3 ms·

You can imagine LLM fingerprinting to be part of future pentest workflows where they identify the model and know it's weaknesses and vulnerabilities etc...

• varispeed 1y ago
  Will be cat and mouse game, because models can be trained to mimic response of other models under such conditions to obfuscate.
  • dboreham 1y ago
    See also: browser user-agent field.