability.

For research and reference only

  • A result is a statistical ranking within the reference bank, not an identity check. Confidence compares only the models in the bank, and a model outside the bank still gets a "closest" candidate.
  • Results can be wrong. Model versions, inference settings, prompts, channel processing, and the coverage of the bank all change the result.
  • A matching name does not prove that an API is genuine. A different match does not prove that an API is fake or has substituted its model, because some models have similar fingerprints.

No assessment of any provider

  • Every detection is configured and started by its user. This site does not test, list, or rank any relay, API provider, or channel on its own.
  • Neither this site nor the author decides whether any provider is genuine, endorses any provider, or accuses any provider of fraud.
  • Screenshots and exported images that others publish come from those publishers. The author cannot verify their request settings, their channels, or whether the images were edited.

Responsibility

  • Do not use a result as the only basis for an accusation, complaint, claim, or other dispute. To judge whether a channel is trustworthy, fix the request settings, repeat several rounds, and combine the result with other evidence.
  • The author takes no responsibility for results or for how anyone uses or shares them. The parties involved settle any resulting dispute themselves, and the author does not handle appeals or arbitration about the results for a particular channel. If the reference data or the algorithm is wrong, open a reproducible issue on GitHub.
  • Detection requests use the API key that you provide. You are responsible for their cost and for whether they follow the provider's terms.
  • This site is provided as is, without any express or implied warranty of accuracy, completeness, availability, or fitness for a particular purpose. To the extent permitted by applicable law, the author is not liable for any loss caused by using or being unable to use this site. The source code is released under the MIT license.

Trademarks

Model names, company names, and logos on these pages belong to their owners. This site is not affiliated with or endorsed by those companies.

Last updated on

On this page