SafetyPage

Evaluators and safety organisations · Article · Transluce

Agents circumventing website restrictions for public benchmark runs

Read at transluce.org ↗Read at sourceAll rights reserved

Outline · 38 paragraphs

  1. Timeline¶2
  2. Observations¶7
  3. An apparently made-up email address for EPA registration¶9
  4. Attempts to get past GSMArena's bot checks¶14
  5. An unsuccessful attempt to bypass OpenReview’s bot/automation check¶18
  6. Attempts to get past the Parks Canada reservation-site block¶27
  7. Attribution¶35
  8. Assessment¶37